scieee Science in your language
[en] (orig)

Lipid fingerprint-based histology accurately classifies nevus, primary melanoma, and metastatic melanoma samples

Author: Huergo Baños, Cristina,Velasco, Verónica,Garate Yeregui, Jone,Fernández Regueira, Roberto Antonio,Martín Allende, Javier,Zabalza Estévez, Ignacio,Artola Igarza, Juan Luis,Martí Laborda, Rosa María,Asumendi Mallea, Aintzane,Astigarraga, Egoitz,Barreda Góm
Publisher: Wiley
Year: 2024
DOI: 10.1002/ijc.34800
Source: https://addi.ehu.eus/bitstream/10810/67337/1/Intl%20Journal%20of%20Cancer%20-%202023%20-%20Huergo%e2%80%90Ba%c3%b1os%20-%20Lipid%20fingerprint%e2%80%90based%20histology%20accurately%20classifies%20nevus%20primary.pdf
RESEARCH ARTICLE
Inno a i e Tools and Me hods
Lipid inge p in -based his ology accu a ely classi ies ne us,
p ima y melanoma, and me as a ic melanoma samples
C is ina Hue go-Baños
1
| Ve 
onica Velasco
2,3
| Jone Ga a e
1
|
Robe o Fe nández
1
| Ja ie Ma ín-Allende
4
| Ignacio Zabalza
3,5
|
Juan L. A ola
3,6
| Rosa M. Ma í
7,8
| Ain zane Asumendi
3,9
|
Egoi z As iga aga
10
| Gab iel Ba eda-G
omez
10
| Ola z F esnedo
11
|
Begoña Ochoa
11
| Ma ia D. Boyano
3,9
| José A. Fe nández
1
1
Depa men o Physical Chemis y, Facul y o Science and Technology, Uni e si y o he Basque Coun y (UPV/EHU), Leioa, Spain
2
Depa men o Pa hology, C uces Uni e si y Hospi al, Ba akaldo, Spain
3
Bioc uces-Bizkaia Heal h Resea ch Ins i u e, C uces Uni e si y Hospi al, Ba akaldo, Spain
4
Languages and Compu e Sys ems, School o Enginee ing Uni e si y o he Basque Coun y (UPV/EHU), Bilbao, Spain
5
Depa men o Pa hology, Galdakao-Usansolo Uni e si y Hospi al, Galdakao, Spain
6
Depa men o De ma ology, Galdakao-Usansolo Uni e si y Hospi al, Galdakao, Spain
7
Depa men o De ma ology, A nau de Vilano a Uni e si y Hospi al, Ins i u e o Biomedic Resea ch (IRBLleida), Uni e si y o Lleida, Lleida, Spain
8
Cen e o Biomedical Resea ch on Cance (CIBERONC), Ins i u o de Salud Ca los III (ISCIII), Mad id, Spain
9
Depa men o Cell Biology and His ology, Facul y o Medicine and Nu sing, Uni e si y o he Basque Coun y (UPV/EHU), Leioa, Spain
10
Depa men R+D, IMG Pha ma Bio ech S.L., De io, Spain
11
Depa men o Physiology, Facul y o Medicine and Nu sing, Uni e si y o he Basque Coun y (UPV/EHU), Leioa, Spain
Co espondence
Ma ia D. Boyano, Depa men o Cell Biology
and His ology, Facul y o Medicine and
Nu sing, Uni e si y o he Basque Coun y
(UPV/EHU), Ba io Sa iena, s/n, 48940 Leioa,
Spain.
Email: [email p o ec ed]
José A. Fe nández, Depa men o Physical
Chemis y, Facul y o Science and Technology,
Uni e si y o he Basque Coun y (UPV/EHU),
Ba io Sa iena, s/n, 48940 Leioa, Spain.
Email: josea. e [email protected]
Funding in o ma ion
Agencia Es a al de In es igaci
on, G an /Awa d
Numbe : RTC-2015-3693-1; Basque
Go e nmen , G an /Awa d Numbe s:
IT1491-22, KK2017-041, KK2020-00069;
Euskal He iko Unibe si a ea, G an /Awa d
Numbe : GIU17/066
Abs ac
P obably, he mos impo an ac o o he su i al o a melanoma pa ien is ea ly
de ec ion and p ecise diagnosis. Al hough in mos cases hese asks a e eadily
ca ied ou by pa hologis s and de ma ologis s, he e a e s ill di icul cases in
which no consensus among expe s is achie ed. To deal wi h such cases, new
me hodologies a e equi ed. Following his mo i a ion, we explo e he e he use o
lipid imaging mass spec ome y as a complemen a y ool o he aid in he diagno-
sis. Thus, 53 samples (15 ne us, 24 p ima y melanomas, and 14 me as asis) we e
explo ed wi h he aid o a mass spec ome e , using nega i e pola i y. The ich
lipid inge p in ob ained om he samples allowed us o se up an a i icial
in elligence-based classi ica ion model ha achie ed 100% o speci ici y and p e-
cision bo h in aining and alida ion da a se s. A deepe analysis o he image da a
shows ha he echnique epo s impo an in o ma ion on he umo mic oen i-
onmen ha may gi e in aluable insigh s in he p ognosis o he lesion, wi h he
co ec in e p e a ion.
