Immune Biomarkers In Peripheral Blood In Patients With HER2 Negative Advanced Breast Cancer Treated With A Chemoimmunotherapy Schedule Based On Pembrolizumab And Gemcitabine. – Translational Substudy Of The PANGEA- Breast Cancer Trial (GEICAM/2015-04)

TESIS DOCTORAL Immune Biomarkers In Peripheral Blood In Patients With HER2 Negative Advanced Breast Cancer Treated With A Chemoimmunotherapy Schedule Based On Pembrolizumab And Gemcitabine. – Translational Substudy Of The PANGEA- Breast Cancer Trial (GEICAM/2015-04) ISABEL ARAUJO FERNANDEZ Sevilla 2021 II III TESIS DOCTORAL Immune Biomarkers In Peripheral Blood In Patients With HER2 Negative Advanced Breast Cancer Treated With A Chemoimmunotherapy Schedule Based On Pembrolizumab And Gemcitabine. – Translational Substudy Of The PANGEA- Breast Cancer Trial (GEICAM/2015-04) Biomarcadores Inmunológicos en Sangre Periférica en Pacientes con Cáncer de Mama Avanzado HER2 Negativo Tratado con un Esquema de Quimioinmunoterapia basado en Pembrolizumab y Gemcitabina - Subestudio Traslacional del Ensayo Pangea-Breast (GEICAM/2015-04) ISABEL ARAUJO FERNANDEZ DIRECTORES Dr. Luis De La Cruz Merino Dr. Víctor Sánchez Margalet Programa de Doctorado Biología molecular, Biomedicina e Investigación Clínica Sevilla 2021 IV V ‘Rien dans la vie n'est à craindre, tout doit être compris. C'est maintenant le moment de comprendre davantage, afin de craindre moins’ Marie Curie VI VII A mi madre, que cultivo en mí las ganas de aprender Source: Modified from Instagram seen at @amandaoleander VIII IX TABLE OF CONTENT TABLE OF CONTENT ..............................................................................................................................IX ABSTRACT .......................................................................................................................................... XIII 1. INTRODUCTION ............................................................................................................................. 1 1.1. INTRODUCTION TO THE IMMUNE SYSTEM ................................................................................ 1 1.1.1. THE IMMUNE RESPONSE ....................................................................................................... 2 1.1.2. THE COMPONENTS OF THE IMMUNE SYSTEM ....................................................................... 3 a. Immune System Cells ............................................................................................................................. 3 b. Immune System Organs and Tissues ...................................................................................................... 8 1.1.3. THE ACTIVATION OF T CELLS AND THE IMMUNE SYNAPSE .................................................... 9 1.2. CANCER AND THE IMMUNE SYSTEM ....................................................................................... 12 1.2.1. TUMOUR CELLS CHARACTERISTICS ...................................................................................... 12 1.2.2. THE THEORY OF IMMUNE SURVEILLANCE ........................................................................... 13 1.2.3. IMMUNE ACTORS IN THE ANTITUMOR RESPONSE .............................................................. 14 1.2.4. CANCER IMMUNOEDITING .................................................................................................. 15 1.3. CANCER IMMUNOTHERAPY .................................................................................................... 17 1.3.1. IMMUNOMODULATORY MONOCLONAL ANTIBODIES ......................................................... 17 1.4. BIOMARKERS IN CANCER IMMUNOTHERAPY .......................................................................... 21 1.4.1. IMMUNE BIOMARKERS ....................................................................................................... 22 1.4.2. IMMUNE BIOMARKERS IN PERIPHERAL BLOOD .................................................................. 24 a. Myeloid-Derived Suppressor Cells ....................................................................................................... 25 b. Regulatory T cells ................................................................................................................................. 28 c. Other examples .................................................................................................................................... 30 1.5. BREAST CANCER ...................................................................................................................... 31 1.5.1. EPIDEMIOLOGY AND GENERAL CHARACTERISTICS .............................................................. 31 a. Incidence .............................................................................................................................................. 31 b. Prevalence ........................................................................................................................................... 32 c. Risk factors ........................................................................................................................................... 33 d. Mortality and morbidity (natural history) ............................................................................................ 34 e. Classification of Breast Cancer ............................................................................................................. 34 1.5.2. HER-2 NEGATIVE ADVANCED BREAST CANCER .................................................................... 35 2 1.1.1. The Immune Response The immune response is a coordinated reaction between cells and molecules from both innate and adaptive immunity. Therefore, it is generally considered that there is a close interaction between the components of both types of immunity3. The immune response is normally triggered by a danger signal that could be divided into three phases: early, intermediate and late. The innate immunity leads to a quick response that is involved in the early and intermediate phases whilst the adaptive response is becoming operational. The innate response is based on humoral mechanisms, based on antibodies production, and cellular mechanisms, involving cytotoxic or regulatory actions. Its activation lead to an inflammatory response3. The main cells involved in the adaptive immune response are B and Tcells lymphocytes participating in both humoral and cellular immunity. The adaptive immunity is involved in a later phase of the immune response 3. Normally, lymphocytes are circulating around the body in a virgin state. Only if they meet a recognisable antigen, they can be activated. In the case of no antigen encounter, these cells have a short live and are eliminated by cell programmed death3. B lymphocytes are able to recognise the antigen in a virgin state, whereas T lymphocytes can only recognise antigens when they are introduced by major histocompatibility complex (MHC) molecules. These MHC molecules are displayed on the surface of all the cells with endocytic and/or phagocytic activity, known as antigen-presenting cells (APC)4,5. Once activated, they start to proliferate (clonal expansion) and can differentiate into effective or memory cells. The effective cells will participate in the immediate response triggered after antigen exposition, and due to the memory cells generated after this first effective response, any subsequent exposure to the same antigen will trigger a more powerful, durable and faster response (secondary response)2. 3 1.1.2. The Components of The Immune System As previously discussed, the immune system involves a set of different organs and tissues that are interconnected through the circulation of different immune cells and a variety of proteins and molecules, such as antibodies or cytokines. a. Immune System Cells All the immune cells are generated from a common precursor, the haematopoietic stem cells (HSC), by process known as haematopoiesis. Figure 1. Cellular differentiation during Haematopoiesis This diagram summarises main cell lineage in the body and shows how all of them derived from a common precursor. Source: Version 8.25 from the Textbook OpenStax Anatomy and Physiology. Published May 18, 2016 4 Haematopoietic stem cells are able to self-renewal and to differentiate into precursors. The differentiation is mostly led by the proportion of different cytokines in the environment. Cytokines are a heterogeneous group of proteins of low molecular weight that are secreted by or membrane-bounded to cells. They play an important role in the induction and regulation of the immune response by regulating the growth, differentiation and activation of immune cells acting in an autocrine, paracrine or endocrine fashion6. Among others, main groups of cytokines include colony-stimulating factors (CSFs), tumour necrosis factors (TNFs), interferons (IFN), interleukins (IL)7. There are two main families of precursors: 1) originated from myeloid stem cells such as granulocytes, monocytes, macrophages and dendritic cells; 2) originated from lymphoid stem cells such as B, T and Natural Killer (NK) lymphocytes3. Depending on their roles and functions, these specialised cells can participate in the innate or in the adaptive immune response. The cells involved in the innate immune response are mostly myeloid and NK cells. These cells have a phagocytic capacity to eliminate any foreign molecule and altered or infected body cell3.  Granulocytes. These are the most numerous types of cells in the blood stream. They are characterised by the presence of multiples granules in the cytoplasm that contain elements to accomplish their activities. There are three different subtypes of granulocytes: o Neutrophils. They constitute the largest population of granulocytes in blood. Basically, they are phagocytes that play an active role in antimicrobial response and against acute inflammation. Their granules contain, multiple enzymes, antiseptics and vasoactive substances3. o Eosinophils. They can be mainly found in tissues and their main role is crucial in the anti-parasitic response and some hyper sensibility 5 reactions7.Their granule contain cytotoxic and pro-inflammatory substances. o Basophils. They have an anti-infectious role. Their granules contain acid cytotoxic and pro-inflammatory substances. Mastocytes are their equivalent in tissues. Both, basophiles and mastocytes have an important role in hyper sensibility diseases7.  Monocytes. Monocytes are released from the bone marrow to the peripheral blood where their median live is around 2 – 3 days. As granulocytes, they present cytoplasmic granules that contain different enzymes. Due to chemo attractive factors, circulating monocytes migrate to tissues where they can be differentiated into macrophages3.Macrophages are mainly phagocytic cells that play an important role in the elimination of apoptotic neutrophils, death cells and cellular debris, thus, favouring the resolution of inflammatory processes and the return to homeostatic equilibrium. Furthermore, they harbour cytotoxic properties and can produce huge amounts of pro-inflammatory or anti-inflammatory cytokines, depending on the type of environment where they are located3. In some circumstances, they can play a role as APC. In that case they can differentiate into dendritic cells that are able to activate naive T lymphocytes3.  Dendritic cells (DC). They are considered as they are the most important APC. They are specialised in capture, transportation, preparation and presentation of the antigens to the lymphocytes T, being the only type of cells capable to stimulate naive T lymphocytes. They are ubiquitous cells that can be present in any body compartment. They migrate from the sites where the antigen was captured to locations where the immune responses can be triggered. These cells are considered the main bridge between the innate and the adaptive immunity. In addition, DC constitute a heterogenic group of cells that differ in phenotype and functions8.  NK cells. These lymphocytes participate in the innate response. Their main role relies on the destruction of abnormal (neoplastic or infected) 6 cells. Moreover, they express a killer activation receptor (KAR) which can recognize microbial antigens. In addition, they can amplify the adaptive immune response as they have a high capacity for cytokine secretion, like IFN-γ9. Main cells intervening at the adaptive immune response are B and T cells.  B cells. These cells are originated and maturated in the bone marrow and express the B cell receptor (BCR). BCR possess the ability to bind circulating antigens directly, without the need of APC. B lymphocytes are responsible of the humoral immune response, being specialised in the production of immunoglobulins (antibodies).The immunoglobulins are glycoproteins synthesized by B cells or plasmatic cells in response to an antigenic stimulus. Their immunogenicity can be passively transferred, if they are ever administered to any subject. 10 B cells are also able to play an APC role to T cells. This constitutes the basis of the cellular cooperation between B and T lymphocytes5,7.  T cells. These cells mature in the thymus and express a membrane receptor, the T cell receptor (TCR). They are responsible of the cellular immune response. Regarding its function, T cells can be classified as: i) T helpers, once activated they segregate cytokines that lead to the activation of other immune cells, ii) T cytotoxic, they can eliminate abnormal cells from our own body, as infected or cancer cells, and iii) T regulators (Treg), they participate in the immune suppression5. As previously discussed, the TCR (CD3) can only bind to antigens when they are presented as peptides by the MHC molecules. In human and big vertebrates, MHC is called human leukocyte antigen (HLA). There are basically two types of HLA molecules: i) Type I (HLA I), they present endogenous peptides and are present in the cell membrane of all nucleated cells and platelets. 7 ii) Type II (HLA II), they introduce exogenous peptides and are present in the surface of all APC4. The TCR receptor has two chains. Regarding the type of the chains (γδ or αβ) there are two types of receptor; i) TCR-1,with chains γδ, and ii) TCR-2, with chains αβ. The type of surface receptors lead to a difference in the type of HLA molecules that activate the different type of T cells5. T cells can only express one of these two types of receptors. The majority (95%) of the T lymphocytes express TCR-2 (T-αβ lymphocytes). These are the classical T cells that can respond to pathogens and tumours. Phenotipically, they are marked by coreceptors; CD8 and CD4. CD4+ T cells can recognize the antigens when presented by type II HLA molecules and has a helper function(T helper cells), whilst CD8+ T cells show a cytotoxic function (T cytotoxic cells) when presented by type I HLA molecules5. Less of the 5% of T lymphocytes express TCR-1 (T-γδ lymphocytes). These are specialised cells that are not HLA restricted. Their function is not well known but they seem to be implicated in tolerance phenomenon5. 8 Figure 2. Components of the Immune Response. This figure provides distribution of immune components participating in the innate response (on the left) and adaptive response (on the right). Source: modified from Dranoff G. Nat Rev Cancer 2004; 4:11-22 b. Immune System Organs and Tissues The organs of the immune system are divided into two groups, primary and secondary lymphoid organs. In primary lymphoid organs, the generation and differentiation of the immune cells takes place. Secondary lymphoid organs are the hosts of the immune cells after differentiation and maturation, what render them ready to play their roles in the defence of the human body. All these organs are connected by blood and lymph vessels leading to a well communicated and coordinated system. Primary lymphoid organs are;  Bone Marrow. The bone marrow is the main hematopoietic organ. It contains hematopoietic stem cells (HSCs). Depending on the microenvironment of the bone marrow, HSCs can differentiate into the myeloid or lymphoid lineage11. Further, the bone marrow is where the maturation of B lymphocytes takes place3. 9  Thymus. It is a glandular organ with the ability to release hormones. It is in this organ where the maturation and selection of T lymphocytes takes place.3,7 Secondary lymphoid organs are;  Spleen. The main function of the spleen relies in the destruction of blood cells. It filters the blood, store blood cells and, in special circumstances, it can lead to extramedullary haematopoiesis. It is in the spleen where some specific immune cells interact with blood circulating antigens and thus can trigger the adaptive immune response12.  Lymph nodes. They are small structures located throughout the human body connected one to another by lymph vessels. Their main function is to filter the different substances collected by the lymphatic vessels7.  Mucosa-associated lymphoid tissue (MALT).This organ is made from aggregates of lymphoid tissue that are located in the submucose of some tissues of our body. These aggregates are composed by immune cells that are able to initiate local immune responses when any danger signal is detected. MALT is the most extensive component of the lymphoid tissue in our body. Examples of MALT are the amygdales, the appendix or the Peyer plaques3,13. 