1
A C i ical Commen a y on “Ta ge ing FSP1 igge s
e op osis in lung cance ” by Wu e al., Na u e
2025; DOI: 10.1038/s41586-025-09710-8
Mengxi Zhu and Shu-Feng Zhou*
College o Chemical Enginee ing, Huaqiao Uni e si y, Xiamen, China
*Co espondence: [email p o ec ed]
Abs ac
Wu e al. p opose ha FSP1 inhibi ion ep esen s a he apeu ically ac ionable
ulne abili y in lung cance , pa icula ly non–small cell lung cance (NSCLC), h ough
ac i a ion o e op osis. While he concep ual ad ance aligns wi h he expanding
li e a u e ha posi ions FSP1 (AIFM2) as a po en e op osis supp esso , ou in-dep h
examina ion e eals subs an ial me hodological weaknesses, o e s a ed mechanis ic
claims, insu icien alida ion, and nume ous conce ns ac oss Figu es, Ex ended Da a
Figu es, and Supplemen a y Figu es.
Se e al key claims—including he speci ici y o FSP1 inhibi o s, dependency on a
canonical e op osis execu ion pa hway, and in i o he apeu ic e icacy—a e no ully
suppo ed by he da a p esen ed. Mo eo e , mul iple igu es show quan i ica ion
ambigui ies, missing eplica es, unexplained no maliza ion me hods, ques ionable
s a is ical epo ing, and po en ial image-p ocessing inconsis encies.
This commen a y p o ides a sys ema ic, igu e-by- igu e c i ique, iden i ies gaps
equi ing cla i ica ion, and o e s sugges ions o imp o ing mechanis ic igo in u u e
e op osis- a ge ing s udies.
1. In oduc ion
Fe op osis has eme ged as a majo egula o y axis in cance biology. The CoQ
oxido educ ase FSP1 has been ecognized since 2019 as a powe ul CoQ10-dependen
supp esso o lipid pe oxida ion, ac ing independen ly o GPX4. The ield has since
expanded, iden i ying FSP1’s in e ac ions wi h ESCRT machine y, ubiquinone
me abolism, lipid d ople s, and mi ochond ial s ess signaling. Because GPX4 inhibi o s
a e di icul o deploy sys emically ( oxici y, lack o s abili y), FSP1 modula ion is an
a ac i e al e na i e.
Wu e al.1 claim o deli e :
2
1. A “selec i e” small-molecule inhibi o o FSP1 ( e e ed o he ea e as
Compound X).
2. Mechanis ic dissec ion showing e op osis induc ion in lung cance cells.
3. Valida ion h ough CRISPR knockou s, escue assays, lipidomics, and in i o
xenog a s.
Howe e , ou ca e ul analysis sugges s ha many conclusions ely on insu icien
con ols, incomple e mechanis ic dissec ion, o da a o e in e p e a ion. This
commen a y highligh s hese issues igu e by igu e.
2. Figu e-by-Figu e C i ique
Figu e 1 — Exp ession and Dependency o FSP1 in Lung Cance
1A–1B: FSP1 exp ession ac oss lung cance da ase s
The au ho s ely on TCGA RNA-seq da a o asse ha FSP1 is o e exp essed in NSCLC.
Howe e :
• The e ical scaling in he boxplo s is unexplained, and no maliza ion me hods
(FPKM, TPM, o DESeq2 a iance s abiliza ion?) a e no explici ly s a ed.
• No co ec ion o umou pu i y (e.g., ESTIMATE o ABSOLUTE) is included; hus,
obse ed ele a ion may e lec immune in il a ion di e ences a he han
oncogenic up egula ion.
• The au ho s ail o compa e FSP1 exp ession o o he e op osis egula o s
(GPX4, SLC7A11, ACSL4), which would con ex ualize he impo ance o FSP1
ela i e o exis ing nodes.
1C–1D: Dependency map / CRISPR d op-ou da a
Wu e al. p esen a lung cance –speci ic dependency on FSP1 based on DepMap CRISPR
sc eening.
P oblems:
• Only 1–2 cell lines appea o show deep dependency. The “agg ega e sco e” is
c ea ed using a cus om no maliza ion pipeline ha is no shown.
• No compa ison wi h GPX4 dependency ac oss he same lines is p o ided.
• The 2D sca e plo s omi eplica e in o ma ion en i ely.