Recei ed: 25 May 2023 Re ised: 24 Oc obe 2023 Accep ed: 2 No embe 2023
DOI: 10.1002/ijc.34800
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion-NonComme cial License, which pe mi s use, dis ibu ion and ep oduc ion in any
medium, p o ided he o iginal wo k is p ope ly ci ed and is no used o comme cial pu poses.
©2023 The Au ho s. In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o UICC.
712 In . J. Cance . 2024;154:712–722.
wileyonlinelib a y.com/jou nal/ijc
KEYWORDS
bioma ke s, lipid imaging mass spec ome y, melanoma, molecula his ology
Wha 's new?
He e, he au ho s e alua e he po en ial o ma ix-assis ed lase deso p ion/ioniza ion lipid
imaging mass spec ome y (LIMS) oge he wi h classi ica ion models buil using a i icial in elli-
gence o classi y samples o ne us, melanoma, and me as a ic melanoma. They ind ha , looking
a he al e a ions in he lipid p o ile o he issues and ha ing buil a lib a y o lipid signa u es
using LIMS, i is possible o au oma ically de ec he p esence o umo cells and e en de e mine
i he sample is a p ima y umo o a me as asis. The indings pa e he way o he de elopmen
o as , accu a e, and au oma ized p o ocols o he sc eening o melanoma samples.
1|INTRODUCTION
Al hough cu aneous melanomas ep esen only 5% o all skin cance s,
i is he mos le hal due o i s high a e o me as asis and he lack o
e ec i e ea men s in ad anced s ages. In clinical p ac ice, p ognosis o
melanoma is based on he B eslow index, he p esence o ulce a ion, and
sen inel node e alua ion.
1
A ound 10% o melanoma ecu ences wi hin
5 yea s o ollow-up ha e been desc ibed
2
o ea ly s age lesions (I and II
s ages acco ding o AJCC 8 h edi ion).
3
Fu he mo e, a ecen s udy
showed ha in a coho o 784 melanoma pa ien s, 53.8% o all me a-
s a ic pa ien s had an I–II ini ial s age.
4
These ac s suppo he iew ha
many ea ly melanomas wi h a biological abili y o me as asize a e no
iden i ied by classical pa hological ma ke s.
De ec ion o melanoma, e en a ea ly s ages, is ou inely pe o med
in he diagnosis se ices by pa hologis s and de ma ologis s. Howe e ,
he s eady aise in he numbe o possible cases o melanoma is inc eas-
ing he p essu e in such se ices. Fu he mo e, despi e he accu acy o a
well- ained pa hologis , he e a e s ill complex cases in which diagnosis
is no simple.
5
The e o e, de eloping new complemen a y me hodologies
ha may help wi h he diagnosis and p ognosis is a e y ac i e esea ch
ield. Fo example, in he las yea s se e al op ical and p o eomic
app oaches ha e appea ed,
6,7
aiming a imp o ing diagnosis.
A di e en app oach o he p oblem o achie ing an accu a e diagno-
sisiso e edbylipidimagingmassspec ome y(LIMS).This echnique
enables di ec explo a ion o issue sec ions using a mass spec ome e .
8
The s a ing poin is a esh ozen sample, which is sec ioned using a c yo-
mic o ome (Figu e 1). The sec ions, usually 10–20 μm- hick, a e deposi ed
in a mic oscope glass slide and co e ed wi h a ma ix ha enables he
ex ac ion o he molecules wi h he aido alase .Then,ag ido coo di-
na esisde ined,whichwillbecome hepixelsin he inalimages.Themass
spec ome e scans he sample, acqui ing a mass spec um a each coo di-
na e. The esul is he dis ibu ion map o each o he analy es ex ac ed
om he issue, wi hou he need o a p e ious labeling.
9
When he a -
ge ed molecules a e lipids, he echnique becomes a kind o digi al molecu-
la his ology ha enables isualiza ion o each cell ype in a gi en issue,
ul ima ely in base o hei me abolic signa u e.
10,11
Ce ainly, lipids, which a e no gene ically encoded, a e, di ec ly
o indi ec ly, p ima ily o seconda ily, in ol ed in all main me abolic
p ocesses o a cell: om s uc u e o signaling, apop osis, and so
o h.
12–15
The e o e, each cell in a gi en me abolic s age p esen s a
well-de ined, igh ly egula ed, lipid inge p in .
16
In o he wo ds, sub-
le a ia ions in he me abolic s a us o a cell in ol e modi ica ions in
lipid exp ession. Fo example, we demons a ed in he pas ha he
ma u a ion o he colonocy es as hey di e en ia e along he c yp s
in ol es a quan i iable modi ica ion o he lipidome.
17
Some o he
species exhibi a g adien along he c yp ha is main ained among
indi iduals and ha is di e en om he lipid inge p in o he cells in
he lamina p op ia. Fu he mo e, he inge p in o he colonocy es is
s ongly al e ed in a sample o neoplas ic issue and he e o e, i may
se e o ea ly de ec ion o he disease.