1.1.3. The Activation Of T Cells And The Immune Synapse The term immune synapse is used to designate a virtual space where the molecular interactions that regulate the immune system take place. It is basically a stable adhesive junction between a polarized effector immune cell, normally a T cell, and an APC, most commonly a DC.As previously described, T lymphocytes are needed to get activated in order to be able to perform their activity. Therefore, the appropriateness of the regulation of these interactions is crucial to ensure the defence of the host against pathogens and even neoplastic cells. In case these interactions were deregulated, the host would not be safe from pathogens attack, tumour cells or, even, autoimmunity14,15. 10 Mainly, there are three different signals that take place during the activation of T cells and so in the immune synapse; engagement of the TCR, costimulation and functional differentiation16. The specificity of the immune response is controlled by the interaction between TCR and the proteins from MHC. As previously described, different type of T cells are able to recognise the antigen when they are presented by T cells matching MHC proteins16. This interaction should be prolonged in the time and of high intensity to lead to an effective activation of the naïve T cells. The affinity between the TCR and the MHC is the basis for the stability of this synapse together with additional engagement of co-receptors, such as CD4 and CD8, to other MHC molecules. This junction is further stabilised by adhesion molecules, which basically are members of the integrin and immunoglobulin families. Further, costimulatory receptors are involved, as well16. Costimulatory receptors have a slight signalling capacity or adhesive activity by themselves. However, costimulation enhances the TCR signalling to produce independent signals that will result in the full T cell activation14,15. In this sense, T cells are circulating by the lymphoid organs. Once they find a peptide-MHC displayed in the surface of an APC matching with its TCR, there is an interaction between both cells. This interaction leads to the first activation signal that induces the formation of the immune synapse16. The second signal is triggered by the binding of the costimulatory molecule CD28, in the T cell membrane, to the molecules CD80 and CD86 expressed by the APC. This binding is indispensable for the activation of the naïve T cell as it amplifies the intracellular signalling that leads to an optimal production of IL-2 needed for the T lymphocyte proliferation16. Specifically, the intracellular signalling induces the expression of CD40- ligand in the surface of the T cells. This ligand binds to the molecule CD40 in the surface of the ACP, what leads to the increase of CD80/CD86 expression strengthening the CD28 signal. Therefore, this is a positive loop 11 that lead to a high proliferation of the T cells against the specific antigen, driving to their clonal expansion16. In absence of CD28 costimulation, T cell enter into an anergy status, and therefore are unable to circulate and/or to be activated later16. After some days of close interaction and signals exchange between the T lymphocyte and the APC, the activated T cell express the inhibitory molecule CTLA-4 (Cytotoxic T Lymphocyte Antigen, CD152). The CTLA-4 molecule is expressed as a negative feedback mechanism, known as checkpoint blockade, in order to avoid an uncontrolled proliferation of the T cells. It is expressed during the early stages of T cell activation to control its amplitude. This molecule binds to CD80 and CD86 with a higher affinity than CD28. This binding represents the end of the activation signal and induces immune suppression17. Other important inhibitory molecule is PD-1 (Programmed Cell Death 1).PD- 1 appears later in the immune response and primarily limits the activity of T cells around the site of interaction. PD-1 receptor is up-regulated on activated T cells and is recognised by two ligands, PD-L1 and PD-L2. Both ligands are induced in peripheral tissues by inflammatory signals. The negative feedback via PD-1/PD-L1 axis leads to the decrease of the activation and the effective functions of the T cells and so, it limits collateral tissue damage in at the site of interaction17. This negative feedback could limit the protective efficacy of the immune system in case of chronic exposition to the antigen or the increase of inhibitory signals that lead to the escape of T lymphocytic responses16. These interactions must happen in the context of a third signal made up of immune-activating cytokines as IL-12, type I (IFNα/β) or type II (IFNγ) interferon. These cytokines leads to the functional differentiation of the T cells. After antigen recognition and activation, CD4+ T lymphocytes proliferate into effectors or helpers (Th) whereas CD8+may differentiate into cytotoxic effector CD8+ T cells. The activation of naïve CD8 T cells follows the same process already described but depends on the signals not only received from DC and APC but also from CD4+Th cells. 15,16. 18 response but also promotes cancer immune evasion in a suppressive environment. Different monoclonal antibodies with this immunomodulatory activity have been marketed across different tumour types28. First indications of the first marketed checkpoints inhibitors were in different settings in advanced melanoma. Initially, ipilimumab was the first checkpoint inhibitor marketed in Europe in 2011. This initial marketing authorisation was issued for the second line treatment of patients with advanced melanoma. This is a humanized monoclonal antibody against the Cytotoxic T-Lymphocyte Antigen 4 (CTLA-4). Results supporting this marketing authorisation came from the pivotal study MDX010-20,a phase III randomised controlled trial (RCT) comparing ipilimumab monotherapy, ipilimumab in combination with a melanoma peptide vaccine (gp100) and gp100 in monotherapy. The main objective of the trial was to compare the efficacy in terms of overall survival (OS) of patients treated with ipilimumab in combination with gp100 versus gp100 plus placebo. Secondary objectives included the comparison of OS between patients treated with ipilimumabalone versus gp100-alone and between the two ipilimumab groups. ipilimumab in monotherapy showed an OS HR of 0.66(95% CI: 0.51, 0.87; p-value= 0.0026; median OS of10.1 months vs. 6.4 months) compared to gp100 in monotherapy 29,30. In 2013, an extension of indication was issued in Europe for ipilimumab in for advance melanoma without restriction regarding the line of treatment. Supporting results came from a double-blind phase III RCT in untreated patients with stage III (unresectable) or stage IV melanoma (study CA184024). The main objective of the study was to compare OS between both treatment groups; ipilimumab in combination with dacarbazine versus dacarbazine alone. Ipilimumab plus dacarbazine showed an OS HR of 0.72 (95%CI; 0.59, 0.87;p= 0.0009,median OS 11.2 months vs.9.1 months) compared to dacarbazine alone in previously untreated patients31,32. Back in 2015, nivolumab, a humanised monoclonal antibody which binds to the PD-1, was the second checkpoint inhibitor been marketed in Europe for the treatment of advance melanoma. Results supporting this indication 19 came from studies CA209066 and study CA209037. Study CA209066 was a double-blind phase III RCT of nivolumab compared to dacarbarzine in previously untreated advance melanoma patients. Nivolumab met its primary endpoint showing a reduction of risk of death of 58% over dacarbazine(HR = 0.42; 99.79%CI: 0.25, 0.73; p-value < 0.0001). OS rates at 12 months were also superior for nivolumab arm; 73% (95%CI: 9.33 –12.09) and 42.1% (95%CI: 65.5, 78.9) for nivolumab and dacarbazine, respectively 33.Study CA209037 was an open-label phase III RCT of nivolumab compared to investigator‘s choice in unresectable melanoma patients who had progressed on or after ipilimumab treatment. The dual primary endpoints for this study were overall response rate (ORR) and OS followed by progression-free survival as key secondary endpoint. In previously treated patients, nivolumab showed improvements in ORR with respect to investigator‘s choice of treatment (31.7% vs. 10.6%, nivolumab vs. investigator‘s choice, respectively) and in PFS with a descriptive HR of 0.74 (95% CI: 0.57, 0.97) over the comparator arm. Overall, OS results were not assessed at that time due to immaturity34. The third of these drugs gaining a marketing authorisation in Europe was pembrolizumab, a humanised monoclonal antibody which binds to PD-1 receptor. The initial indication granted to pembrolizumab was for the treatment of advance melanoma in adults. The use of prembrolizumab in naive patients was mostly supported by results of phase III RCT to evaluate the safety and efficacy of two different dosing schedules of pembrolizumab compared to ipilimumab. Selected co-primary endpoints were PFS and OS. Pembrolizumab showed improvements in both co-primary endpoints compared to ipilimumab; PFS HR=0.58 (95%CI: 0.47, 0.72; pvalue<0.00001) and OS HR of 0.69 (95%CI: 0.52, 0.90; p-value<0.00358). Overall, OS rates at 12 months of 74.1%, when administered each 14 days, and 68.4%, when administered each 21 days, compared to 58.2% obtained for ipilimumab and in the first line setting. In patients previously treated with ipilimumab, supporting results from a phase II RCT of pembrolizumab compared to chemotherapy showed a PFS HR of 0.50 (95%CI: 0.39, 0.0.64; p-value<0.0001)35,36. 20 Taking into account the dismal prognosis of melanoma patients at that time, these results were considered impressive and so promising. The impact on survival together with the manageable safety profile of this type of drugs opened the door to their development in different indications. Since then, checkpoint inhibitors in monotherapy have been authorised for several indications, such as, non-small cell lung cancer (NSCLC), renal cell carcinoma (RCC), classical Hodgkin lymphoma (cHL), squamous cell cancer of the head and neck (SCCHN), urothelial carcinoma (UC), oesophageal squamous cell carcinoma (OSCC), Merkel cells carcinoma (MCC) or colorectal cancer (CRC)37–42. Further strategies were also studied in order to increase the success of these therapeutic interventions. This included combinations targeting multiples checkpoints that might prove synergism and increase opportunities for more effective beneficial clinical effects. For instance, nivolumab in combination with ipilimumab in advanced melanoma patients for the first line treatment of advanced RCC were also authorised39,42. However, this strategy also increase the risk of immune-related adverse events (irAEs)25. Figure 3. Anti-CTLA4 & Anti-PD-1 Mechanism of Action This picture summarises the mechanism of action of anti-CTLA-4 and anti-PD-1 drugs. Source: modified from Opdivo EPAR43. Evidence emerged that chemotherapy could have immunostimulatory properties44 and so, other approach was the combination to between checkpoint inhibitors and chemotherapy, as for example; i) pembrolizumab in combination with carboplatin and paclitaxel or nab-placlitaxel in first line for metastatic squamous non-small cell lung carcinoma (NSCLC) showed a median OS of 17 months and an ORR of 63%38,45,46,or ii) durvalumab in 21 combination with etoposide and carboplatin or cisplatin in first line of extensive-stage small cell lung cancer (SCLC) showed a median OS of 12.9 months and an ORR of 67.9%37,47. An additional approach is the combination of checkpoints inhibitors with target therapies; i) pembrolizumab in combination with axitinib compared to sunitinib showed an OS HR of 0.53 (95% CI:0.38, 0.74) and an ORR of 59%, or ii) nivolumab in combination with cabozantinib showed an OS HR of 0.60 (95% CI:0.40, 0.89; p-value=0.0010) and an ORR 55.7% both in the first line of advanced renal cell carcinoma (RCC)38,39,48,49. Taking into account all of the above, these agents are considered to constitute a breakthrough in the treatment landscape for advanced cancer stages. New horizons are to bring these to early stages as in the periadjuvant setting. However, their efficacy may vary not only depending on the tumour type but also on the setting, where these therapies often fails or provide a relatively low proportion of patients respond. In this sense, the identification of subpopulations of patients more likely to reach the benefit from these agents constitutes, nowadays, an unmet need. Translational research and the development of biomarkers are thus essential to reach this objective50. 1.4. BIOMARKERS IN CANCER IMMUNOTHERAPY A biomarker is defined as a biological characteristic that could be measured as an objective indicator of a normal biological process, a pathological process, or the response to an exposure or intervention that can guide therapeutic interventions. Some specific molecular, histological, radiographic, or physiological characteristics can be considered biomarkers. They can be found in blood, other body fluids or tissues. There are different diagnostic approaches by which a suitable biomarker may be identified and measured51–53. Biomarkers have a broad utility as they are able to predict the susceptibility 54diagnose a certain disease (diagnostic biomarker), to assess the stage and 22 the evolution of a certain disease (prognostic biomarker), to predict the clinical outcome to a given treatment (predictive biomarker) or to assess the status of a disease or the evidence of the exposure or effect to a drug or to an environmental agent (monitoring biomarker)51,55. Biomarkers are widely used in drug development as they serve to facilitate the availability of new safer and more effective drugs 54. Even if biomarkers can have an exploratory nature when tested in early phases of the clinical development, the technical or quantitative reliability is essential to be ensured. In addition, in the case that biomarkers would be necessary in confirmatory trials, they must be well validated, ideally in well-designed prospectively clinical trials52. With this regard, biomarkers are broadly used nowadays in clinical development to identify populations, to guide the dose finding process, as an early proof of concept or mechanism of action, to assess toxicity or as indicators of natural course of a disease52. 1.4.1. Immune Biomarkers As previously discussed, biomarkers research is essential to identify the ideal target population for immune checkpoint inhibitors. Efforts are focussing in evaluating immune biomarkers with a predictive role. An example of these biomarkers could be the expression of PD-L1 assessed in tumour samples. The most effective marketed checkpoints inhibitors target the PD-1/PD-L1 interaction. PD-1 is expressed on tumour-infiltrating cells and PD-L1 is expressed on antigen-presenting cells and in tumour cells. Hence, PD-L1 expression can be found in both immune and tumour cells. The level of expression in tumour samples can be easily determined by immunohistochemistry (ICH). The role of PD-L1 expression as a predictive biomarker has been extensively evaluated. Evidence suggested that high PD-L1 expression correlated with increased response rates or OS in RCC, NSCLC or melanoma56–59 23 However, despite many attempts the role of PD-L1 expression as predictive biomarker is not yet clear. For instance, we recently published an article analysing the role of PD-L1 expression in relation with the authorised indications in Europe of atezolizumab as part of its initial application in 2017. In this sense, atezolizumab was initially approved for its use in monotherapy in the treatment of adult patients with (a) locally advanced or metastatic NSCLC after chemotherapy or (b) locally advanced or metastatic UC after chemotherapy or ineligible to cisplatin therapy. The indication in NSCLC was granted disregarding the PD-L1 expression as the benefit was considered established for all the PD-L1 expression subgroups although a trend to better treatment outcomes was seen for patients with higher PD-L1 expression. In the case of the UC indication, the approval of the indication was supported by results from two different trials; one single-arm multicohort Phase II trial and one phase III RCT of atezolizumab compared to chemotherapy. These results were considered conflicting with regards the role of PD-L1 expression. This was because results from cohort 1 of the phase II study (patients treated in first-line) showed a proportion of responders (ORR) consistent across the different PD-L1 expression subgroups whereas a trend towards a longer OS in subgroups with lower PD-L1 expression was seen. Results from cohort 2 of the same study (patients treated in second-line or more) showed better ORR and OS delay in patients with higher expression of PD-L1 compared to the overall population. Results from the phase III trial (platin pre-treated patients) were no statistically significant for OS in the population with higher PD-L1 expression. Controversially, the benefit in both settings, first and second line, was considered established based on relatively similar results response rates to chemotherapy but substantially longer duration of the responses (DOR). However, a post-authorisation measure was adopted and the marketing authorisation holder was requested to submit results from a third study, an ongoing phase III RCT with unavailable results at the time of the procedure. In 2018, interim results of this third study submitted to the European Medicines Agency (EMA) for evaluation, concluded that OS for atezolizumab in monotherapy compared to chemotherapy was inferior for 24 patients with low PD-L1 expression. Consequently, the initially approved indication in UC was amended to exclude these patients50. Other similar marketed products with restrictions per PD-L1 expression are pembrolizumab in UC, in HNSCC and in NSCLC, or durvalumab in NSCLC. Additional uncertainties to be added to the demonstration of its clinical value is the fact that still there are not harmonization to assess the PD-L1 expression. Each marketed drug of this type has its own companion diagnostic test. They cannot be used interchangeably as not all assays have comparable levels of sensitivity and reproducibility. Indeed, analysis of same samples using different test may lead to discordant results that could hamper the eligibility to patients to ICIs in clinical practice60,61. 