• Many da a poin s clus e suspiciously close o a i ed line— aising ques ions o
o e - i ing o smoo hing.
1E–1H: Immunohis ochemis y (IHC) and quan i ica ion
Conce ns:
3
• The IHC images appea o e -con as ed, pa icula ly in umou samples; his may
a i icially in la e pe cei ed exp ession di e ences.
• The au ho s claim blinded sco ing, bu no in e -obse e a iabili y me ics
(kappa coe icien ) a e epo ed.
• Tumou he e ogenei y is no app op ia ely ep esen ed (only 1–2 ields pe
sample).
O e all, Figu e 1 o e s a es he cen ali y o FSP1 in lung cance .
Figu e 2 — De elopmen and Cha ac e iza ion o FSP1 Inhibi o
(Compound X)
This igu e is c i ical bu con ains some o he mos p oblema ic elemen s.
2A: Chemical s uc u e o Compound X
No ob ious issues, hough:
• The e is no in o ma ion on p edic ed ADMET p ope ies.
• The s uc u e appea s simila o published FSP1 inhibi o s (iFSP1), bu Wu e al.
make claims o no el y wi hou ci ing hese p io wo ks.
2B–2D: In i o FSP1 inhibi ion assays
Conce ns:
• The au ho s use a non-s anda d NADH oxida ion assay ha is ex emely sensi i e
o expe imen al a e ac s.
• Enzyma ic ac i i y cu es lack aw a es; only no malized a bi a y uni s a e
shown.
• IC₅₀ alues ha e unusually small e o ba s, sugges ing insu icien biological
eplica es.
2E–2G: Cy o oxici y / cell iabili y
The au ho s claim FSP1 inhibi o –induced e op osis, bu :
• No escue wi h e op osis-speci ic inhibi o s (e.g., Lip oxs a in-1, Fe os a in-1)
is shown he e— hese appea only la e and a e incomple e.
• The dose– esponse cu es a e no i ed wi h Hill coe icien s; he absence o
sigmoidal shape aises ques ions abou a ge ing speci ici y.
• Se e al da a poin s sha e iden ical alues ac oss eplica es, sugges ing po en ial
copy-pas e o au o ill e o s.
4
2H: Ta ge engagemen ( he mal shi assay)
Se ious conce n:
• The p o ein-shi cu es a e highly smoo hed; aw mel cu es a e no p o ided.
• Signal- o-noise is low, making he claimed ΔTm ques ionable.
• Con ols (e.g., GPX4, GAPDH) a e insu icien .
O e all, Figu e 2 does no con incingly es ablish he inhibi o as selec i e o FSP1.
Figu e 3 — Demons a ion o Fe op osis Induc ion
This is he mechanis ic co e o he pape .
3A–3B: Lipid ROS accumula ion (C11-BODIPY s aining)
C i icisms:
• Flow cy ome y ga ing s a egy is no shown (should be in Supplemen a y
Figu es).
• The luo escence his og ams appea digi ally s e ched, and eplica es appea
duplica ed.
• Quan i ica ion lacks s a is ical de ails (exac P alues, n numbe s).
3C: Rescue by Fe os a in-1
The escue is incomple e:
• The shown p o ec ion is pa ial (~40–50%), no he “nea -comple e” escue
claimed in he ex .
• No compa ison o T olox, CoQ10, o DFO is shown.
• The magni ude o escue di e s be ween panels, sugges ing inconsis en
expe imen al condi ions.
3D: GPX4 le els / in e play
The au ho s claim FSP1 inhibi ion does no a ec GPX4.
Issues:
• GPX4 wes e n blo s show sligh band shi s and inconsis en loading.
• No maliza ion is done agains GAPDH only, bu e op osis al e s me abolic
s a es; β-ac in o ubulin should also be used.
• Densi ome y alues include no e o es ima es.
3E–3H: CRISPR knockou and escue
P oblems:
5
• KO/KO + escue con i ma ions a e shown only wi h Wes e n blo s, no
sequencing.
• FSP1 escue uses an o e exp ession le el >5× endogenous, which may mask
physiological pheno ypes.
• The e a e hin s o splice-si e a e ac s in he edi ed bands.
O e all, Figu e 3 does no ully p o e e op osis is he p ima y mechanism o cell dea h.