18,19
We ha e also demons a ed ecen ly ha i is possible o es ablish
a leas se en di e en segmen s in he neph on acco ding o he lipid
inge p in and ha he, in some case sub le, di e ences in lipid p o ile
be ween hese segmen s a e p ese ed among indi iduals.
20
Simila
obse a ions we e also made ega ding ne us.
21
In ha case, we dem-
ons a ed ha epide mis, de mis, and melanocy es p esen also well-
de ined inge p in s. Deep analysis o he esul s om ne i poin ed o a
di e ence in he lipid exp ession be ween supe icial melanocy es and
hose ha ha e mig a ed o deepe a eas o he de mis. Ve y likely,
such a ia ion is due o he ma u a ion p ocess o he melanocy es.
22,23
We exploi he e he malignancy s a e-associa ed changes in lipid
me abolism o melanocy ic cells o diagnos ic bene i and explo e he
use o LIMS o classi y pa ien samples in o ne us, p ima y melanoma,
and me as asis om melanoma. P e ious wo ks al eady explo ed he
me aboli e exp ession in sec ions o p ima y melanoma using mass
spec ome y,
24
demons a ing ha he echnique enables ex ac ing he
me abolic signa u e om di e en cellula popula ions: melanoma cells,
connec i e issue, mac ophages, and lymphocy es, al hough he spa ial
esolu ion a which images we e eco ded, 50 μm/pixel, limi ed some-
how p ecise ex ac ion o such signa u es. In a ecen wo k, Lazo a e al
used mass spec ome y o p obe di e en poin s ac oss sec ions and
TMA co es o ne us and p ima y melanoma, concluding ha i is possible
o build classi ie s using he p o ein inge p in eco e ed di ec ly om
he issue.
25
Using he same app oach, Lazo a e al. demons a ed ha
he echnique is also able o iden i y p oli e a i e nodules wi hin a benign
congeni al ne us
26
and o dis inguish be ween Spi z ne i om Spi zoid
malignan melanoma.
27
Using DESI in a educed coho o samples, Qi
e al. iden i ied choles e ol as a possible bioma ke in human melanocy ic
ne i,
28
al hough he conclusions eached in ha s udy need u he ali-
da ion in a la ge collec ion o samples and hei echnique did no o e
HUERGO-BAÑOS ET AL.713
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
FIGURE 1 Flow diag am o he p o ocol used in his wo k in LIMS expe imen s. Samples a e esh- ozen, ying o a oid he use o OCT o
any o he subs ance ha may al e he lipid dis ibu ion. Then, 16 μm- hick sec ions a e ob ained wi h he aid o a c yomic o ome and a e
deposi ed on a plain glass slide. Nex , hey a e co e ed wi h 1,5-diaminonaph halene and in oduced in he MALDI sou ce o he mass
spec ome e , whe e hey a e scanned ollowing a g id o coo dina es sepa a ed 25 μm. The spec ome e acqui es one mass spec um a each
coo dina e. The o iginal dis ibu ion o he de ec ed species is econs uc ed by in eg a ing each m/z and ep esen ing i s in ensi y agains he
coo dina es. Finally, a segmen a ion analysis is pe o med o de e mine he lipid inge p in s (=cell popula ions) in he sec ion.
714 HUERGO-BAÑOS ET AL.
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
enough spa ial esolu ion o associa e choles e ol changes o cell popula-
ions. O he s udies using echniques wi hou spa ial localiza ion also
epo ed he possibili y o using lipid bioma ke s o de ec ion o mela-
noma.
29
As i can be seen, none o hose p e ious s udies included sam-
ples om ne i, p ima y melanoma, and me as a ic melanoma. On he
o he hand, ou own s udies in cell cul u es poin o a di e en ial lipid
p o ile o benign and umo melanocy ic cells, and be ween he la e ,
depending on hei p oli e a ion po en ial.
30
Demons a ing ha LIMS is
no only able o classi y samples in o umo and non- umo bu also o
iden i y i a sample is a me as asis o a p ima y umo is o pa amoun
impo ance, because i would se he ounda ions o he de elopmen
o new and powe ul diagnos ic echniques.
2|MATERIALS AND METHODS
Samples and da a om pa ien s we e collec ed om 2017 o 2020 and
p o ided by he Pa hology Depa men s o C uces (Ba akaldo, Spain) and
Galdakao-Usansolo (Galdakao, Spain) Uni e si y Hospi als, and by he
Depa men o De ma ology o he A nau de Vilano a Uni e si y
Hospi al (Lleida, Spain). Disease s ages we e classi ied acco ding o
AJCC, 8 h edi ion.
3
Clinical and diagnos ic da a o each pa ien
we e collec ed e ospec i ely om cen alized elec onic and/o
pape medical eco ds. A o al o 15 ne i, 24 p ima y melanomas
and 14 me as asis we e collec ed, along wi h he clinical in o ma-
ion, including gende , age a he p ima y umo , localiza ion o he
p ima y melanoma, localiza ion o me as asis, s age, and his ologi-
cal sub ypes (SSM—supe icial sp eading melanoma, NM—nodula
melanoma, ALM—ac olen iginous melanoma, LMM—len igo maligna
melanoma, Table 1).