1.4.2. Immune Biomarkers in Peripheral Blood Specifically in oncology, biomarkers in tumour samples are supposed to be the major source of information. Due to tumour heterogeneity, it is extremely difficult to capture all the tumour landscape from a single biopsy52. Therefore, biomarkers from peripheral blood or other tissues would provide further complementary information. In the case of biomarkers in peripheral blood, advantages over tissue biomarkers are huge, being the easier sample access one of the most valuable62. Ideally, they could be routinely monitored in clinical practice and thus, lead to guide clinical decisions63. Several strategies and techniques are being studied in order to explore the applicability of immune biomarker from liquid samples; the count of specific immune cells by flow cytometry or the detection of circulating tumour cells, circulating DNA from the tumours or exosomes by the so called liquid biopsy are examples of them63,64. With regards to the development of immune biomarkers in cancer immunotherapy, a possible explanation to the failure of immune therapy is that patients present an immunosuppressive tumour microenvironment. Taking into account the dynamic changes of immune cells in the tumour 25 microenvironment, the analysis of circulating immune cells in peripheral blood is consider a valuable non-invasive alternative 65. Examples of immune cells with suppressive activity present in the tumour microenvironment that can be easily detected in peripheral blood with a potential clinical value could be myeloid-derived suppressors cells (MDSC), regulatory T cells (Tregs) or CD8 T cells, among others. a. Myeloid-Derived Suppressor Cells Overall, MDSCs are a heterogeneous group of cells with immunosuppressive properties. This group includes myeloid progenitors and immature myeloid cells.66. MDSCs cells seem to be the result of an altered myelopoiesis induced by an abnormal concentration of different mediators at the tumour microenvironment (TME). Given a stressful situation in the body, such as an infectious process, the bone marrow is stimulated to release immature myeloid cells to peripheral tissues where the aggression is detected in order to prevent consequences from an excessive immune stimulation and to limit autoimmune responses against antigens. In normal circumstances once in peripheral tissues, these cells can differentiate by myelopoiesis into mature cells from the myeloid lineage. This process is controlled by different cytokines as Granulocyte Macrophage-Colony Stimulating Factor (GM-CSF), M-CSF, SCF or IL-3. In pathological conditions, as cancer disease, characterised by a chronic inflammation associated to immune suppression, myelopoiesis seemed to be altered result in the accumulation of immature myeloid precursors in the affected tissues. Tumours use the expansion of suppressive cells as mechanism of immune escape. In this sense, the concentration of a wide variety of suppressive cytokines, chemokynes or pro-inflammatory factors released by tumour and stromal cells led to the enhancement of altered myelopoiesis preventing myeloid cells maturation but inducing their activation and acquisition of immunosuppressive functions. Examples of chronic inflammatory factors participating in this process could be GM-CSF, tumour necrosis factor-α (TNF-α), IL-1β, IL-6, IFN-γ, transforming growth factor-β (TGF-β) or vascular endothelial growth factor (VEGF). This process also led to the recruitment of additional MDSCs to the TME by the release of 26 chemokines like C-C chemokine ligand 2 (CCL2), CCL5, IL-8 or C-X-C motif chemokine ligand 12 (CXCL12).66–69 The main immunosuppressive function of the MDSCs are their capacity of inhibit the proliferation, differentiation and activation of Tcells; a) making T cells exhausted and incapable of recognizing tumour cells by releasing reactive oxygen species (ROS) or nitrogen oxide (NO) or exhausting the availability of L-arginine and L-cystein, essential amino acids for their well functioning, and ii) controlling the NK cell cytotoxicity by releasing TGF-β and IL-10. In addition, these cells may promote tumour angiogenesis and seem also to play an important role in tumour spread, facilitating the colonization step of the metastatic cascade by enhancing the migratory potential of the tumour cells67,68,70.MDSCs could also stimulate the clonal expansion of Tregs71. MDSC are known to have an important role in promoting cancer immune evasion. These cells have been described to be expanded not only in the TME but also in peripheral blood in different types of cancers. In this sense, circulating MDSCs have been proposed as an immune biomarker. However, available information on circulating levels, clinical value and differences between tumour types, is still limited.72 An explanation of this lack of data could be the challenges historically found in their identification, as: i) this is an heterogeneous population, ii) with different stages of differentiation, and iii) with confounding phenotypic and functional characteristics common to other myeloid cell population.72 These cells are generally characterised by the expression of CD33, CD11b and alack or low expression of HLA-DR. Mainly, three different subtypes of MDSCs had been described in humans: early-stage MDSCs, granulocytic (G- MDSCs), and monocytic (M-MDSCs)68. The mechanisms of suppression could differ depending on the subtype and proportions could also vary depending on the tumour type.66Subpopulations seem to differ in the expression of CD15 and CD14. 27 To note that G-MDSC are phenotypically different to steady-state neutrophils as they present less granules, increased arginase 1 (ARG 1), and express CD11b and CD66b. For this reason, it seemed more appropriate to change the term from G-MDSC into polymorphonuclear MDSC (PMN-MDSC) 72,73.Overall, phenotype of PMN-MDSCs phenotype is defined as CD11b+HLA-DR−/lowCD14−CD15+ or CD11b+CD14−CD66b+ whereas M-MDSCs is currently defined as CD11b+HLA-DR−/lowCD14+CD15−. In order to capture more immature cells, the market Lin-is used and includes CD3, CD14, CD15, CD19, CD56.Hence, early-stage MDSC are defined asLin−HLA-DR−CD33+ cells73. Related to its clinical value, an increased frequency of circulating MDSCs in peripheral blood seemed to be associated with poor prognosis and improved risk of disease recurrence. Furthermore, MDSC could also compromise the effectiveness of immune-based therapies by possibly playing a role in resistance to this type of therapies66,69,74,75. With regards to therapy with checkpoint inhibitors, there is evidence suggesting circulating MDSC could potentially predict response to this type of therapy. This evidence is maintained across the tumour types described. For instance, melanoma patients presenting a low level of MDSC at baseline were reported to reach better clinical responses and improved OS results while treated with ipilimumab. In addition, it is also reported that melanoma responders to ipilimumab experienced decreased in MDSC levels compared to baseline levels76–78 In prostate cancer, a study of metastatic castration-resistant prostate cancer patients treated with a combination therapy of a GM-CSF-secreting cancer vaccine (i.e. GVAX) with ipilimumab reported a low frequency of pretreatment MDSC for those patients who presented delays in OS79. Similar conclusion were also reported in NSCLC; a low level of MDSC seems to correlate with improved clinical outcomes to anti-PD-1 therapy65,80–82. Furthermore, higher levels of MDSC seem to correlate to a late state of the disease; a higher numbers of MDSC was reported in NSCLC patients with stage III and IV compared to stages I and II.65 34 maintained exposure along the years to these hormones may also be linked to other risk factors that play an important role in the development of breast cancer, as low parity or breast feeding109. Lifestyle could also contribute to the development of breast cancer per hormonal mechanisms; i)obesity or overweight is associated with a higher risk for postmenopausal breast cancer, ii)alcohol intake is associated with a greater risk for both premenopausal and postmenopausal breast cancer, iii) diet based on high fat, low fibre and high simple carbohydrates could also increase this risk, iv) sedentary lifestyles was also found to increase risk when comparing both sedentary and active populations110,111. d. Mortality and morbidity (natural history) Breast cancer was the 4th leading cause of death worldwide and the 3rd in Europe in 2020. According to GLOBOCAN 2020, Asia was the region with mortality rates by regions were the highest, followed by Europe, Africa, Latin America & the Caribbean, Northern America, and Oceania97,101,103. In Europe, a total of 91,826 deaths were reported in 2020. Specifically in Spain, breast cancer was also the 4th leading cause of death with a total of 6, 606 reported deaths in 2020102,103. Mortality rate seems to have decreased in the last decades, especially in younger populations, what could be explained by the earlier detection of the disease and the approval and implementation of new systemic treatments103,107. e. Classification of Breast Cancer Breast cancer is a heterogeneous disease. Many differences can be found in terms of molecular alterations, cellular composition or clinical outcomes. The development of useful clinical tumour classification with a prognosis or predictive value was deemed necessary112. Extensive genomic analysis and tests assessing gene expression profile could better define breast cancer subtypes. Indeed, multi parameter molecular tests (PAM-50 or MammaPrint/BluePrint) were developed in this regard. Nevertheless, the 35 worldwide availability of these tests was uncertain as they did not seem largely affordable. Therefore, in order to simplify the characterization of the disease, standard pathological features using immunohistochemical methods (IHC) were agreed to be used to define subgroups of breast cancer. At this respect, tests as hormone (estrogens& progesterone) receptor (HR), human epidermal growth factor receptor 2 (HER2) or proliferation index (ki67) expressions, were agreed to be performed in the context of breast cancer diagnosis. As a consequence, 5 different molecular subtypes have been internationally accepted: ‗triple negative‘ (HR-/HER2-), ‗luminal A‘ (HR+/HER2-), `luminal B‘ (HR+/HER2-/high ki67), ‗luminal B-HER2‘ (HR+/HER2+) and HER2-enriched (HR-/HER2+). This classification represent nowadays the main tool for clinicians to tailor systemic therapies113. 1.5.2. HER-2 Negative Advanced Breast Cancer In the following section, we will describe luminal and triple negative breast cancer types as they represent the populations in which our work is mainly focused. a. Luminal breast cancer Luminal breast cancer represent up to 75% of the invasive breast cancers. These tumours are supposed to have a slower growth and are mainly characterized by the presence of Estrogens (ER) and Progesterone (PgR) Receptors114. Under this designation, two main subgroups of tumours are described: luminal A tumours and luminal B tumours. Luminal A tumours are characterized by high expression ER-related genes, low expression of HER2 genes, low expression of Ki-67 (proliferation marker) and high expression PgR-related genes. Luminal B tumours are characterized by low expression of ER-related genes, variable expression of HER2-related genes, and high expression of Ki-67. Luminal B tumours present a worse prognosis than luminal Atumours115. 36 Due to this biological feature of expressing hormonal receptors, the endocrine therapy (ET) is considered as the therapeutic backbone for this type of cancer either in the adjuvant or the metastatic setting. Endocrine therapies include selective non steroidal aromatase inhibitors (e.g. letrozole or anastrozole), steroidals aromatase inhibitors (e.g. exemestane), selective ER modulators (e.g. tamoxifen), and ER antagonists (e.g. fulvestrant).116 According to guidelines, ET is the preferred treatment strategy for advanced HR+ disease, except for patients who presented visceral crisis or presumed endocrine resistance. Chemotherapy is usually preferred in cases with visceral involvement or when the disease is considered aggressive116. Overall, sequential ET with different endocrine agents should be used as long as the patient is benefiting from this kind of therapy. Nevertheless, approximately, 20-25% of the advanced breast cancer patients with this tumour subtype show an early and rapid progression. This fact can be explained by the development of mechanisms of therapeutic resistance to this type of therapies. Although not fully elucidated, ER loss or ER mutations seem among the main mechanisms leading to endocrine resistance. In this sense, the choice and sequence of different available endocrine agents will depend on which agent was previously used and the outcome obtained114,116–118. The development of endocrine resistance and the early disease progression is considered a critical issue. In order to delay or circumvent endocrine resistance, a new therapeutic class of agents has been introduced, the CDK4/6 inhibitors (CDKi). These new therapeutic agents are aimed to be administered in combination with endocrine agents, as for example, aromatase inhibitors. This combination seems to have a positive impact in progression free survival (PFS) and OS in different settings of advance breast cancer, according to results from Phase II and III trials. However, uncertainties remained on the optimal integration in terms of the need of biomarkers to understand whether combination therapies or stand-alone treatments are better for each individual patient, the optimal sequence of these drugs and the optimal treatment after progression to CDKi119–125. 37 Another possible therapeutic strategy relies in the combination of endocrine agents with mTOR inhibitors. This kind of therapy is used when other therapeutic alternatives are exhausted as it only demonstrated a modest impact in PFS but not in overall survival (OS) with an unfavourable safety profile. This combination, however, remain a valid option for patients who developed endocrine resistance117,126. Recently, alpelisib is an α-specific class Iphosphatidylinositol 3-kinase (PI3K) inhibitor has been approved by the European Commission in combination con fulvestrant for the treatment of advance luminal-like breast cancer with PIK3CA mutation after progression to any endocrine agent in monotherapy. This approval was supported by results from a pivotal phase III double-blind placebo-controlled trial, SOLAR-1. SOLAR-1 study included postmenopausal women and men with HR+HER2- advanced breast cancer whose disease had progressed or recurred to an aromatase-inhibitor-based treatment (with or without CDK4/6 combination). Alpelisib in combination with fulvestrant prolonged PFS in PIK3CA mutated patients compared to placebo and showed no detrimental trend in OS. Safety results suggested a low tolerability of the combination based on high rates of discontinuations but presumably higher than for chemotherapy.SOLAR-1 study included a limited number of patients who had previously received CDK4/6 inhibitors which hampered any conclusion of clinical benefit in such population127–129. b. Triple negative breast cancer Triple-negative breast tumours are characterised by the lack of expression of hormonal receptors(ER or PgR) and lack of over-expression or amplification of HER2. These tumours are large in size, poorly differentiated, with highly lymph node involvement at diagnosis and exhibit a more invasive capacity. Compared to other forms of breast cancer, they show a worse prognosis; i) more aggressive clinical course, ii) increased likelihood of local and distant recurrence and iii)higher risk of early recurrence from 1 to 4 years after diagnosis. Metastasis are mostly occurring in visceral organs and brain rather than in bone as for other breast cancer subtypes 130,131. 38 In relation to disease management, it constitutes a high medical need. Clinical guidelines recommend the use of sequential single-agent chemotherapy except in patients with visceral crisis or rapidly progressing disease. However, the optimal sequence is not defined as no single chemotherapy agent demonstrated clear superiority over others. The choice of specific cytotoxic agents seems to be conditioned by a large number of factors and needs to be considered individually. Available agents for these patients are taxanes, anthracyclines, capecitabine, gemcitabine, platinum salts, vinorelbine or ixabepilone116,118,132. Clinical trials with translational substudies seem to be the best option in the metastatic setting, as the only way to find an adequate treatment for this population relies in a better biological characterization of these tumours116,126. Until very recently, there were no drugs available specifically approved for this disease apart from chemotherapy. PARP inhibitors become recently available for TNBC patients with BRCA1/2 in the late setting132. Additionally, atezolizumab has been approved by the European Commission in combination with nab-paclitaxel for the treatment of unresectable locally advanced or first line metastatic triple-negative breast cancer (TNBC) with PD-L1 expression ≥ 1% 133. This approval was supported by results from study IMpassion 130 (WO29522), a phase III randomised clinical trial double blind placebo-controlled study aimed to evaluate the efficacy and safety of atezolizumab in combination with nab-paclitaxel in patients with unresectable locally advanced or metastatic triple-negative breast cancer who had not received prior chemotherapy for metastatic disease. In patients whose tumours expressed PD-L1 ≥1%, atezolizumab in combination with nab-paclitaxel compared to nab-paclitaxel alone showed a clinically meaningful and statically significant delay in PFS (HR 0.62; 95% CI0.49, 0.78; p<0.0001; mPFS 7.5 vs. 5.3 months). This combination also demonstrated an important impact in OS (HR 0.67; 95% CI0.53, 0.86,mOS 25.4 vs. 17.9) for the same population. The safety profile was considered acceptable for this specific setting133. 