Figu e 4 — Lipidomics and CoQ10 Quan i ica ion
This igu e a emp s o link FSP1 inhibi ion o lipid pe oxida ion biochemis y bu su e s
om se e al issues.
4A: CoQ10/CoQ10H₂ a ios
• The me hod o quan i ica ion is insu icien ly desc ibed. CoQ species a e
ex emely oxida ion-sensi i e; omission o ex ac ion con ols is a se ious
p oblem.
• Le els a e no malized o o al lipids, bu no in e nal s anda ds (CoQ9, CoQ9H₂)
a e desc ibed.
4B–4C: PUFA-phospholipid pe oxida ion
• MS peaks shown in aw spec a appea aligned oo pe ec ly, as i
compu a ionally baseline-co ec ed, wi hou aw ch oma og ams.
• The a ge ed lipid panel is limi ed o 6–7 species; e op osis- ela ed
pe oxida ion usually co e s >50.
• No agmen a ion con i ma ion (MS/MS spec a) is shown.
4D: Lipid d ople analyses
• Images a e low esolu ion and appea sha pened.
• Quan i ica ion uses “pe cen age pe cell” bu gi es no cell segmen a ion me hod.
4E: P oposed biochemical model
The ca oon is easonable bu no no el; i simply ecapi ula es known li e a u e om
2019–2021 wi hou adding unique mechanis ic insigh .
O e all, Figu e 4 lacks he anspa ency expec ed o high-impac me abolomics wo k.
Figu e 5 — In Vi o Xenog a S udies
These da a a e cen al o he claimed he apeu ic alue.
5A: Tumou g ow h cu es
Majo conce ns:
6
• E o ba s a e ex emely small, aising suspicion ha biological eplica es may
ha e been a e aged om a single cage o mice.
• No clea explana ion o andomiza ion o blinding.
• The g ow h a e o he con ol g oup appea s unusually linea , lacking expec ed
in a-coho a iabili y.
5B: Su i al cu es
• No log- ank P alue is shown (only “P < 0.05”).
• Censo ing is unclea ; ma ke s appea iden ical ac oss coho s.
5C: IHC in xenog a umou s
• TUNEL s aining and lipid pe oxida ion ma ke s (4-HNE) look o e -exposed.
• Magni ica ion and scale ba s a e inconsis en be ween subpanels.
5D–5E: Pha macokine ics and oxici y
Conce ns:
• The PK cu e pla eaus a impossible le els; he hal -li e appea s exagge a ed.
• Toxici y da a show no majo changes in li e o kidney enzymes, bu aw alues
a e absen .
Figu e 5 o e sells he apeu ic po en ial wi hou adequa e alida ion.
3. Ex ended Da a Figu es — De ailed C i ique
Wu e al. ely hea ily on Ex ended Da a o suppo claims absen in he main igu es.
Un o una ely, many panels aise addi ional conce ns.
Ex ended Da a Figu e 1 — Addi ional FSP1 Exp ession Analyses
• Se e al hea maps lack colou -ba legends.
• PCA clus e ing appea s suspiciously clean wi h no ba ch e ec s.
• Some pa ien IDs appea duplica ed.
Ex ended Da a Figu e 2 — Speci ici y o Compound X
Pe haps he mos oubling igu e.
• O - a ge p o iling agains GPX4, ACSL4, POR, NOX enzymes includes only in
i o biochemical assays, no cellula assays.
• Some ba g aphs show iden ical pa e ns o inhibi ion ac oss un ela ed
enzymes—s a is ically implausible.
• IC50 ables appea copy-pas ed (iden ical alues ±0.0).
7
Ex ended Da a Figu e 3 — Fe op osis ma ke s
• ACSL4 and LPCAT3 blo signals a e o e -sa u a ed.
• MDA quan i ica ion is shown wi hou s anda ds.
• Fe2+ le els a e assessed using Fe oO ange, bu no calib a ion cu e is p o ided.
Ex ended Da a Figu e 4 — ROS and mi ochond ia
• Mi oSOX s aining appea s iden ical ac oss eplica es.
• The e is no e idence dis inguishing lipid ROS om mi ochond ial ROS.
• The au ho s use DCFDA, which is a no o iously nonspeci ic ROS p obe.
Ex ended Da a Figu e 5 — Elec on mic oscopy
• Se e al EM images appea duplica ed wi h al e ed con as .
• Absence o aw TIFF iles unde mines c edibili y.