The age o he pa ien s anged om 19 o 96 yea s, wi h a simila
gende dis ibu ion. Rega ding he his opa hological classi ica ion,
mos o he ne us we e in ade mal (n=11), al hough a compound
ne us and a junc ional ne us we e also included. The p ima y mela-
noma samples co esponded o SSM (n=19), NM (n=2), LMM
(n=1), and ALM (n=2). Finally, all me as asis samples we e diag-
nosed as loco egional in ansi me as asis.
2.1 |Immunohis ochemis y
Biopsies we e ozen a 80C, a oiding he use o OCT o any o he
compound ha could pe u b he o iginal lipid dis ibu ion. S epwise
sec ioning o he ozen issues was conduc ed by a con en ional
c yomic o ome. Fi s , 4 μm- hick sec ions we e collec ed o
hema oxylin-eosin (H&E) s aining (Geminis AS Au oma ed S aine ,
Einho en, Ne he land) and o Immunohis ochemis y (IHC) o Melan
A and HMB45 melanocy e bioma ke s. Melan A, also known as
MART-1 (melanoma an igen ecognized by T cells) ma ks bo h mela-
nocy es om ne i and melanoma, while HBM45 speci ically ma ks
melanoma cells. IHC was pe o med using he En isionTMGj2 Sys-
em/AP ki . Op ical images we e eco ded using a NanoZoome S210
Digi al slide scanne (Hamama su C13239-01). Melan A (An i-MelanA
an ibody [EP1422Y]) and MB45 (An i-Melanoma gp100 an ibody
[EPR4864]), bo h om Abcam (Camb idge, CB2 0AX, UK) we e used
o iden i y melanoma cells because hey ecognize he g oup o p o-
eins gp-100/Pmel-17 speci ic o melanosomes. Second, sec ions o
16 μm hickness we e ob ained o LIMS. The a eas o be scanned
we e selec ed using he H&E and IHC op ical images. One mus ake
in o accoun ha he whole a ea o he sec ions could no be
explo ed, due he speed o he mass spec ome e (see below), and
lipid p opensi y o oxida ion. Finally, he sec ions explo ed by LIMS
we e s ained wi h H&E, so he pa hologis s could anno a e he his o-
logical a eas and s uc u es in o de o co ela e hem wi h he seg-
men s ob ained om he analysis o he LIMS images. In a small
numbe o cases, he pos -MALDI H&E could no be used because
echnical p oblems (mainly, he issue olded du ing s aining,
because he glass slides did no ha e any ixa ion o a oid al e ing he
o iginal lipid dis ibu ion). In hose cases, he H&E o a consecu i e
sec ion had o be used, which sligh ly complica ed he anno a ion o
he segmen a ion images.
TABLE 1 Clinical in o ma ion o he pa ien s included in
his wo k.
Ne us Melanoma Me as asis
Numbe 15 24 14
Age ( ange) 19–84 32–96 25–91
Gende
Female 8 11 5
Male 7 13 9
Localiza ion
Head and neck 2 2 2
T unk 7 7 4
Uppe ex emi y 1 6 2
Lowe ex emi y 0 5 3
Hand and oo 4 4 1
ND 2
His ology
In ade mal 11
Compound 3
Junc ional 1
LMM 1 1
SSM 19 1
NM 2 3
ALM 2 2
ND 7
S age (AJCC)
In si u 1
I4
II 16
IV 3
Abb e ia ions: ALM, ac olen iginous melanoma; LMM, len igo maligna
melanoma; ND, no de e mined; NM, nodula melanoma; SSM, supe icial
sp eading melanoma.
HUERGO-BAÑOS ET AL.715
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
2.2 |LIMS expe imen s
His ological sec ions om 53 di e en samples we e p epa ed and
analyzed by LIMS as desc ibed in Ga a e e al.
21
A schema ics o he
p o ocol may be ound in Figu e 1. B ie ly, 1,5-diaminonaph halene
(DAN) was used as ma ix and deposi ed wi h he aid o ou in-house
designed sublima o .
31
The sec ions we e scanned in nega i e pola -
i y, using he o bi ap analyze o a MALDI-LTQ-O bi ap XL (The mo
Fishe , San Jose, CA, USA), equipped wi h a modi ied MALDI
sou ce.
32
P e ious wo ks ha e demons a ed ha such pola i y
enables he de ec ion o a la ge numbe o lipid classes, wi h less
in e e ence o adduc s o ma ion.
33
Da a we e acqui ed wi h a mass esolu ion o 60,000 a m/
z=400. Two mic oscans o 10 lase sho s we e a e aged o each
pixel using a 25 μm as e size. Wi h hese se ings, he spec ome e
eco ded one pixel e e y 2 s, and he e o e, i was necessa y o limi
he a ea o he sample o be scanned, o a oid long exposu es o he
issue o oom empe a u e condi ions. Spec a we e p ocessed using
in-house de eloped so wa e, buil in Ma lab (Ma hWo ks, Na ick,
USA). Lipid assignmen was ca ied ou using he m/z alue, he on-
issue MS/MS and MS
3
(whene e he signal in ensi y pe mi ed i )
da a and UHPLC/ESI-MS/MS esul s ob ained om he analysis o
he ex ac s om he same samples. Wi h his p ocedu e, in mos
cases i was no possible o dis inguish be ween e he and inyl-e he
lipids.