39 Furthermore, sacituzumab govitecan-hziy was recently approved by FDA under accelerated approval based on tumour response rate and duration of response. This is a new antibody-drug conjugate (ADC), an anti-Trop-2 humanized monoclonal antibody coupled to a topoisomerase I inhibitor, the active metabolite of irinotecan (SN-38), by a hydrolysable linker (CL2A). This approval was based on results from a Phase I/II multicenter single arm trial, IMMU-132-01. Study IMMU-132-01, enrolled a total of 108 metastatic triple-negative breast cancer patients who had received at least two prior treatments for metastatic disease. Results from this study showed an ORR of 33.3% (95% CI: 24.6, 43.1) and median duration of response of 7.7 (95% CI: 4.9, 10.8) months. The majority of the responses were partial responses (30.6%). Sacituzumab govitecan-hziy seemed to present manageable safety profile with low rates of adverse events leading to discontinuation. Most common adverse event of CTCAE grade >3were neutropenia, anaemia and leucopoenia134,135. These results and the role of this agent in the management of advance triple-negative breast cancer are yet to be confirmed in Phase III trials132. Preliminary results of a Phase III randomized controlled trial (RCT), ASCENT study, showed clinically and statistically significant improvements in PFS(5.6 vs. 1.7 months; HR: 0.41; P<0.0001) and OS(12.1 vs. 6.7 mo; HR: 0.48; P<0.0001) of sacitumab govitecan-hziy compared to single-agent standard chemotherapy (capecitabine, eribulin, vinorelbine or gemcitabine)of physician's choicein advance triple-negative breast cancer. Response rates, ORR (35% vs. 5%) and CBR, were also greater in the active arm compared to the comparator arm. Safety findings were consistent with previous studies and no important safety signals were indentified. The ASCENT study enrolled a total of 529 metastatic triple-negative breast cancer patients who had received a median of four prior anticancer therapies136–138. 1.5.3. Cancer Immunotherapy in Breast Cancer Traditionally, breast cancer has not been considered an immunogenic tumour. However, evidence emerged suggesting immune activity in some breast cancer subgroups, mostly in triple negative tumours, based on the 40 better understanding on the components of the breast TME. Therefore, development of immune checkpoints inhibitors in breast cancer was initiated. First approach was to test immunomodulatory monoclonal antibodies for their used as monotherapy. Several early-phase studies of showed some degree of activity monoclonal antibodies in different settings of TNBC and luminal breast cancer patients in the metastatic setting. Results were promising when selecting patients for PD-L1 expression although still modest taking into account the limited sample size and the exploratory nature of these type of designs. Results for ORR were low when the population was unselected for PD-L1 expression 139. Currently, there is only one phase III trial reported that tested an immune checkpoint inhibitor in monotherapy against chemotherapy. This study is KEYNOTE-119; a randomized, open-label, phase III study of pembrolizumab compared to single agent chemotherapy per physician's choice for metastatic TNBC. The study aimed to randomise over 600 patients in a 1:1 ratio into pembrolizumab arm or treatment per physician‘s choice arm (capecitabine, eribulin, gemitabine or vinorelbine). Patients were stratified by PD-L1 status (positive [CPS ≥1] vs. negative [CPS <1]), and history of prior(neo)adjuvant treatment vs. de novo metastatic disease at initial diagnosis. Eligible patients were those with confirmed stage IV/M1 TNBC, who were previously treated with anthracycline and/or taxanes in the adjuvant or in the metastatic setting and presented adequate organ function and an ECOG PS score between 0 and 1. Primary endpoints for the study were OS analysed in participants with PD-L1 with combined positive score (CPS) ≥10, ≥1 and in all participants (ITT). Secondary endpoints for efficacy included ORR per RECIST v1.1, PFS, DOR, DCS all tested in the three population140,141. Finally, the study randomised 622 patients treated in second or third line. At median FUP of 9.9 months for pembrolizumab arm and 10.9 months for the comparator arm, results did not showed statistically significant improvements in OS or in PFS for none of the three populations tested142. 41 Based on the manageable safety profile of these drugs, a second approach was to seek for synergisms with other effective drugs in order to improve clinical outcomes. This opened the door for a wide choice of strategies to be tested depending on the type of the tumour but also on the settings. In addition, this could also allow patients to benefit from immune checkpoints inhibitors regardless of PD-L1 expression. In this sense, the ideal chemotherapy agent for the combination could be chosen based on their expected immune activity. It is well known that chemotherapy agents may induce changes in the TME, such as, PD-L1 up - regulation, increase antigen release by tumour cells or enhance the expression of immunogenic surface markers as MHC class I. These effects may also enhance the anti-tumour activity of immune check point inhibitors when both type of drugs are administered in combination. Specific immune activities reported in literature for chemotherapy agents are as follows44;  Anthracyclines can increase the proliferation of CD8+ Tcells and are capable to induce immunogenic cell death that may lead to anti-tumour immune response by dendritic cells activation and specific Tcells response.  Taxanes, specifically docetaxel and paclitaxel are reported to increase TIL recruitment in breast cancer but also to reduce the immune suppression in the TME by selectively decrease Treg cells and MDSC levels in the TME.  Cyclophosphamide can induce immunogenic cell death and suppress Treg cells and increase proliferation of CD8+ T cells and NK cells.  Gemcitabine can decrease MDSC levels but also to increase the antitumour activity of CD8+ T cells.  Platinum salts can induce immunogenic cell death, to promote T cell activation and down regulate MDSC activities. 42 In the specific case of advance breast cancer, some examples of strategies that are being tested in phase I/ II trials; i) atezolizumab with bevacizumab and/or with nab-paclitaxel in a cohort of locally advance or metastatic TNCB pre-treated patients, ii) pembrolizumab and trastuzumab in metastatic patients, trastuzumab-resistant with HER2+ breast cancer, iii) pembrolizumab in combination with eribulin mesylate in metastatic TNBC patients139. Examples of strategies tested in phase III trials are; i) atezolizumab in combination with nab-paclitaxel and ii) pembrolizumab plus chemotherapy compared to placebo plus chemotherapy, both in TNBC first line setting. As previously discussed, atezolizumab has been recently approved in combination with nab-paclitaxel for the first line treatment of metastatic triple negative disease with PD-L1 expression ≥1%. Until now, it is the only checkpoint inhibitor that has been approved in breast cancer. This approval was supported by the results from one single pivotal study, Impassion 130. Impassion 130 was a phase III, randomised, placebo-controlled, doubleblinded, two-arm study to evaluate the efficacy and safety of atezolizumab in combination with nab-paclitaxel compared to placebo plus nab-paclitaxel in patients with locally advanced or metastatic TNBC, who have not received any prior chemotherapy regimen for metastatic disease. Eligible patients were those who presented metastatic or locally advanced, histologically documented TNBC per American Society of Clinical Oncology- American College of Pathologists (ASCO-CAP) criteria, with absence of HER2, ER or PR expression. Patients should had not received any previous treatment with chemotherapy or targeted systemic therapy for inoperable locally advance or metastatic disease. In case of having received neoadjuvant or adjuvant treatment, it needed to have been completed with more or equal to 12 months prior to randomization. Patients needed to be eligible to taxane monotherapy treatment understanding that these patients would have absence of rapid clinical progression, life-threatening visceral metastasis, or the need for rapid symptom and/or disease control. Furthermore, patients needed to have measurable disease per RECIST v.1.1 43 and available tissue sample evaluable for tumour PD-L1 expression. Patients needed to present an adequate hematologic and end-organ function and an ECOG PS score up to 1. The study was design with co-primary endpoints. These were PFS and OS tested in both, ITT and PD-L1 selected populations. Key secondary endpoints included ORR, DOR and time to deterioration (TTD) in global health status/ health-related quality of life (HRQoL). Overall, PD-L1 expression was assessed in both fresh and archival tumour samples. Patients were classified as PD-L1 status positive or negative. PD- L1 positive status was defined as ≥ 1% of staining of tumour-infiltrating immune cells (IC) in the tumour area. Randomisation to study or comparator arms was performed into a 1:1 ratio. Patients were stratified by presence of liver metastasis, prior taxane treatment or PD-L1 status. The study randomised a total of 902 patients; from them, 369 patients had PD-L1 positive tumours. The study met their co-primary endpoints showing clinically meaningful results in both populations. These were supported by results from key secondary endpoints. However, the benefit observed was mostly driven by the PD-L1 ≥ 1% population as when excluding them from the analysis, results in PD-L1 < 1% population did not showed any improvements in PFS (HR 0.94; 95% CI 0.78, 1.13) nor in OS (HR 1.02; 95% CI 0.84, 1.24). Although the subgroup benefitting from these therapy was considered that could have been better defined than only related to PD-L1expression ≥ 1%, the addition of atezolizumab to nab-paclitaxel was not considered to provide any additional benefit to patients with PD-L1 expression < 1% . The approved indication was, thus, limited to patients with TNBC with PD-L1 expression ≥ 1%133,143. In the case of pembrolizumab, the study KEYNOTE-355 is an ongoing randomised, placebo-controlled, double blind, phase III trial of pembrolizumab plus chemotherapy versus placebo plus chemotherapy (nabpaclitaxel; paclitaxel; or gemcitabine plus carboplatin) for previously 50 Pembrolizumab is a human PD-1 blocking antibody with several indications gained in different type of oncologic diseases. As previously discussed, PD- 1isa receptor expressed on activated T cells that plays a major role in the maintenance of T cell tolerance limiting effective T cell responses. Currently, there are two known ligands of PD-1, PD-L1 and PD-L2 (or B7-H1 and B7- H2). Overall, PD-L1 is considered the most important one and it is aberrantly expressed in some tumours including breast cancer. Therefore, it can induce immunosuppression through signalling PD-1. In breast cancer PD-L1 expression (in tumour tissue and Tumour infiltrating lymphocytes (TILs)) has been shown to be correlated with worse clinic pathological characteristics of the disease, such as larger tumour size, histology grade III tumours, or negative hormone receptors. Blocking PD-1/PD-L1 interaction might have an immunogenic effect in breast carcinoma patients. In addition, a set of immune biomarkers in peripheral blood, such as MDSCs, Tregs and cytokines, among others, were selected to be assessed during the trial. This translational substudy, named PANGEA-BreastT study, was designed to analyze the above mentioned set of immune biomarkers in the clinical setting, looking at their basal level and monitoring their evolution at different tine points during the treatment with gemcitabine and pembrolizumab. This translational research may shed light on the putative mechanisms of the eventual clinical activity of this combination, in order to identify biomarkers that could be monitorised and maybe also predict clinical activity. Study Drug/Medication: Enrolled patients were treated with: - Pembrolizumab at a dose of 200mg as an intravenous (IV) 30 minutes infusion on day 1of each 21-day cycle. in combination with - Gemcitabine at a dose of 1.250mg/m2 as an intravenous (IV) 30 minutes infusion on day 1 and 8 of each 21-day cycle. Treatment continued until objective disease progression, clinical progression (under investigator criteria), unacceptable toxicity, death or withdrawal of 51 consent, whichever occurred first. Patients that completed 24 months of uninterrupted treatment with pembrolizumab or 35 administrations of study medication were supposed to stop pembrolizumab treatment (though may continue with gemcitabine). Subjects who stopped pembrolizumab after 24 months were supposed to be eligible for up to one year of additional pembrolizumab treatment if progressed after finishing treatment. Prophylactic antiemetic therapy was administered under site criteria. An initial exploratory run-in-phase was performed to test the safety of the combination and to determine the Recommended Phase II Dose (RP2D) of gemcitabine in combination with fixed doses of pembrolizumab. Design and targeted population: This was a multicenter phase II trial, with an initial exploratory run-in- phase, to evaluate the efficacy and safety of pembrolizumab in combination with gemcitabine in patients with HER2- negative ABC that had previously received anthracyclines and taxanes(unless clinically contraindicated). In hormone receptor positive patients, previous treatment with 2 or more lines of hormone therapy was also a requirement. Patients must have had at least one measurable lesion that could be accurately assessed at baseline and was suitable for repeated assessment by CT, MRI or plan X-ray and a metastatic lesion accessible for biopsy. Approximately 53 patients (up to a maximum of 65 patients depending on the results of the run-in-phase) were expected to be included in this trial. The study included two cohorts of patients: i) Triple Negative and ii) Luminal A+B, with an approximate 1:1 distribution between both groups. A safety dose testing or ―run-in-phase‖, with a 6+6 design, in which toxicity was aimed to be evaluated within the first cycle, was performed since pembrolizumab in combination with gemcitabine was not being previously tested. Initially 6 patients were included in the study at dose level 0 (gemcitabine at a dose of 1,250mg/m2 as an IV infusion on day 1 and 8 of each 21-day cycle and pembrolizumab at a dose of 200mg as an IV infusion 52 on day 1 of each 21-day cycle). The rules of the ‗run-in-phase‘ were as follows; - If ≤ 2 patients experienced a Dose Limiting Toxicity (DLT), 6 additional patients were supposed to be included at the current dose level. If there was a confirmation of this dose to be safe (≤ 3 patients experiencing DLT), this dose would have been considered as the RP2D and it would have been used for the following recruited patients. - If ≥ 3 patients experienced a DLT within the first 6 patients, or ≥ 4 within the first 12 patients included at dose level 0, a de-escalation to dose level - 1 (gemcitabine at a dose of 1,000mg/m2 as an IV infusion on day 1 and 8 of each 21-day cycle and pembrolizumab at a dose of 200mg as an IV infusion on day 1 of each 21-day cycle) was expected to be performed. In this case, a group of 12 additional patients were supposed to be included at dose level -1, if ≥ 4 experience DLT, this combination would been considered too toxic and the study would have been stopped. If ≤ 3 experiences DLT with this combination, this would have been considered the RP2D and it would have been used for the following recruited patients. Initially 3 patients were allowed for inclusion simultaneously. These 3 patients were followed closely during the first cycle to observe the occurrence of any DLT. At the times these 3 patients are completing the first cycle, the next 3 patients were included one by one, until the first cohort is completed. If none of these 6 patients had a DLT, up to 4 patients more could be allowed for inclusion from the second cohort of 6 patients; they were supposed to follow the same procedure as in the first cohort. If one of these patients from the second cohort had a DLT, the inclusion would have been performed in smaller groups (with a maximum number of 4 patients with DLT) and following the same procedure as in the first cohort of 6 patients. An internal committee periodically reviewed the safety data in order to take the decision to maintain or decrease the dose level. This internal committee consisted of the chief investigator, the GEICAM medical monitor and the study statistician. The meetings were performed by teleconference to take 53 these decisions as quickly as possible once the last patient finished the first cycle of treatment. Other meetings were considered ad-hoc whenever necessary (i.e. when new DLTs appear). Patients included in the run-in-phase at the same dose than that in the phase II were considered for the phase II analysis. Objectives of the PANGEA-breast cancer trial: Primary Objectives: - Run-in-phase: To determine the Recommended Phase II Dose (RP2D) of gemcitabine in combination with fixed doses of pembrolizumab. - Phase II: To assess the efficacy of pembrolizumab in combination with gemcitabine in terms of Objective Response Rate (ORR) in patients with HER2-negative ABC. Secondary Objectives: - To assess other efficacy measures of the combination in patients included in the phase II (including those in the run-in-phase at the same dose that in the phase II). - To determine safety and tolerability of the combination in all patients included in the study. Exploratory Objectives: - To assess other efficacy measures of the combination based on immunerelated (ir) response criteria in patients included in the phase II (including those in the run-in phase at the same dose that in the phase II). - To search for tumour tissue and peripheral blood biomarkers of clinical activity. - To compare biomarkers data from cohorts of healthy volunteers (if available) with data from patients included in the study. End-points: Primary End-points: 54 - Run-in phase: To determine the incidence rate of Dose Limiting Toxicity (DLT) with in the first cycle of the combination. - Phase II: ORR is defined as CR plus PR according to Response Evaluation Criteria in Solid Tumours (RECIST) version 1.1. Secondary End-points: - Efficacy:  Progression-Free Survival (PFS) assessed according to RECIST version 1.1 by the investigator.  