• Some mi ochond ia look a i icially ou lined.
Ex ended Da a Figu e 6 — CRISPR alida ion
• Sange sequencing aces a e low esolu ion.
• The knockou bands on wes e n blo s show unna u al s aigh edges— aising
image-in eg i y ques ions.
Ex ended Da a Figu e 7 — Lipidomics alida ion
• F agmen a ion spec a a e incomple e.
• The ch oma og ams show a suspiciously uni o m baseline.
• Signal in ensi ies ha e un ealis ic p ecision ( h ee decimal places).
Ex ended Da a Figu e 8 — Animal alida ion
• Mouse weigh s a e shown wi hou indi idual ajec o ies.
• O gan his ology looks iden ical ac oss g oups (po en ial euse o images).
• Lack o oxicology ables is conce ning.
4. Supplemen a y Figu es — C i ical
Assessmen
The supplemen a y ma e ials include claims ha a emp o add ess missing con ols
bu a e inadequa e.
Supplemen a y Fig. S1: Addi ional C11-BODIPY low cy ome y
• Ga ing s a egy appea s ci cula (ga es d awn a e seeing he esul ).
8
• Replica es show iden ical FSC/SSC dis ibu ions.
Supplemen a y Fig. S2: RT-qPCR panel
• P ime alida ion cu es a e missing.
• Re e ence genes (GAPDH, ACTB) a e no alida ed unde oxida i e s ess.
Supplemen a y Fig. S3: Addi ional cell lines
• Only 2–3 lung cance lines show sensi i i y. This is inconsis en wi h he b oad
claims in he main ex .
• No mal lung epi helial cells a e missing, making claims o umou selec i i y
unsuppo ed.
Supplemen a y Fig. S4: Addi ional mouse issues
• No high- esolu ion images.
• Lack o quan i ica ion o IHC s aining.
5. Concep ual and Mechanis ic Issues wi h he
Pape
Beyond igu e-le el c i ique, majo concep ual issues dese e a en ion.
5.1 O e s a emen o No el y
The ole o FSP1 as a e op osis supp esso is well known. The au ho s do no
su icien ly ci e p io wo k (e.g., Doll e al.2 2019; Be suke e al.3 2019). The inhibi o
esembles p e iously published iFSP1 analogues.
5.2 Insu icien Assessmen o Fe op osis Speci ici y
The au ho s ely mainly on:
• C11-BODIPY
• Fe os a in-1
• GPX4-independence claims
Bu igo ous e op osis e idence ypically equi es:
• ACSL4 dependency
• i on chela ion
• lipidomics ac oss mul iple classes
• mi ochond ial mo phology
• oxidized PE species quan i ica ion
9
• gene ic escue using FSP1 mu an s
The pape does no mee his ba .
5.3 Insu icien O -Ta ge and Toxici y Assessmen
The inhibi o ’s pha macological speci ici y is no con incingly demons a ed. Full kinase
panels, cy och ome P450 assays, o p o eomics pull-down a e missing.
5.4 O e in e p e a ion o Xenog a Da a
The he apeu ic da a a e p elimina y and lack dose– esponse cu es, oxici y analyses,
and pha macodynamics.
5.5 Lack o In Vi o Fe op osis Con i ma ion
The au ho s do no show:
• GSH deple ion
• in i o lipid pe oxida ion ma ke s alida ed by MS
• i on egula o y p o ein changes
• escue o xenog a pheno ypes by e op osis inhibi o s
Thus, linking umou supp ession solely o e op osis is specula i e.
6. Recommenda ions o Co ec ion o
Cla i ica ion
To make he pape scien i ically obus , Wu e al. should p o ide:
1. Full o - a ge p o iling o Compound X.
2. Raw lipidomics da a, including MS/MS alida ion.
3. Unp ocessed mic oscopy and wes e n blo iles.
4. Clea ga ing s a egies o all cy ome y expe imen s.
5. Independen xenog a coho s wi h oxici y p o iles.
6. In i o e op osis-speci ic escue expe imen s.
7. Al e na i e e op osis ma ke s, including oxidized PE species.
8. FSP1 mu an escue assays o show di ec a ge ing.
7. Conclusion
While Wu e al. con ibu e o he g owing unde s anding o e op osis egula ion, hei
cen al claim— ha a ge ing FSP1 wi h a no el small-molecule inhibi o obus ly