Da a om each sec ion we e analyzed using a segmen a ion algo-
i hm (HD-RCA)
34
o isola e and iden i y he lipid signa u es o each
his ological a ea in he sec ion. To es ablish he numbe o segmen s
on each image, a heu is ic app oach was used: he ini ial numbe o
segmen s was se o 10 o ne us and melanoma and 5 o me as asis.
Then, he segmen s sugges ed by he algo i hm we e e i ied by
examining hei co ela ion: hose segmen s wi h co ela ions highe
han 95% we e g ouped oge he , because such a high co ela ion
seems o indica e ha hey de ine simila his ologic a eas. Each seg-
men was colo ed using a colo scale ba and he co ela ion be ween
segmen s. The wo segmen s wi h he lowes co ela ion
be ween hei lipid signa u es we e assigned he colo s a bo h ends
o he scale and he es o he segmen s ecei ed colo s acco ding o
hei co ela ion. Thus, hose segmen s ha p esen colo s ha a e
close in he scale, p esen mo e simila lipid inge p in . Besides, each
sec ion p esen s a unique composi ion and he e o e, simila colo s in
di e en sec ions may co espond o di e en cell popula ions. The
signa u es om he segmen s ob ained om each sec ion we e la e
used in subsequen mul i-expe imen s a is ical analysis.
The spec a we e il e ed be o e s a is ical analysis, o a oid
in oduc ion o excessi e noise. O he 488 mass channels, only
90 exhibi ed a SD lowe han hei mean in ensi y in he h ee condi-
ions. Then, he MS/MS and MS
3
(i a ailable) spec a o each species
was analyzed o p o ide a sound assignmen . To e alua e he s a is i-
cal signi icance o he di e ences in he lipid inge p in s among
ne us melanocy es, p ima y melanoma and me as asis, Le ene es ,
ANOVA uni a ia e s a is ical analysis and Tukey/Games Howell pos
hoc we e compu ed using SPSS S a is ics 17.0 (IBM, A monk, NY,
USA). The Le ene es de e mines he homogenei y o a iance
(H
0
=g oups ha e equi alen a iance) o choose he pos hoc
me hod: Tukey i Le ene p≥.05 and Games-Howell i Le ene p≤.05.
PCA analysis and classi ica ion models we e ca ied ou using O ange
Biolab V.2.7.8 (Ljubljana, Slo enia).
3|RESULTS
3.1 |Iden i ica ion o melanoma his ological
s uc u es by LIMS
Unde s anding he his ology o a ne us using he lipid inge p in p o-
ided by a LIMS expe imen is ela i ely simple. Ne us a e usually
well-o ganized and s uc u ed lesions in which epide mis, de mis, and
melanocy e- ich egions a e eadily iden i ied. This is clea ly seen in
Figu e 2, whe e he compa ison be ween he H&E image o a ne us,
he segmen a ion o a LIMS expe imen and he dis ibu ion o h ee
ep esen a i e lipids is shown.
The ne us sec ion shows sligh papilloma ous g ow h, a he
expense o melanocy es in he de mis, wi h g ow h in mo e supe icial
agg ega es. Melanocy es in he de mal dep h end o sepa a e o o
be mo e di usely a anged. A ce ain ib oscle osis in he papilla y
de mis is also obse ed. The epide mis is hin and an annex o he
middle de mis can be iden i ied. The segmen a ion image is buil
g ouping oge he pixels wi h simila lipid inge p in and assigning
each g oup (segmen ) a colo , using he colo ba in he igu e. The
esul ing image highligh s he issue's a chi ec u e om a molecula
poin o iew. Fo example, clea di e ence be ween he epide mis
(whi e segmen ) and he es o he issue is eadily seen. The melano-
cy es and he su ounding s oma exhibi a high deg ee o he e oge-
nei y, highligh ed by hei di ision in se e al segmen s ha ollow he
melanocy ic agg ega es (ligh blue, yellow, and o ange) and he su -
ounding s oma (black and da k blue). Ul ima ely, he segmen a ion
images e lec he di e en ial spa ial dis ibu ion o he lipid species
de ec ed. As an example, he dis ibu ion o h ee ep esen a i e spe-
cies is shown in he igu e. Sphingomyelin (SM) 34:1 p esen s a highe
abundance in he s oma o ne us, while is sligh ly less abundan in
melanocy es. Con e sely, phospha idyle hanolamine (PE) 36:2 ollows
he opposi e end, being mo e abundan in he epide mis o ne us.
The concen a ion o phospha idylinosi ol (PI) 38:3 is also mo e abun-
dan in he s oma, bu wi h a di e en dis ibu ion, poin ing o
changes in exp ession be ween cell popula ions. Ac ually, he segmen-
a ion image in Figu e 2is a so o summa y o hese lipid dis ibu-
ions: each cell popula ion p esen s a well-de ined lipid p o ile.