Clinical Benefit Rate (CBR) defined as Complete Response (CR) plus Partial Response (PR) plus Stable Disease (SD) lasting ≥ 24 weeks according to RECIST version 1.1.  Response Duration (RD) assessed according to RECIST version 1.1.  Overall Survival (OS). With special interest on long term responders (i.e. alive and without disease progression after 24 months of study treatment). -Safety was assessed by standard clinical and laboratory tests (haematology, serum chemistry). Adverse Events (AE) grade were defined by the NCI CTCAE (National Cancer Institute Common Terminology Criteria for Adverse Events) version 4.0. Exploratory End-points: - Efficacy:  ORR is defined as irCR plus irPR according to irRECIST.  Progression-Free Survival (PFS) assessed according to irRECIST by the investigator.  Clinical Benefit Rate (CBR) defined as irCR plus irPR plus irSD lasting≥ 24weeks according to irRECIST.  Response Duration (RD) assessed according to irRECIST. - A set of immune biomarkers were analysed and correlated with evolution of the disease and efficacy of pembrolizumab in combination with gemcitabine (ORR and other efficacy end-points: PFS, CBR and RD), paying special attention to long term responders. 55 - This set of immune biomarkers was to be compared with those from healthy volunteers (if available). Study population: Patients with HER2 negative advanced breast cancer. (See complete eligibility criteria in Annex I) Justification of Sample size determination: A Simon minimax two-stage design was employed with the possibility of stopping early due to lack of response. Results from previous studies showed that gemcitabine produced a response rate of around 20%, this will be our expected H0. With the combination of pembrolizumab and gemcitabine, it was expected that this rate increased to 35% what will be our H1 (an absolute increase of 15%), with an alpha error of 0.05 and a statistical power of 80%, 53 evaluable patients were needed to be included in this trial. The first stage was supposed to include 31 evaluable patients, if at least 7present a response, recruitment was expected to continue to include the 53 evaluable patients. The null hypothesis of H0=20% was supposed to be rejected if 16 or more responses were observed in 53 patients. Statistical Analyses:  Demographics and Baseline Characteristics - Standard descriptive statistics, such as the mean, median, range and proportion, were agreed to be used to summarize the patient sample and to estimate parameters of interest.  Safety Analyses - AEs and Serious Adverse Events (SAE) were agreed to be reported in frequency tables (overall and by intensity). The safety analysis were decided be performed in the population that has received at least one dose of any of the study drugs/medications. 56  Efficacy Analyses - All efficacy endpoints were agreed to be evaluated in the Efficacy and ITT populations.  Biomarker analysis - The biomarker analysis was designed as exploratory and descriptive. For continuous variables, mean, standard deviation, median, minimum and maximum values were agreed to be provided. Categorical variables to be summarized by numbers and proportions. Biomarker endpoints to be evaluated in all patients enrolled in the study with available samples. Study Duration: The end date of study was set as the date of last patient´s death or the date when there is sufficient data to achieve the primary and secondary objectives and all patients have ended the study treatment, whichever comes first. Performing exploratory objectives were independent of the date of the end of the study. 57 58 2. HYPOTHESIS AND OBJECTIVES 2.1. HYPOTHESIS The hypothesis of this sub-study of the PANGEA breast cancer trial is that some specific immune cells detectable in peripheral blood could be useful biomarkers to monitor the clinical response in advanced breast cancer patients treated with a chemo-immunotherapy strategy based on gemcitabine plus pembrolizumab. 2.2. OBJECTIVES This is a sub-study of the PANGEA-Breast trial that aimed to explore the clinical role of a set of immune biomarkers in peripheral blood. Primary objective;  To explore any modifications along the course of the treatment of a set of immune biomarkers in peripheral blood (MDSCs, Tregs, CD8+Tcells, lymphocytes and CD4/CD8 ratio) Secondary objectives  To explore any relationship between the basal count and/or modifications throughout the course of the treatment of this set of immune biomarkers(MDSCs, Tregs, CD8+ Tcells, lymphocytes and CD4/CD8 ratio)and the clinical outcome achieved along the course of the treatment.  To explore any relationship between the basal count and/or modifications throughout the course of the treatment of this set of immune biomarkers(MDSCs, Tregs, CD8+ Tcells, lymphocytes and CD4/CD8 ratio)and histology. 59  To explore any differences in MDSC counts at baseline of PANGEA patients and a cohort of healthy volunteers. 66 In order to facilitate the process, CD127 was selected to detect theses cells as the expression of this marker inversely correlates with the expression of FOXP3 in both healthy subjects and patients86–88. In this sense the presence of Tregs cells defined as CD4+CD25+CD127- guarantees the presence of FOXP3 in the selected cells. The commercial antibodies used to detect these cells by flow cytometry are described in Table 2. Table 2. Antibodies used for flow cytometry Cell Type Flow cytometry antibodies MDSC PerCP-CyTM 5.5 Mouse Anti-Human CD45 (Ref:564105; BD PharmingenTM) PE-CyTM7 Anti-Human HLA-DR (Ref: 560651; BD PharmigenTM) APC- CyTM7 Rat Anti-CD11b (Ref: 557657; BD PharmigenTM) PE Mouse Anti-Human CD33 (Ref: 555450; BD PharmigenTM ) FITC Mouse Anti-Human CD14 (Ref: 555450; BD PharmigenTM ) APC Mouse Anti-Human CD15 (Ref: 551376; BD PharmigenTM ) Treg cells Human Regulatory T Cell Cocktail (Ref: 560249; BD BDPharmigenTM )a T cells/ CD4 T cells/ CD8 T cells MultitestTM 6-Color TBNK (Ref: 337166; BD Multitest) a. Includes FITC Anti-Human CD4, PE-CyTM7 Anti-Human CD25 y Alexa Fluor® 647 anti-Human CD127  Procedure – The flow cytometer used was FACS CantoTM II. Samples were analyzed per the standard protocol of the Flow Cytometry Laboratory of the Clinical Biochemistry Service at Hospital Virgen Macarena University Hospital. This protocol was as follows: 1. Pipette 50μL of the peripheral blood sample into three different test tubes; one for each cytometric analysis (MDSC, Treg and others lymphocytic subpopulations). 67 2. Add 5μL of the corresponding monoclonal antibodies to each test tube. 3. Vortex and incubate 15 minutes in the dark. 4. Add 2mL of BD FACSTM Lysing Solutions 10X Concentrated into each test tube to lyse red blood cells. 5. Vortex and incubate 3 minutes in the dark. 6. Centrifuge 5 minutes at 1500 rpm and discard supernatant. 7. Wash the cell pellet with 2mL of BD FACSTM Sheath Fluid and agitate with vortex. 8. Centrifuge 5 minutes at 1500 rpm and discard supernatant. 9. Add 0.5mL of BD FACSTM Sheath Fluid and agitate with vortex. Samples were the analyzed in the flow cytometer and up to 100.000 events were recorded for the final analysis.  Results – The software used for data collection was BD FACS DivaTM. The flow cytometer provided relative values. Absolute counts (cel/μL) were obtained by multiplying relative values from the flow cytometer by the total number of leucocytes from complete blood counts. Total MDSC was calculated as the sum of M-MDSC plus PMN-MDSC. Ratio was calculated by dividing CD4 counts by CD8 counts for CD4/CD8 ratio. 3.6.2. Complete Blood Count Test A complete blood count test was performed to every sample in order to assess the absolute count of total leucocytes, neutrophils, total lymphocytes and monocytes. The hematology analyzer XN-1000 Pure (Sysmex Europe GMBH) was used for this purpose. Total lymphocytes counts were obtained from complete blood counts. 68 3.7. STATISTICAL CONSIDERATIONS 3.7.1. Sample Size There were not specific sample size calculations for this specific sub-study. Details of the A Simon minimax two-stage design employed for the size calculation of PANGEA-Breast trial are provided in section PANGEA- Breast STUDY (GEICAM/2015-04) 3.7.2. Analysis Populations Three main analysis populations were defined as follows;  The Intent-to-treat (ITT) Population included all patients who were enrolled into the PANGEA-Breast trial.  The Efficacy Population is a subset of the ITT population and included all the enrolled patients that have measurable disease, have received at least one dose of study treatment and have at least one tumour response assessment according to RECIST v.1.1 criteria (unless a progression or death or unacceptable toxicity had been seen before the first tumour response assessment) and without certain major protocol deviations according to the protocol deviation manual from PANGEA- Breast study protocol.  The Biomarker Population is a subset of the ITT population and included all the patients that had available samples for biomarkers assessment. The biomarker population was the primary population for the analysis of this sub-study. 3.7.3. Statistical Methods An exploratory and descriptive analysis was conducted for each variable in the study. All continuous variables were summarized using the following descriptive statistics: n, mean, median, standard deviation, maximum and minimum and confidence interval (CI) was performed when the information 69 was considered relevant to describe a unique variable. The frequency and percentages of observed levels were reported for all categorical measures. Ninety-five percent confidence intervals were provided for estimates of interest wherever possible. All statistical tests were performed with a significance level of 5%, unless otherwise specified. For the quantitative samples Mann-Whitney-U and Wilcoxon test Mann-Whitney U tests were used. a. Patient Disposition A detailed description of patient disposition was provided by GEICAM. The following items were included:  Total number of patients entered.  Total number of patients enrolled.  Summary of reasons for patients entered, but not enrolled.  Total number of patients treated.  Summary of reasons for patients enrolled, but not treated. b. Patient Characteristics Patient characteristics provided by GEICAM included a summary of the following:  Patient demographics.  Baseline disease characteristics.  Preexisting conditions/secondary conditions.  Prior anti-cancer therapy. c. Efficacy Analysis All efficacy analyses were based on the Efficacy population.Among the efficacy analyses of the PANGEA-Breast trial provided by GEICAM, only the 70 data from the BOR were used. For each patient, BOR was summarized by response category.Information of changes in tumor burden was derived by using local radiologist‘s/investigator‘s assessment. The BOR was defined as the best response achieved by patients from study enrolment until study discontinuation or last time seen assessed by investigator. Patients could have achieved complete response (CR), partial response (PR), stable disease (SD) or progression disease (PD) according to RECIST v.1.1 definitions (see Table 3. Criteria for Response Assessment)163. Table 3. Criteria for Response Assessment Response Definition of response criteria by RECIST v1.1163 CR - Disappearance of all target lesions. - If any pathological lymph nodes, a reduction in short axis to<10 mm. PR - Decreased sum of diameters of target lesions at least in 30%, taking as reference the baseline sum diameters. PD - Increased sum of diameters of target lesions at least in 20%, taking as reference the smallest sum on study. - The sum must also demonstrate an increase in absolute values of at least 5 mm. - The appearance of one or more new lesions is also considered progression. SD - Neither sufficient shrinkage to qualify for PR nor for PD, taking as reference the smallest sum diameters while on study. For the purpose of this sub-study, patients achieving CR, PR or SD of any duration were considered as having reached clinical benefit. Results of ORR and CBR at 24 months were also provided for informative purposes. d. Biomarker Analysis Data analyses were performed using the statistical software SPSS version 26.0 (SPSS Inc., Chicago, IL, USA). 71 The biomarker analyses of the present sub-study were exploratory and primarily made use of descriptive statistical methods. All biomarkers analyses were based on the Biomarker population. Descriptive analyses were conducted to report the distribution of the immune cells of interest along the course of the study treatment. For continuous variables, descriptive statistics including the mean, standard deviation, median, minimum and maximum values were provided. Categorical variables were summarized by numbers and proportions (if possible, the 95%confidence intervals were calculated). Association of the distributions of immune cells population of interest with the outcome to the study treatment strategy, clinical benefit (yes/no), and the disease histology (triple negative/luminal) were explored by running separate analysis for each category. Information to group patients by clinical benefit was obtained from BOR results. Patients reached CR, PR at SD of any duration were grouped into the ‗yes‘ clinical benefit group. Patients on PD or other were grouped into the ‗no‘ clinical benefit group. To note, patients not evaluable for BOR assessment and/or without efficacy assessment of response were considered as having a PD as best overall response for the purpose of this study. In the case of MDSC, results were also compared with data from a cohort of 20 healthy volunteers that were available from studyPI-0502-2014 FPS- 2014 (study population/healthy volunteers). Normality was tested applying Shapiro Wilk and Lilliefors (Kolmogorov- Smirnov) tests. As results did not show a normal distribution, non parametrical tests as Wilcoxon test and Mann-Whitney U test were then performed. For statistical contrast between subgroups NO difference was set as null hypothesis and statistical significance was pre-specified at p≤ 0.05. 72 3.8. STUDY LIMITATIONS Some limitations can be mentioned with respect to this substudy. With regard to the study design, it should be noted the lack of previous clinical trials with the same treatment strategy. Further limitations on this group are related to its exploratory nature; single-arm trial, open design, lack of statistically controlled pre-specified hypothesis and the small sample size. Among the limitations related to the underlying disease are the heterogeneity of the disease, the advanced stage of the study population and the heavily pretreated setting that could preclude providing optimal outcomes. Related to the conduct of the study, the lack of some data due to patients‘ withdrawals and unavailable samples along the course of the study should be noticed. Taking all of the above into consideration, it is likely that the established relationships could not show statistically significant results, unless differences between compared groups would be extraordinarily high. 3.9. ETHICS 3.9.1. Ethics review The study was approved by the independent Institutional Review Board (IRB)/Independent Ethics Committee (IEC) associated with each study centre. Any substantial changes to the study protocol were documented in study protocol amendments and, where required, in a new version of the Clinical Study Protocol (CSP). The amendments were approved by the relevant IEC/IRB and, if applicable, the national regulatory authority, before implementation. Local requirements were followed for revised protocols. The sponsor distributed subsequent amendments and new versions of the CSP to each investigator. 73 3.9.2. Ethical Conduct of the Study This study was performed in accordance with the ethical principles that have their origin in the Declaration of Helsinki and that are consistent with the International Conference of Harmonization (ICH) guidelines on Good Clinical Practice (GCP) and the applicable laws and regulations. 3.9.3. Patient Information and Consent The Principal Investigator at each centre ensured that each patient was given full and adequate oral and written information about the nature, purpose, possible risk and benefits of the study, including any information on the mandatory and optional sampling. Patients were notified that they were free to discontinue from the study at any time, give the opportunity to ask questions and allowed time to consider the information provided. Each patient‘s signed and dated informed consent was obtained and documented before conducting any procedure specifically for the study, including screening/baseline evaluation. A copy of the signed informed consents was provided to the patient and another copy was kept with the study records. 74 75 Figure 5. Overview of Subject Disposition 4. RESULTS 4.1. STUDY PATIENTS 4.1.1. Disposition of Patients The disposition of the patients in this study is summarised in Figure 5. Overview of Subject Dispositionand Table 4.A total of 15 patients were finally not eligible to be enrolled in the study. The most common reasons for screening failure were related to inclusion criteria number 8 (5 patients did not have an adequate organ function- 33.3%), number 7 (3 patients without 0 – 1ECOG PS status- 20.0%) and number 4 (2 patients without documented disease by IHQ-13.3%). Of the 36 patients enrolled into the study, all patients received the study treatment. Table 4. Overview of Patients Disposition (As of CDCO 11Apr 2019) Assessed for eligibility (n=51) Excluded (n=15)  Not meeting inclusion criteria (n=15) Discontinued study treatmenta (n=34)  Progressive disease (n= 29)  Death (n= 2)  Adverse Events (n=1) Treated (n=36) Enrolled (n=36) Study Discontinuationa (n=25)  Death (n= 22)  Lost to FUP (n= 1)  Patient decision (n=1) 82 Table 6. Disease characteristics at baseline (ITT) All enrolled (N=36) ECOG PS, n (%) 0 23 (63.9) 1 12 (33.3) 2 1 (2.8%) Histology type, n (%) Ductal 33 (91.7%) Lobular 2 (5.6%) Other: Squamous Invasive Cancer 1 (2.8%) HER-2, n (%) Negative 36 (100.0%) Hormonal Receptors, n (%) Negative 21 (58.3%) Positive 15 (41.7%) Local Subtype by IHC, n (%) Triple negative 21 (58.3%) Luminal B-like (Her2-) 15 (41.7%) Status at Initial Diagnosis, n (%) M0 32 (88.9%) M1 4 (11.1%) Time since initial diagnosis of BC to enrolment (years) N 36 Mean (SD) 6.83 (7.44) Median 4.19 Min ; Max 0.95 ; 37.17 Time since diagnosis of M1 to enrolment (years) N 36 Mean (SD) 3.15 (3.08) Median 2.33 Min ; Max 0.18 ; 14.04 Target lesions at baseline, n (%) 36 (100.0%) Sum of the longest diameters (mm) N 36 Mean (SD) 72.33 (55.81) Median 63.50 Min ; Max 11.00 ; 279.0 Baseline number of metastatic sites 1 9 (25.0%) 2 14 (38.9%) ≥3 13 (36.1%) Metastatic sites, n (%) Visceral 28 (77.8%) Bone 20 (55.6%) Soft Tissue 5 (13.9%) Lymph Node 13 (36.1%) 83 Breast 2 (5.6%) Skin 4 (11.1%) Previous disease-related therapies are summarized in Table 7. Summary of previous disease-related therapies at baseline (ITT) Table 7. Summary of previous disease-related therapies at baseline (ITT) All enrolled (N=36) Prior metastatic chemotherapy, n (%) 34 (94.4) Prior metastatic hormone therapy, n (%) 19 (52.8) Prior metastatic targeted therapy, n (%) 24 (66.7) Prior Therapies, n (%) CT + HT 7 (19.4) CT + HT + TT 19 (52.8) CT + TT 6 (16.7) CT 4 (11.1) Number of regimens received for metastatic breast cancer, n (%) None 1 (2.8) 1 9 (25.0) 2 2 (5.6) >3 24 (66.6) Median 4.00 Abbreviations: CT=Chemotherapy , HT= Hormone Therapy, TT= Targeted Therapy Almost all of the patients (94.4%) had received a previous treatment with chemotherapy for metastatic disease with a median of 4 numbers of regimens of chemotherapy. Two third of the patients (66.7%) had previously received a targeted therapy and half of the patients (52.8%) had received hormone therapy at baseline. Overall, 52.8% of the patients had previously received the three main types of therapy: chemotherapy, hormone therapy and targeted therapy. These patients had most commonly been treated with analogues of pyrimidines (80.6%), anthracyclines (41.7%) and taxanes (33.3%).The most common hormonal agents received were aromatase inihibitors, as letrozole (38.9%) or exemestane (27.8%), andanti-estrogens, fulvestrant (36.1%) or tamoxifen (25.0%). In the case of target therapies, the most commonly used drugs were anti-angiogenics as bevacizumab (25.0%), m- 84 TOR inhibitors as everolimus (16.7%), CDK4/6 inhibitors as palbociclib(11.1%). 4.1.5. Patients analysed (analysis sets) The analysis sets and the number of patients in each analysis set are summarized in Table 8. Analysis sets Definitions of the analysis sets are given in Section 3.7.2. Analysis Populations. Table 8. Analysis sets Number (%) of patients Patients enrolled 36 (100%) Patients included in the ITT population 36 (100%) Patients included in the Biomarker population 35 (100%) Patients included in the Efficacy population 33 (91.7%) a. Percentages were calculated from number of patients enrolled As planned, all the data collected from patients enrolled and treated with available samples for biomarkers analysis in peripheral blood were analysed for the purpose of this sub-study, even if a tumour assessment was not available. 4.2. EFFICACY EVALUATION Efficacy analyses were performed in the Efficacy population. At the time of CDCO, ORR was 15.2% (95% CI5.1-31.9) with none of the patients reached CR, 15.2% of the patients reached PR, 36.4% remained in SD. From the patients who reached SD as BOR, there was only one patient who maintained this outcome for at least 24 weeks. Overall, progressive disease (PD) as best overall response was reported for 36.4% of the patients. Results showed a CBR including SD of any duration of 51.5% (95% CI 33.5, 69.2) and a median DOR of 4.3 months (95% CI: 2.3, 7.4). Best overall response (BOR) results are summarised in Table 9. Best objective response (Efficacy population) 85 Table 9. Best objective response (Efficacy population) Best Objective Response Number (%) of patients (N=33) Yes CB CR 0 PR 5 (15.2%) SD 12 (36.4%) SD at 24 months 1 (3.0%) No CB Progression 12 (36.4%) No post baseline response assessment 1 (3.0%)a Unable to assess 3 (9.1%)b Abbreviations: CB= Clinical Benefit Note: Patients ESP0032005 and ESP0031003 reached SD but were excluded from Efficacy populations. Both patients are included in the biomarker population. a. PatientESP0044005: Patient had a clinical progression and after that she died. b. Patient ESP0031001: End of treatment by progression. Patient ESP0051006: End of treatment by Death. Patient ESP0055001: End of treatment and study by Adverse Event. It should be noted that 1 patient did not have any response assessment post baseline and for 3 patients the response could not be assessed (9.1%). From these 4 patients, 3 patients left the study before the first radiological assessment due to clinical progression or death after one or two cycles of study medication. The remaining patient discontinued after one cycle due to toxicity, being the adverse event classified as ‗respiratory failure‘ grade 3 by the investigator. The event was considered not related to study medication and the outcome of the event was unknown. 4.3. BIOMARKER EVALUATION Cell analyses were performed in the Biomarker population. All the samples received were analysed for biomarker assessment. A total number of 34, 30 and 23 samples from baseline, cycle 3 and cycle 6 (or end of treatment), respectively, were received. Study discontinuation, study withdrawal or protocol deviations were the main reasons why the sample sets were incomplete. 86 4.3.1. MDSC Analyses a. MDSC in comparison with healthy volunteers Baseline counts of all the subgroups of MDSC from patients in the PANGEA study were compared to 20 samples from healthy volunteers without known current or previous history of cancer. Results are shown in Table 10. Median results showed that the levels of MDSC of healthy volunteers were lower than those assessed for patients in the PANGEA trial; median count for total MDSC in patients was more of 2 fold higher than in the healthy volunteers group (44.46 vs.17.53 cel/μL, patients vs. healthy volunteers, respectively). This was seen in all of the subtypes of MDSC. Differences between median values were statistically significant for the mononuclear subpopulation and for the total MDSC population. Table 10. MDSC values of the Study Population compared to Healthy volunteers MDSC at baseline M-MDSC (cel/μL) PMN-MDSC (cel/μL) Total MDSC (cel/μL) SP (N= 35) HV (N=20) SP (N= 35) HV (N=20) SP (N= 35) HV (N=20) n (%) 34 (97) 20 (100) 34 (97) 20 (100) 34 (97) 20 (100) Missing 1 0 1 0 1 0 Mean (SD) 65.50 (133.57) 15.87 (12.12) 22.99 (42.45) 9.28 (12.55) 88.49 (160.50) 25.15 (21.63) Median 33.46 10.28 8.65 4.99 44.46 17.53 Q1:Q3 17.99: 58.71 7.74: 25.22 3.23: 25.72 2.52: 8.74 24.33: 79.27 12.37: 29.32 Min; Max 1.56; 787.60 0.70; 50.43 0.18; 228.40 1.17; 54.43 2.64; 897.70 1.88; 84.20 p-valuea, b (2-sided) 0.001 0.185 0.002 Abbreviations: M-MDSC= Mononuclear Myeloid-derived suppressor cells, PMNMDSC=Polymorphonuclear Myeloid-derived suppressor cells, SP= Study patients, HV= Healthy Volunteers, SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum a. Between groups b. The analysis was performed using U de Mann-Whitney for independent samples 87 These results are also presented in Figure 6. MDSC values of the Study Population compared to Healthy volunteersIn this graph, a higher dispersion of the individual values was seen in the group of the study patients compared to the healthy volunteers. This dispersion is general to the two different subpopulations of MDSC. Figure 6. MDSC values of the Study Population compared to Healthy volunteers b. General description of MDSC The numbers of MDSCs present in the collected samples were counted, including M-MDSC, PMN-MDSC and total MDSC. Results by time point are presented in Table 11. Overall, results showed a drop in the median count of the total population of MDSC along the time. Median values of total MDSC ranged from 44.46 to 19.33 cel/μL at baseline and at cycle 6, respectively. Drop was more notorious and sharp from baseline to cycle 3, being maintained thereafter. Abbreviations: M-MDSC= Mononuclear Myeloid-derived suppressor cells, PMNMDSC=Polymorphonuclear Myeloid-derived suppressor cells, SP= Study patients, HV= Healthy Volunteers 88 Differences in counts along the course of the treatment were not statistically significant. 89 Table 11. General description of MDSC by MDSC subpopulation (Biomarker Population) Biomarker Population (N= 35) M-MDSC (cel/μL) PMN-MDSC (cel/μL) Total MDSC (cel/μL) Baseline Cycle 3 Cycle6 Baseline Cycle 3 Cycle6 Baseline Cycle 3 Cycle6 n (%) 34 (97) 30 (86) 23 (66) 34(97) 30(86) 23(66) 34(97) 30(86) 23(66) Missing 1 5 12 1 5 12 1 5 12 Mean (SD) 65.50 (133.57) 69.34 (155.05) 21.31 (26.27) 22.99 (42.45) 16.12 (33.48) 10.41 (9.18) 88.49 (160.50) 85.46 (179.74) 31.72 (28.51) Median 33.46 13.31 11.17 8.65 5.04 11.09 44.46 20.41 19.33 Q1:Q3 17.99:58.71 4.08:60.61 4.12:28.63 3.23:25.72 1.96:12.48 2.90:12.60 24.33:79.27 6.72: 65.41 9.65:52.36 Min; Max 1.56;787.60 0.78;712.10 0.67;104.50 0.18;228.40 0.45;175.70 0.20;41.19 2.64;897.70 1.90;757.10 4.55; 116.90 p-valuea,b (2-sided) 0.058 0.548 0.108 Abbreviations: M-MDSC = Mononuclear Myeloid-derived suppressor cells, PMN-MDSC=Polymorphonuclear Myeloid-derived suppressor cells, SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum. a. Between baseline and Cycle 6 samples. b. The analysis was performed using a Wilcoxon test for related samples. 90 The trend seemed to be mostly driven by the M-MDSC sub-population as the proportion of M-MDSC was higher compared to the proportion of PMN-MDSC at baseline. Median counts obtained for PMN-MDSC sub-population barely moved and seemed to present a different trend along the study treatment. Figure 7 represents the variation of the two subpopulations and the total count of MDSC in the three different time points along the course of the study. As previously discussed, there was a notorious drop in the total count of MDSC from baseline to cycle 3 maintained thereafter. This seemed to be mainly due to the proportion of the M-MDSC as this subgroup constitutes the majority group of this group of cells compared to PMN-MDSC. Figure 7. General description of MDSC by MDSC subpopulation (Biomarker Population) c. MDSC by clinical benefit Total MDSC values were assessed by the clinical outcome presented by the patients in the study. Results are presented in Table 12. Abbreviations: B= Baseline, C3= cycle 3, C6 = cycle 6, M-MDSC = Mononuclear Myeloidderived suppressor cells, PMN-MDSC=Polymorphonuclear Myeloid-derived suppressor cells 91 Overall, there was an important decrease in the median count of MDSCs for the subpopulation of patients that obtained a clinical benefit along the time. This decrease is more noticeable from baseline to cycle 3. Levels at cycle 3 were maintained in cycle 6. This is in line with the main analysis. For the subpopulation that presented progressive disease, a decreased was seen from baseline compared to cycle 6. Median counts from baseline to cycle 3 were mostly constant; median count minimally increased from 57.01 to 60.66 cel/μL. Values decreased dramatically from cycle 3 to cycle 6. Table 12. Total MDSC by clinical benefit (Biomarker Population) Biomarker Population (N= 35) Total MDSC (cel/μL) CB (N=19) PD (N=16) Basal Cycle 3 Cycle6 Basal Cycle 3 Cycle6 n (%) 19 (100) 19 (100) 18 (95) 15 (94) 11 (69) 5 (31) Missing 0 0 1 1 5 11 Mean (SD) 56.70 (85.93) 31.57 (41.04) 27.72 (23.27) 128.8 (219.40) 154.3 (237.00) 46.13 (42.86) Median 38.62 17.38 17.82 57.01 60.66 23.16 Q1:Q3 16.52: 60.88 5.61: 36.42 10.64: 38.81 33.96: 128.90 11.94: 259.30 19.33: 56.50 Min; Max 2.64; 396.60 3.14; 169.80 4.55; 83.67 16.78; 897.70 1.90; 757.10 14.75; 116.90 p-valuea, b (2-sided) 0.078 0.715 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Figure 8 (see below) presents comparative results for both subgroups by time point. Results showed that patients with progressive disease presented a higher median level of MDSC at baseline than patients who reached clinical benefit; median counts were of 38.62 cel/μL vs. 57.01 cel/μL for patients that reached 98 In the analysis of PMN-MDSC, results showed increased counts from baseline to cycle 6 in the luminal subpopulation and bare changes in the triple negative population. Overall, triple negative PMN-MDSC median counts were higher compared to luminal population at baseline but reached comparable magnitude at cycle 6. None of these differences were statistically significant; p-values (per Mann-Whitney U test) between both groups at baseline, cycle 3 and cycle 6 were 0.259, 0.439 and 0.833, respectively Table 17. PMN-MDSC by histology (Biomarker Population) Abbreviations: TN = Triple Negative, L= Luminal, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Biomarker Population (N= 35) PMN-MDSC (cel/μL) TN (N=21) L (N=14) Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 20 (95) 18 (86) 12 (57) 14 (100) 12 (86) 11 (79) Missing 1 3 9 0 2 3 Mean (SD) 31.44 (53.33) 21.78 (42.26) 9.37 (5.58) 10.92 (12.24) 7.63 (8.08) 11.54 (12.18) Median 12.17 6.29 10.71 5.86 3.67 11.81 Q1:Q3 4.02: 35.39 3.27: 17.02 4.48: 12.19 2.73: 15.84 1.42: 12.19 1.17: 17.68 Min; Max 0.18; 228.40 0.68; 175.74 1.45; 21.61 0.53; 40.67 0.45; 27.21 0.20; 41.19 99 4.3.2. Lymphocytes Analyses a. General description of lymphocytes Absolute white blood cells counts were assessed in the samples received. Results are shown in Table 18. and Figure 10. Overall, the median values of lymphocytes were in within the normal ranges at the different time points assessed. With respect to the fluctuation of values along the course of treatment, results showed a continuous decrease from baseline until cycle 6. Median counts dropped from 1290 cel/μL to 1190 cel/μL. This decrease was not statistically significant. Table 18. General description of total lymphocytes (Biomarker population) Biomarker Population (N= 35) Lymphocytes (cel/μL) Baseline Cycle 3 Cycle 6 n (%) 34 (97) 30 (86) 23(66) Missing 1 5 12 Mean (SD) 1407.94 (781.02) 1330.33 (577.54) 1191.30 (530.49) Median 1290.00 1280.00 1190.00 Q1:Q3 940.00: 1585.00 925.00: 1757.50 760.00: 1590.00 Min; Max 290; 4900 330; 2560 480; 2380 p-valuea, b (2-sided) 0.053 p-valuea,b (2-sided) 0.859 0.286 Abbreviations: TN = Triple Negative, L= Luminal, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. 100 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum. Notes: all the counts are in cel/µL a. Between baseline and Cycle 6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Figure 10. General description of total lymphocytes (Biomarker population) b. Lymphocytes by clinical benefit Lymphocytes counts were assessed by the clinical outcome and results are in Table 19. Overall, median counts at baseline were mostly within a similar magnitude for the subpopulation who reached clinical benefit. Although this difference was not statistically significant, results showed a decrease of the mean values from baseline to cycle 6 in the population that experienced a clinical benefit. Median counts varied from 138 cel/μL at baseline to 1205 cel/μL at cycle 6. Table 19. Total lymphocytes by clinical benefit (Biomarker population) Biomarker Population (N= 35) Lymphocytes (cel/μL) 101 CB (N=19) PD (N=16) Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 19 (100) 19 (100) 18 (95) 15 (94) 11 (69) 5 (31) Missing 0 0 1 1 5 11 Mean (SD) 1528.95 (942.21) 1325.79 (637.40) 1219.44 (574.65) 1254.67 (501.37) 1338.18 (485.69) 1090 (357.77) Median 1380 1220 1205 1150 1470 970 Q1:Q3 950: 1850 800: 1810 717.50: 1632.50 835: 1465 1060: 1500 790: 1450 Min; Max 640; 4900 410; 2560 480; 2380 290; 2290 330; 2110 730; 1610 p-valuea, b (2-sided) 0.071 0.465 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Similarly, a non statistically significant decrease from baseline to cycle 6 was also observed in the patients who did not obtained a clinical benefit; lymphocyte count decreased from 1150 cel/μL at baseline to 970 cel/μL (lymphopenia CTCAE grade 1) at cycle 6. However, the trend was different as the median count at cycle 3 (1470 cel/μL) increased when compared to median values at baseline. Figure 11presents comparative results of both subgroups per time point. Overall, baseline median counts were higher for the subpopulation who reached the clinical benefit compared to the population who progressed. However, this difference was not maintained at the following time point, were median counts were greater for the group of patients who progressed compared to the group who reached the clinical befit. At cycle 6, the difference reverted again with median counts being greater for the clinical benefit subgroup. None of these differences were statistically significant. 