Consequen ly, when he pixels in he LIMS image a e g ouped based
on hei lipidome, as in he segmen a ion image, he whole issue
a chi ec u e eme ges. In some sense, LIMS is a kind o molecula
his ology.
Figu e 2(and Figu e S1 o he supplemen al ma e ial) also show
examples o melanoma and me as asis sec ions. The p ima y umo
p esen s an epide mal componen ha a ec s he hai ollicle a a
le el deepe han he in undibula po ion. The mos eye-ca ching
716 HUERGO-BAÑOS ET AL.
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License

ea u e o he image is he hai ollicle wi h cys ic dila ion in addi ion
o ke a in and b own pigmen . The ollicula epi helium is hype plas ic
and occupied by a diso de ly g ow h o indi idual melanocy es, o m-
ing hecae. The umo al pa ha a ec s he de mis shows la ge
agg ega es, wi h a di use g ow h o epi helioid melanocy es wi h
some pigmen , melanophages, and occasionally in a umo al lympho-
cy es. The LIMS image ai h ully ollows he his ology in he H&E
image. On he le side o he segmen a ion image, he dila ed hai ol-
licle appea s in black, while i s luminal ke a in and e en he ke a in o
he s a um co neum appea as a whi e segmen . Tumo melanocy es
a e p edominan ly in he blue segmen , bo h in de mal agg ega es and
in he epide mis. Mo eo e , a pe i umo al egion can be iden i ied o
umo melanocy es and s omal issue wi h di e en lipid p o ile. The
s oma is coinciden wi h he ed and o ange segmen s. In his la e
segmen , lymphocy es and melanophages appea as spo s. When u -
he segmen s a e c ea ed (no shown), he immune cells become
mo e e iden . In addi ion, pe i ollicula collagen egions can also be
de ec ed.
The example o me as asis in Figu e 2, is a collagenous- ich
s oma in il a ed by a umo wi h sca e ed and he e ogeneous p es-
ence o pigmen . Tumo cells a e dis ibu ed almos con inuously,
some imes o ming e ical ails. Collagen and o he elemen s o he
s oma, such as blood essels, can be seen be ween hese ails. In
he segmen a ion image, he melanocy es appea as g een and b own
pa ches and he s oma in blue and whi e. Co espondence be ween
his ological and segmen a ion images is no s aigh o wa d because
he H&E co esponds o a consecu i e sec ion and he di e ences
be ween his ological egions is mo e e iden in he segmen a ion
image. The s iking con as be ween he colo s o he segmen s
inc eases he isibili y o he small colo di e ences in he H&E image.
A look a he dis ibu ion o he h ee lipid species shown as
example in Figu e 2, highligh s once mo e ha he segmen a ion
analysis is a e lec ion o he di e ences in lipid p o ile among he
cells composing he issue. SM 34:1 p esen s a s ong di e ence in
ela i e abundance be ween s oma and umo cells in he melanoma
sec ion, while i p esen s lowe concen a ion in me as a ic cells. PE
36:2 is mo e abundan in umo cells in he melanoma sec ion and is
maximized in me as a ic umo cells. Meanwhile, PI 38:3 also shows a
di e en end: i s ela i e abundance s ongly inc eases in he umo
nodules in melanoma and shows a ela i ely highe abundance in he
s oma o me as asis and in he umo cells ha co ela e wi h
he g een segmen , highligh ing he exis ence o di e en umo cell
popula ions.
The inhe en complexi y o he umo al issues complica es o a
g ea ex en he p ocess o building a classi ie , as a single sample e y
o en p esen s di e en umo cell popula ions wi h (some imes
sligh ly) di e en lipid signa u es. As a i s app oxima ion, we
ex ac ed om each sec ion all hose signa u es ha can be associ-
a ed o umo cells and included hem in he classi ie s, lea ing aside
hose lipid inge p in s co esponding o o he a eas ( ib oblas s, scle-
o ic issue, nec o ic a eas, e c.). Ex ac ion o lipid signa u es was also
guided by he compa ison wi h he Melan A s aining expe imen s.
Figu e 3shows he compa ison be ween example images o H&E
s aining, Melan A and he segmen a ion o a LIMS expe imen ca ied
ou o e samples o in ade mic ne us, p ima y melanoma and
me as asis.
In he segmen a ion image o he ne us, he h ee main his ologi-
cal egions a e clea ly isible: epide mis in whi e, de mis in yellow and
he melanocy e- ich egions in ed and b own. Con e sely, melanomas
a e, in gene al, highly he e ogeneous, p esen ing a a iable numbe o
egions in each sample. Figu e 3, middle ow, shows a his ological
sec ion o a supe icial sp eading p ima y melanoma in s age IIB.
3
The
lesion exhibi s an asymme ic g ow h and a zone o de mal in asion
wi h la ge nodules, sepa a ed by connec i e issue and some
FIGURE 2 Lipid inge p in highligh s he his ology o he issue. H&E op ical images, segmen a ion and dis ibu ion images o SM 34:1, PE
36:2 and PI 38:3 o e sec ions o ne us, melanoma and me as asis om melanoma. The LIMS expe imen was eco ded in nega i e pola i y a
25 μm/pixel o spa ial esolu ion and he segmen s we e colo ed using he colo ba shown in he igu e and he deg ee o co ela ion be ween
segmen s: he p oximi y o he colo s in he scale indica es he co ela ion be ween he lipid inge p in o he segmen s. Rela i e abundance o
he lipids ollows a black-blue- ed-yellow-whi e scale. Scale ba =1 mm.