102 Figure 11. Total lymphocytes by clinical benefit (Biomarker population) c. Lymphocytes by histology Lymphocytes were also assessed by the histology of the breast cancer disease (see Table 20). In the case of triple negative disease, median values increased from baseline (1045 cel/μL) to cycle 3 (1220 cel/μL) and remained at mostly the same level thereafter (1205 cel/μL). No statistically significance difference was found when comparing median values from baseline to cycle 6. Results showed a decrease in the median values from baseline (1565 cel/μL) until cycle 6 (970 cel/μL) in the luminal subpopulation. This decrease is within normal ranges until cycle 6 where median values correlated with lymphopenia CTCAE v5.0 grade 1. Overall, the observed difference in median values from baseline to cycle 6 was not statistically significant. Abbreviations: CB= Clinical Benefit, PD= Progressive Disease Note: p-values (per Mann-Whitney U test) between both groups at baseline, cycle 3 and cycle 6 were 0.560, 0.832 and 0.857, respectively. 103 Table 20. Total lymphocytes by histology(Biomarker population) BK Population (N= 35) Lymphocytes (cel/μL) TN (N=21) L (N=14) Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 20 (95) 18 (86) 12 (57) 14 (100) 12 (86) 11 (79) Missing 1 3 9 0 2 3 Mean (SD) 1147 (485.03) 1235 (554.15) 1140.83 (415.37) 1780.71 (974.78) 1473.33 (606.31) 1246.36 (650.34) Median 1045 1220 1205 1565 1385 970 Q1:Q3 832.50: 1447.50 815: 1560 737.50: 1547.50 1247.50: 1910 1085: 1907.50 810: 1910 Min; Max 290; 2330 330; 2240 560; 1760 880; 4900 410; 2560 480; 2380 p-valuea, b (2-sided) 0.374 0.091 Abbreviations: TN = Triple Negative, L= Luminal, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Comparative results per time point presented in Figure 12 showed that median counts were lower for triple negative patients than for luminal patients at baseline. This difference was statistically significant (p-value = 0.005). This difference was maintained in cycle 3 but reverted in cycle 6. In both cases, differences did not show statistical significance. 104 Figure 12. Total lymphocytes by histology (Biomarker population) Abbreviations: TN= Triple Negative, L= Luminal Note: p-values (per Mann-Whitney U test) between both groups at baseline, cycle 3 and cycle 6 were 0.005, 0.368 and 0.928, respectively. 105 4.3.3. Regulatory T cells Analyses a. General description of Treg cells Results for Treg cells count are shown in Table 21. Overall, a statistically significant decrease of the median values of Treg from baseline to cycle 6 was seen. Median counts minimally changed from baseline to cycle 3. Table 21. General description of Treg cells (Biomarker population) Biomarker Population (N= 35) Treg cells (cel/μL) Baseline Cycle 3 Cycle 6 n (%) 34 (97) 30 (86) 23(66) Missing 1 5 12 Mean (SD) 36.83 (22.133) 38.53 (29.802) 26.40 (18.361) Median 29.98 31.02 23.26 Q1:Q3 21.83: 45.99 16.92: 51.04 13.96: 36.10 Min; Max 7.55; 91.74 2.32; 140.15 1.75; 85.39 p-valuea, b (2-sided); 0.028 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum. Notes: all the counts are in cel/µL a. Between baseline and Cycle 6 samples. b. The analysis was performed using a Wilcoxon test for related samples. A graphical representation of these results is provided in Figure 13, where a descending trend of the medians counts from baseline to cycle 6 is evidenced. 106 Figure 13. General description of Treg cells (Biomarker population) b. Treg cells by clinical benefit Treg lymphocytes were also assessed by clinical benefit. Results are shown in Table 22. Overall, median counts of Treg cells at baseline were of similar magnitude for the population who reached clinical benefit and the population who progressed. For those patients who reached a clinical benefit, results showed a decreased median count of Treg along the course of the treatment with median values varying from 30.48 cel/μL at baseline to 24.71 cel/μL at cycle 6. This decrease is more evident between cycle 3 and cycle 6 values, as median counts barely changed between baseline and cycle 3. Overall, median counts also decreased from baseline to cycle 6 in the group of patients who progressed. However, there was not a clear trend observed as mean values initially increased from baseline (28.63 cel/μL) to cycle 3 (31.91 cel/μL) and decreased from cycle 3 to cycle 6 (15.71 cel/μL). 107 Table 22. Treg lymphocytes count by clinical benefit (Biomarker population) Biomarker Population (N= 35) Treg cells (cel/μL) CB (N=19) PD (N=16) Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 19 (100) 19 (100) 18 (95) 15 (94) 11 (69) 5 (31) Missing 0 0 1 1 5 11 Mean (SD) 36.68 (20.99) 37.27 (25.03) 25.54 (13.77) 37.02 (24.25) 40.70 (37.948 29.50 (32.15) Median 30.48 30.14 24.71 28.63 31.91 15.71 Q1:Q3 21.99: 45.46 14.47: 49.00 14.29: 36.55 21.05: 47.57 19.37: 57.18 9.40: 56.49 Min; Max 11.48; 85.20 5.70; 94.95 1.75; 49.31 7.55; 91.74 2.32; 140.15 7.55; 85.39 p-valuea, b (2-sided); 0.053 0.144 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Obtained results in both groups per time point are compared in Figure 14. There were minimal differences of median counts between both subgroups at baseline and cycle 3. At cycle 6, median counts were higher for the clinical benefit subgroup compared to the progressive disease subgroup, although this difference was not statistically significant. 114 Table 26. CD4 T cells counts by histology (Biomarker population) Biomarker Population (N= 35) CD4 T cells (cel/μL) TN (N=21) L (N=14) Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 20 (95) 18 (86) 12 (57) 14 (100) 12 (86) 11 (79) Missing 1 3 9 0 2 3 Mean (SD) 431.83 (227.88) 478.29 (293.87) 389.28 (229.57) 603.41 (214.11) 517.84 (231.13) 401.09 (207.22) Median 386.65 379.75 343.20 620 509.45 326.40 Q1:Q3 249.10: 650.93 242.00: 657.70 185.93: 609.08 378.15: 753.63 325.88: 716.63 218.40: 541.20 Min; Max 60.90; 926.10 19.80; 1139.40 112.10; 821.10 360.80; 980 147.60; 934.20 148.80; 802.20 p-valuea, b (2-sided) 0.594 0.033 Abbreviations: TN = Triple Negative, L= Luminal, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Comparative results by time points are also presented in Figure 18. Baseline median counts were higher for the luminal subpopulation compared to the triple negative subpopulation. This difference was statistically significant (p-value = 0.039). Although no clinically significant, same trend was maintained at cycle 3 but not at cycle 6, where median values were relatively of similar magnitude. 115 Figure 18. CD4 T cells counts by histology (Biomarker population) 4.3.5. CD8 T cells Analyses a. General description of CD8 T cells Results for CD8 T cells counts are presented in Table 27 and in Figure 19. Overall, a constant decrease of the median values of CD8 T cells from baseline to cycle 6 was observed; median counts passed from 377.25 cel/μL at baseline to 233.70 cel/μL at cycle 6. This decrease was not statistically significant. Abbreviations: TN= Triple Negative, L= Luminal Note: p-values (per Mann-Whitney U test) between both groups at baseline, cycle 3 and cycle 6 were 0.039, 0.692 and 0.833, respectively. 116 Table 27. General description of CD8 T cells (Biomarker population) Biomarker Population (N= 35) CD8 T cells (cel/μL) Baseline Cycle 3 Cycle 6 n (%) 34 (97) 30 (86) 23(66) Missing 1 5 12 Mean (SD) 461.43 (506.66) 412.69 (337.19) 399.33 (350.05) Median 377.25 313.20 233.70 Q1:Q3 221.25: 486.90 172.50: 506.30 161.50: 563.50 Min; Max 63.80; 3038.00 62.70; 1587.20 22.80; 1370.20 p-valuea, b (2-sided) 0.408 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum. Notes: all the counts are in cel/µL a. Between baseline and Cycle 6 samples. b. The analysis was performed using a Wilcoxon test for related samples. 117 Figure 19. General description of CD8 T cells (Biomarker population) b. CD8 T cells by clinical benefit Analysis of CD8 T cells by clinical benefit is showed in Table 28. Overall, median counts of CD8 T cells dropped from baseline along the course of the treatment in the subgroup that reached the clinical benefit; median counts changed from 412.00 cel/μL to 254.55 cel/μL from baseline to cycle 6, respectively. This difference was not statistically significant. In the case of the subgroup of patients who progressed there were a slightly decrease from baseline to cycle 6 with median counts from 296.80 cel/μL to 226.30 cel/μL, respectively. However, median counts at cycle 3 (323.40 cel/μL) increased from baseline. Table 28. CD8 T cells by clinical benefit (Biomarker population) Biomarker Population (N= 35) CD8 T cells (cel/μL) CB (N=19) PD (N=16) 118 Baseline Cycle 3 Cycle6 Baseline Cycle 3 Cycle6 n (%) 19 (100) 19 (100) 18 (95) 15 (94) 11 (69) 5 (31) Missing 0 0 1 1 5 11 Mean (SD) 526.41 (626.11) 415.98 (381.31) 430.14 (384.44) 379.13 (296.78) 407.00 (260.69) 288.40 (161.73) Median 412.00 294.00 254.55 296.80 323.40 226.30 Q1:Q3 237.50: 507.60 166.60: 494.00 155.18: 659.20 203.40: 480.00 253.00: 543.20 177.75: 430.10 Min; Max 176.40; 3038.00 112.70; 1587.20 22.80; 1370.20 63.80; 1305.30 62.70; 953.60 161.50; 563.50 p-valuea,b (2-sided) 0.327 0.465 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. As showed in Figure 20, when comparing results in both subgroups by time point, baseline median counts for the patients that reached clinical benefit were greater compared to the patients who progressed. This difference was not maintained at cycle 3, where median counts were greater for the patients who progressed. At cycle 6, median counts were similar in both subgroups. None of these differences were statistically significant. 119 Figure 20. CD8 T cells by clinical benefit (Biomarker population) c. CD8 T cells by histology Analysis of CD8 T cells by histology is presented in Table 29. In the triple negative group, CD8 T cell counts decreased along the course of treatment. Median values ranged from 338.90 cel/μL at baseline to 261.50 cel/μL at cycle 6. This drop did not show statistical significance. Same trend was observed in the luminal group. Median counts changed from 456.80 cel/μL at baseline to 233.70 cel/μL at cycle 6. This difference was neither statistically significant. Table 29. CD8 T cells counts by histology(Biomarker population) Biomarker Population (N= 35) CD8 T cells (cel/μL) TN (N=21) L (N=14) Abbreviations: CB= Clinical Benefit, PD= Progressive Disease Note: p-values (per Mann-Whitney U test) between both groups at baseline, cycle 3 and cycle 6 were 0.302, 0.582 and 0.914, respectively. 120 Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 20 (95) 18 (86) 12 (57) 14 (100) 12 (86) 11 (79) Missing 1 3 9 0 2 3 Mean (SD) 347.00 (163.04) 382.01 (262.09) 367.65 (257.429) 624.91 (750.90) 458.72 (435.60) 433.89 (440.64) Median 338.90 313.20 261.50 456.80 308.70 233.70 Q1:Q3 214.85: 433.35 186.60: 539.20 161.48: 629.28 221.25: 528.63 166.95: 530.90 161.50: 515.70 Min; Max 63.80; 722.30 62.70; 953.60 22.80; 795.20 155.10;303 8.00 121.10; 1587.20 117.60; 1370.20 p-valuea, b (2-sided) 0.790 0.374 Abbreviations: TN = Triple Negative, L= Luminal, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Comparative results by subgroup per time point are shown Figure 21. Baseline median counts were greater for the luminal population compared to the triple negative population. This difference was not maintained at subsequent time point as at cycle 3 median values were mostly similar for both groups. At cycle 6, median counts were slightly greater for the triple negative population compared to the luminal population. None of the cases showed statistically significant differences. 121 Figure 21. CD8 T cells counts by histology (Biomarker population) 4.3.6. CD4/CD8 T cells Ratio Analyses a. General description of CD4/ CD8 T cells ratio The ratio between CD4 and CD8 lymphocytes count was also determined. Results are shown in Table 30 and Figure 22. The baseline median count of CD4 cells was higher than the median count for CD8 cells which is maintained along the course of treatment. Median values of the ratio were always >1, meaning the initial proportion between both cell populations was therefore maintained. Abbreviations: TN= Triple Negative, L= Luminal Note: p-values (per Mann-Whitney U test) between both groups at baseline, cycle 3 and cycle 6 were 0.245, 0.950 and 0.928, respectively. 122 Table 30. General description of CD4/CD8 T cells ratio (Biomarker population) Biomarker Population (N= 35) CD4/CD8 T cellsratio (cel/μL) Baseline Cycle 3 Cycle 6 n (%) 34 (97) 30 (86) 23(66) Missing 1 5 12 Mean (SD) 1.52 (1.08) 1.62 (1.48) 1.49 (1.16) Median 1.37 1.19 1.31 Q1:Q3 0.86; 1.73 0.93; 1.76 0.71; 1.57 Min; Max 0.32; 5.25 0.32; 7.71 0.34; 5.67 p-valuea, b (2-sided) 0.709 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum. Notes: all the counts are in cel/µL a. Between baseline and Cycle 6 samples. b. The analysis was performed using a Wilcoxon test for related samples. Figure 22. General description of the CD4/ CD8 ratio (Biomarker population) 123 A constant decrease of the median values was seen from baseline to cycle 6 in both cells subtypes. However, no clear trend was observed in the median values of the ratios obtained along the course of the treatment. Mostly no difference was found between median values at baseline and at cycle 6. Median values showed an initial increase from baseline (1.37 cel/μL) to cycle 3 (1.19 cel/μL) and a decrease from cycle 3 to cycle 6 (1.31 cel/μL). b. CD4/CD8 T cells ratio by clinical benefit CD4/CD8 ratio was also assessed by clinical benefit (see Table 31). In both populations, the proportion of CD4 was higher than the proportion of CD8, as the ratio was always over 1. However, in both scenarios minimum values showed ratios under 1, suggesting the CD4/CD8 proportions were different for some patients. Table 31. CD4/ CD8 ratio by clinical benefit (Biomarker population) Biomarker Population (N= 35) CD4/CD8 T cells ratio (cel/μL) CB (N=19) PD (N=16) Baseline Cycle 3 Cycle 6 Baseline Cycle 3 Cycle 6 n (%) 19 (100) 19 (100) 18 (95) 15 (94) 11 (69) 5 (31) Missing 0 0 1 1 5 11 Mean (SD) 1.48 (1.14) 1.77 (1.74) 1.56 (1.30) 1.58 (1.03) 1.37 (0.89) 1.23 (0.29) Median 1.21 1.20 1.30 1.57 1.19 1.35 Q1:Q3 0.86: 1.52 0.98: 1.86 0.67: 1.74 0.86: 1.88 0.87: 1.50 0.99: 1.40 Min; Max 0.32; 5.25 0.35; 7.71 0.34; 5.67 0.35; 4.09 0.32; 3.22 0.71; 1.46 p-valuea, b (2-sided) 0.845 0.715 Abbreviations: SD = Standard deviation, Q=Quartile, Min= Minimum, Max=Maximum, CB= Clinical Benefit, PD= Progressive Disease a. Between baseline and C6 samples. b. The analysis was performed using a Wilcoxon test for related samples. 