HUERGO-BAÑOS ET AL.717
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
lymphocy es and melanophages. Tumo nes s a e clea ly de ined and
ep esen ed in he LIMS segmen a ion image as iole and whi e seg-
men s, indica ing ha a leas wo di e en umo cell popula ions
exis . Mo eo e , he lymphocy es su ounding he melanocy e nod-
ules appea as a blue segmen . The epide mis ma ches he ed seg-
men , while o he cells such as ib oblas s and s oma in gene al
appea in g een.
In e es ingly, he e ogenei y is lowe in mos o he analyzed
me as asis. Fo example, he me as asis sec ion in Figu e 3, bo om
ow, clea ly shows h ee egions: umo nodes (b igh and da k ed), a
bo de segmen ha delimi s he umo nodes ha is no e iden in
he op ical image (g een segmen ) and collagen- ich issue (whi e
segmen ).
3.2 |Di e en ial lipid inge p in o melanocy ic
cells om ne us, p ima y melanoma and me as asis
Figu e 4shows he sco es plo o he p incipal componen analysis
(PCA) o he lipid signa u es ex ac ed om he samples using LIMS.
Melanocy ic lipid signa u es come om ne us sec ions, while mo e
han one lipid inge p in was ex ac ed om each p ima y and me a-
s a ic melanoma sample. Wi hin each condi ion, he samples we e
andomly di ided in o aining and alida ion se s and mul iple com-
pa isons we e ca ied ou . When he h ee condi ions we e included
in he analysis, Figu e 4A,B, a good sepa a ion o melanocy ic samples
om hose o umo cells was obse ed in he aining se , while
incomple e sepa a ion was obse ed be ween melanoma and
FIGURE 3 Segmen a ion images desc ibe he his ology o he issue. Compa ison be ween he images o H&E s aining, Melan A bioma ke
dis ibu ion assessed by IHC and he segmen a ion analysis o LIMS ca ied ou o e he same his ological sec ions. Top ow: a ne us sec ion;
middle: p ima y melanoma and bo om ow: me as asis om melanoma. LIMS expe imen eco ded in nega i e pola i y a a pixel size o 25 μm.
The algo i hm au oma ically de ec ed his ological a eas based on he lipid composi ion and assigned colo s acco ding o he colo ba shown in
he igu e, and he co ela ion be ween lipid inge p in s. Scale ba =1 mm.
718 HUERGO-BAÑOS ET AL.
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
me as asis. Howe e , he analysis o he samples in he alida ion se
using only hose lipid species ha showed s a is ically signi ican di -
e ences in he aining se (po en ial bioma ke s), esul ed in a pe -
ec sepa a ion o he h ee ypes o samples (Figu e 4B). Th ee
classi ica ion models we e es ed using he po en ial lipid bio-
ma ke s: suppo ec o machine (SVM), Naï e Bayes and Logis ic
Reg ession, achie ing wi h he la e ape ec classi ica iono he
samples (see Table S1).
Nex , he lipid inge p in s o ne us melanocy es we e compa ed
wi h hose o p ima y melanoma and me as asis (Figu e 4C–F). A
pe ec sepa a ion was achie ed all cases, highligh ing he s ong me -
abolic changes ha accompany umo ans o ma ion. The classi ica-
ion models also exhibi ed ou s anding pe o mances, bo h in he
aining and alida ion se s, wi h logis ic eg ession achie ing pe ec
sepa a ion.
Classi ica ion o p ima y and me as a ic samples was no pe ec
in he aining se (Figu e 4G,H), wi h only wo o he me as asis mis-
classi ied as p ima y umo s. Howe e , sample classi ica ion was pe -
ec in he alida ion g oup, using he po en ial lipid bioma ke s
(Table S1).
FIGURE 4 Sco es plo s o p incipal componen s
analysis (PCA) o he lipid inge p in s. The samples we e
andomly di ided in o aining and alida ion se s (2/3 and
1/3 o he samples, espec i ely) be o e being subjec ed o
se e al PCA: (A and B) join analysis o all samples; (C and
D) analysis o ne us and p ima y melanoma lipidomes;
(E and F) ne us e sus me as asis and (G and H) me as asis
e sus p ima y melanoma. In all cases, a classi ie buil
using logis ic eg ession and he disc iminan lipids
selec ed in he aining se , was able o achie e a pe ec
classi ica ion o he alida ion g oup samples (Table S1 o
he supplemen al ma e ial).
HUERGO-BAÑOS ET AL.719
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
4|DISCUSSION
The esul s p esen ed he ein clea ly demons a e ha a subs an ial
change in he lipid inge p in accompanies umo ans o ma ion.