130 Patients were treated with a combination of gemcitabine and pembrolizumab IV at 1200 mg/m2 and 200 mg, respectively, as per PANGEA-Breast trial protocol. The main objective of this sub-study was to explore the role of a set of immune biomarkers in peripheral blood by monitoring the fluctuations of their values alongside the treatment course. The set of immune markers analysed were MDSC, lymphocytes, Treg cells, CD4 T cells, CD8 T cells and CD4/CD8 ratio. These immune biomarkers were selected based on the evidence supporting their potential predictive or prognosis role. Limitations of the study related to the design, sample size and methodology used were already discussed in section 3.8 (STUDY LIMITATIONS) Related to the study population, 36 patients were enrolled and treated in the PANGEA Breast trial. Patients were recruited from nine different clinical sites in Spain. At the time of the CDCO (11 Apr 2019), only 2 patients (5.6%) remained on study treatment. The majority of study patients (80.6%) discontinued from study drugs due to disease progression. A total of 25 patients (69.4%) had left the study at CDCO. The most common cause of study discontinuation was death (61.1%). The original version of the study protocol (v. 1.0) dated from the 04 of September 2015. The study was amended six times being the latest three versions the most relevant ones. For instance, amendment protocol v.4.0 from 27 of June 2017, modified alkaline phosphatase upper limit on inclusion criteria number 8 to facilitate the inclusion of patients with hepatic, bone or disease metastases. Furthermore, amendment protocol v.6.0 from 19 of March 2018 modified inclusion criteria restricting the ECOG performance status up to score 1 and adding a maximum number of previous treatments lines of chemotherapy for advance breast cancer treatment. These latest modifications were performed as ECOG status and 131 previous treatment lines could negatively influence the outcomes to the study therapy. These restrictions resulted in a study population relatively young with a median age of 51.56 years being the majority of the patients younger than 65 years old; only 3 (8.3%) patients were equal or older to 65 years old. This age distribution could not be considered representative of the overall population with advance breast cancer in a late stage of the disease base on the low representation of equal or older to 65 years old patients103. The majority of the patients were in a relatively good condition with ECOG PS scores between 0 and 1. There is only one patient who presented an ECOG PS score of 2 as this was allowed in previous protocol versions. Related to the disease characteristics, the majority of the study patients presented a ductal type disease (91.7%). All the patients were confirmed to be HER2 negative. The population could be considered well balanced in terms of negative and positive hormonal receptors; 58.3% of the patients were triple negative and 41.7% were luminal B-like (HER2-). All the patients were metastatic at study entry. However, the majority of the patients were not metastatic at diagnosis with a median time from initial diagnosis to enrolment of 6.83 years and a median time from diagnosis of metastatic disease to enrolment 3.15 years. This seemed to be in line with reported data and to define a population that had lived with the disease for a long time where the disease has likely evaded the immune system and has extended129,143,164. The study population has a high burden of disease with a median sum of longest diameters of 63.50 mm. The majority of the patients presented equal or more than two sites of metastasis being the most common types: visceral, bone and lymph node. This is considered representative of a late stage of the disease129,143. As a consequence of the late stage of the disease, the study population was heavily pre-treated with a median of 4 regimens of chemotherapy received; 132 most of the patients had already been treated with chemotherapy, hormone therapy and targeted therapies. In this scenario with an advanced disease at a late stage and heavily pretreated, chances to reach any positive outcome may be really reduced. The disease is highly developed and may be completely invisible to the immune system where the process of immune edition is increased due to the potential immune changes caused by different treatment strategies44. From the total 36 enrolled and treated patients; i)only 33 patients were evaluable for efficacy with available assessments of BOR, and ii) 35 patients were analysed for biomarkers. The remaining patient withdraw informed consent for biomarkers. Relative to efficacy results, ORR was of 15.2% (95% CI: 5.1-31.9) with a median duration of response of 4.3 months (95% CI: 2.3, 7.4). This result was driven by the proportion of patients who reached partial response (15.2%) as none of the patients reached complete response. Overall, the majority of the patients remained in stable disease (36.4%) as BOR with only one patient maintaining this outcome form at least 24 weeks. The CBR including stable disease of any duration was 51.5% (95% CI 33.5, 69.2). Progressive disease (PD) as best overall response was reported for 36.4% of the patients These results are low considered compared to recent results from IMpassion130. This could be explained taking into account that this is a heavily pretreated population and not selected for PD-L1 expression143,165. To note, study was considered issue of early termination at the futility analysis due to lack of efficacy. With regards to biomarker analyses, a set of biomarkers in peripheral blood were initially planned to be analysed from patients participating in the PANGEA study. Samples were collected from three different time points; baseline, cycle 3 and cycle 6 or end of treatment. However, end of 133 treatment samples were not assessed as cycle 6. This is an additional limitation of the study that may hamper comparative description. All available samples were assessed. A general description of the proportion of these cells along the course of the treatment together with additional comparative analyses per clinical benefit and histology were provided. As previously discussed patients not evaluable for BOR assessment and/or without efficacy assessment of response were considered as progressions for the purpose of the clinical benefit analyses. This could be also seen as an added limitation to the study as these patients only have available samples from baseline. It should also be noted that the 35 patients were included in biomarker population but not all samples sets were complete; a total of 34, 30 and 23 samples at baseline, cycle 3 and cycle 6, respectively, were finally analysed. This limitation renders comparison groups not always balanced in terms of number of patients across the three different time points. Overall, results showed a general depletion of all the cells populations studied along the observation period; total MDSCs, lymphocytes, Treg cells, CD4 T cells, CD8 T cells and the CD4/CD8 ratios. Statistical significance was only reached for Treg cells and CD4 T cells. When assessing results by clinical outcome, first, baseline median counts of total MDSC, M-MDSC, PMN-MDSC, Lymphocytes, CD4, CD8 CD4/CD8 ratio were greater for the subgroup of patients who progressed compared to the ones who reached clinical benefit. Statistical significance was only reached for M-MDSC. Second, similarly to findings in the general analyses, a decrease was observed in total MDSC, M-MDSC, lymphocytes, Treg cells, CD4 T cells and CD8 T cells in the subgroup of patients who reached clinical benefit. Statistical significance was only reached for CD4 T cells. Third, in the subgroup of patients who progressed interpretation of the results is hampered by the sample lost specifically at cycle 6. Furthermore, 134 samples did not differentiate between pre-specified time points or end of treatment; this is, if a patient progressed at cycle 3 the sample was assessed as a cycle 3 time point sample and not as end of treatment sample. Although, time to response (TTR) was not a selected endpoint for this study, it should be noted that responses to ICIs are normally observed between the first two to three months; broadly, medians time to response reported for pembrolizumab and nivolumab varied between 1.9 - 3.5months38,39. In this sense, it could be seen reasonable to think that cycle 3 time point could be more valuable for the subgroup of patients who progressed, and so, counts between baseline to cycle 3 could also help to understand the behaviour of cells count in this subgroups. For instance, an increased or no change was seen for total MDSC, M-MDSC, lymphocytes, Tregs, CD4, CD8 from baseline to cycle 3. Unfortunately, differences between baseline and cycle 3 were not tested for statistical significance. To note, cycle 6 time point could be seen as a valuable end-point to monitor durable responses but also to provide opportunities to late responders. With regards to the analysis by histology, results showed; first, increased baseline median counts for total MDSC, M-MDSC, PMN-MDSC, CD4/CD8 ratio was seen for triple negative compared to the luminal subgroup. Statistical significance was only reached for total MDSC, M-MDSC and CD4/CD8 ratio. Second, increased baseline median counts for lymphocytes, Treg, CD4, CD8 was seen for luminal compared to the TN subgroup. Statistical significance was only reached for lymphocytes and CD4. Third, a decrease in total MDSC, M-MDSC, PMN-MDSC, Treg cells, CD4 T cells, CD8 T cells, CD4/CD8 ratio and lymphocytes was reported for triple negative from baseline to cycle 6. None of the reported differences were statistically significance. 135 Fourth, a decrease in total MDSC, M-MDSC, CD4 T cells, CD8 T cells while an increase in PMN-MDSC, lymphocytes, Treg, CD4/CD8 ratio was reported in the luminal subgroup from baseline to cycle 6. Statistical significance was only reached for the decrease of CD4 T cells. The fact that fluctuations in these immune cells counts were also found in the luminal subgroup may be suggestive of activity of the tested strategy and so, open the door to test immune checkpoint inhibitor in this specific breast cancer subtype. With specific regard to MDSC, these are a heterogeneous group of myeloid progenitors and immature immune cells that have immunosuppressive properties. As previously discussed, their main suppressive role is at the level of T cells. When generally assessing MDSC along the course of the treatment, it should be noted that the major subtype of these cells were M-MDSC compared to PMN-MDSC.M-MDSC median count mostly triplicated the PMNMDSC median count at baselines. The difference was maintained at cycle 3 although it was lower; M-MDSC median count were almost 2 fold lower compared to PMN-MDSC median count at cycle 3. Impressively, median counts at cycle 6 were of comparable magnitude. Information about normal distributions of these two subtypes of MDSC that could help us to contextualise these results is very limited and could seen contradictory across tumour types. For instance, a recent published study of NSCLC patients treated with nivolumab focused its discussion in PMN-MDSC as stated that this was the major group of these cells in NSCLC92. However, previous study also in NSCLC focused its discussion in M-MDSC suggesting these cells may be a predictive and prognosis biomarker for anti-PD-1 therapy in NSCLC patients82. In the specific case of breast cancer, Bergenfelz et al. analysed the peripheral blood of patients with breast cancer and found a significant enrichment of M-MDSC compared to healthy volunteers and sepsis patients158.Similar conclusions were found later by Larsson et al159. This is in line with the results we found when baseline MDSC counts of PANGEA 136 patients were compared to a cohort of 20 healthy volunteers. Statistically significant results were found as levels in the disease population were greater compared to the healthy volunteers; total MDSC median counts in the patient population doubled the ones observed in healthy volunteers being this driven by M-MDSC which was the most enriched subset of MDSC compared to PMN-MDSC. PANGEA PMN-MDSC also presented higher baseline counts compared to healthy volunteers but this difference was not statistically significant. It should be noted that the cohort of healthy volunteers presented similar demographical characteristics as race, sex and with ages comprised between 35 and 60 years old to patients than patients in the PANGEA- breast trial, however, the information was not captured. With regard to the general analysis of total MDSC along the observation period, results showed a decrease of median counts for total MDSCs with values of 44.46cel/µL and 19.33cel/µL at baseline and cycle 6, respectively. This decrease was not statistically significant. As previously explain these results were mainly driven by the M-MDSC subset. When assessing MDSC by clinical outcome, baseline median counts of total MDSCs for patients who did not showed a clinical benefit were almost 1.5 fold superior compared to patients reaching clinical benefit. As expected, the same was found when M-MDSC and PMN-MDSC were assessed separately. Although with a different magnitude, median counts in the subgroup with progressive disease continued to be greater compared to the subgroup of patients reaching the clinical benefit along the course of treatment; the greatest difference was seen at cycle 3. This is in line with literature already discussed for melanoma, prostate and NSCLC patients and supportive of the potential predictive value of baseline MDSC counts to immune checkpoint blockade therapy65,76,78,79,82. Furthermore, a depletion of total MDSC was found in the subgroup of patients that reached clinical benefit. This decrease was seen at cycle 3 and 137 median values were maintained thereafter. This is also in line with reported data already discussed for patients treated in with checkpoint inhibitors in melanoma, prostate and NSCLC. Although no statistically significant, again, these results support the potential predictive value of MDSC to immune checkpoint inhibitors. In this sense, one could have expected to ideally find a different trend in the subgroup of patients who progressed. However, a decreased in median values from baseline compared to cycle 6 was also showed for this subgroup. As previously discussed, these results can be argued noticing that i) sample size between baseline (n= 15) and cycle 6 (n=5) were not comparable due to samples lost, and ii) median MDSC counts barely changed from baseline to cycle 3 whereas they sharply decreased in the clinical benefit subgroup. It should be also noted that in the discussed papers, authors normally assessed MDSC in relation to ORR, i.e. patients who only presented complete response and partial response, whereas in our case we are also including patients with stable disease. This could be limiting as it cannot be discern whether this subgroup of patients could present similar features to the ones that reached a response or the ones who progressed. To note, trends seen in total MDSC results were mostly driven by the M- MDSC subset as PMN-MDSC subset showed different trends along the course of the treatment in both subgroups. Following with results per histology, statistically significant greater median counts were found for the triple negative compared to luminal patients at baseline. Greater proportions were also found in triple negative patients compared to luminal ones in M-MDSC and PMN-MDSC separated analyses. However, statistically significance was only found for M-MDSC. Although statistical significance was not maintained, triple negative median values were greater compared to luminal ones along the observation period, mostly in both subset, M-MDSC and PMN-MDSC. In this sense, results could 138 support prognosis role of MDSC being increased numbers of MDSC associated to more aggressive disease. Overall, both populations mostly showed a similar trend; a decreased in median counts were seen from baseline to cycle 6 being the sharpest decreased reported at cycle 3. Although it is not possible to discern if this activity is attributed to one of component of the combination, and if so which, or to their synergism, the fact that luminal patients showed a similar trend to triple negative population could open the door to new application of immune checkpoint inhibitor in this specific breast cancer subtype. To note trends seen for total MDSC were mostly the reflection of the behavior of M-MDSC cells as PMN-MDSC showed different trends in both luminal and triple negative populations. With regard to the general description of absolute lymphocytes counts, a continuous decrease of the median counts of lymphocytes from baseline until cycle 6 was seen. This difference between baseline and cycle 6 was close to be statistically significant (p-value= 0.053).This decrease in lymphocytes could be due to the effect of pembrolizumab as lymphophenia is reported as a common adverse reaction in the product information (PI) 38 When assessing results by clinical benefit, baseline frequencies were higher for the subgroup of patients reaching clinical benefit compared to those who progressed, albeit not statistically significant. In the subgroup of patients reaching a clinical benefit, a constant decrease from baseline to cycle 6 in median counts is described whereas in the subgroup of patients who progressed an initial increase was seen from baseline to cycle 3 followed by a decrease from cycle 3 to cycle 6. As previously discussed lymphocytes counts are affected by anti PD-1 drugs. In this sense, although results should be interpreted with caution due the lost of samples at cycle 6 in the non-responder subgroup, the fact that a continuous decrease is reported for the subgroup who reached clinical benefit in contrast to an increased from baseline to cycle 3 for the non- 139 responder group, may support the predictive value of lymphocytes counts to immune checkpoint blockade therapeutic strategy. Absolute lymphocyte count ≥ 1000/μL during treatment with nivolumab is reported to correlate with improved overall survival in melanoma patients and to predict response to this drug in NSCLC77,81,94. However, this not applied to our results as all the median absolute lymphocyte count were well above 1000 cel/μL with the exception of median values at cycle 6 for the non-responder subgroup. Analysis by histology showed that showed a statistically significant lower median count of lymphocytes at baseline for triple negative patients compared to the luminal population. Different trends were seen in the median counts from baseline to cycle 6 for both populations; a decrease was seen for luminal patients whereas an increase was reported for triple negative patients. In relation to previous assumptions, the fact that luminal population showed that lymphocytes population is affected by anti PD-1 drugs could open door to new applications of these drugs in this histology. Following with Treg cells, the general analysis showed a statistically significant and constant decrease of the median values of Treg from baseline to cycle 6 was seen. Overall, Treg cells depletion has been reported in breast cancer patients treated with chemotherapy. Specifically, gemcitabine based therapies have been reported to enhance depletion of Treg cells. 115,159 When assessing results per clinical outcome, a decrease along the observation period was seen in Tregs irrespectively of the response achieved. Again the comparability exercised is hampered due to the limited number of samples at cycle 6 (n=5) and so, it should be noted that median counts also increased from baseline to cycle 3. Similarly to other subset of cells, results may suggest the predictive value of Tregs to the tested strategy. In regard to the analysis per histology, Tregs from the TME were reported to be at a higher proportion in triple negative disease compared to other