Such changes enable building classi ica ion models wi h pe o mances
be ween 90% and 100% (depending on he algo i hm) when biopsies
om ne us, p ima y melanoma and me as a ic melanoma a e included
and wi h pe ec pe o mance in bina y compa isons. A consequence
o his obse a ion is ha LIMS can be used in conjunc ion wi h such
classi ie s o iden i y umo cells di ec ly om issue sec ions, opening
he doo o he use o his echnique in clinics o as and au oma ed
sample sc eening.
Compa ison o he ela i e abundance o he main lipid classes
eco ded by LIMS may be ound in Figu e 5, while he po en ial bio-
ma ke s a e collec ed in Figu e S2. The e a e s a is ically signi ican
a ia ions in lysophospha idylinosi ol (LPI), phospha idylcholine/
phospha idyle hanolamine (PC/PE), PC/PE e he /plasmalogens (O/P),
phospha idylglyce ol (PG) and SM classes. PC and PE we e g ouped
oge he in a single g aphics, due o he exis ence o isoba ic species
con ibu ing o each peak in he mass spec um. Simila ly, PC/PE O/P
we e also g ouped in a single class. Appa en ly, he e is an inc ease o
PC/PE in p ima y melanoma, which is pa ly e e ed in me as asis.
In e es ingly, he changes in PC/PE O/P ollow an opposi e end o
hose o he di-acyl PC/PE species. Also signi ican is he dec ease in
LPI de ec ed in p ima y melanoma. Finally, he e is an inc ease in PG
and a dec ease in SM om ne us o melanoma ha a e again pa ially
e e ed in me as asis.
De ailed analysis o he ela i e abundance o he indi idual spe-
cies may be ound in Figu e S2 and Table S2. A close look a he PI
class—signi ican ly in ol ed in signaling—shows a dec ease in he ela-
i e abundance o a achidonic acid (AA, 20:4)-con aining species, and
an inc ease in he species wi h mono- and di-unsa u a ed a y acids
(MUFA and DUFA) om ne i o p ima y melanoma. Fo some o he
species, his end is e e ed in me as asis, which exhibi s exp ession
le els close o hose o ne us. This kind o ans-acyla ion p ocess
would explain he absence o changes in he ela i e abundance o
o al PI be ween he h ee condi ions and i was also obse ed in
o he sys ems.
18
A educed numbe o LPI and LPI-O/P species p esen ed changes
be ween samples. The mos ema kable a ia ion is he dec ease in
he ela i e abundance o LPI 20:4, which mimics hose obse ed
in AA-con aining PI species. I is emp ing o specula e ha his esul s
om a highe demand o AA o p o ision o in lamma ion- ela ed
eicosanoid media o s.
A limi ed numbe o PG and PS species we e also de ec ed, bu in
bo h cases, he e is an inc ease in he ela i e abundance o MUFA/
DUFA-con aining species and a dec ease in PUFA in p ima y mela-
noma. Such end is e e sed in me as asis, showing a mo e simila
composi ion o ha ound in melanocy es.
The analysis in Figu e 4may be ega ded as a simpli ied app oach.
Ce ainly, he e is s ill a weal h o in o ma ion p esen in he LIMS
da a ha was no exploi ed in he p esen wo k, such as he p esence
o in il a ing cells, a eas wi h nec osis, cell densi y o he p esence o
di e en clones o umo cells. Al hough in e p e a ion is no simple,
c ea ion o a lib a y o lipid inge p in s o au oma ically anno a e he
issue sec ions may yield in aluable in o ma ion o he diagnosis and
p ognosis o he disease. Cons uc ion o such lib a y would equi e
o de ailed analysis o hund eds o samples and ca e ul compa ison
wi h he LIMS images. Figu es S3–S5 cons i u e a g adien in com-
plexi y and illus a e how LIMS is able o highligh he he e ogenei y
o umo samples. The Melan A image o a p ima y melanoma in
Figu e S3 shows e y homogenous umo cells, in b own, o ming
nes s, su ounded by lymphocy es and some eosinophils in blue. Such
o ganiza ion is clea ly cap u ed by he segmen a ion image o he co -
esponding LIMS expe imen , in which he umo cells a e g ouped in
he ed segmen , while he in lamma o y cellula i y appea s in blue
and he de mis in whi e. In e es ingly, all umo cells show a simila
lipidome. Ac ually, inc easing he numbe o segmen s o he image
did no di ide he umo nodules in di e en g oups, demons a ing
ha hey uly p esen non-di e en iable lipidomes. The obse a ions
in Figu e S3 con as wi h he indings shown in Figu e S4 om a di -
e en p ima y melanoma. In his second example, he lymphocy es
a e g ouped o ming a lymphoid ollicle su ounded by umo and
appea in he same segmen as he sub-epide mal lymphocy es in e -
mixed wi h he epide mis. He e, he umo melanocy es a e g ouped
in h ee di e en segmen s: ed, whi e and b own. Fu he inspec ion
FIGURE 5 Signi ican changes in lipid
inge p in be ween condi ions. Values a e
exp essed as he ela i e abundance o he
co esponding lipid g oup, 100% being he
abundance in ne us melanocy es. PC and PE a e
p esen ed oge he as “PC”due o he
(p esumably) la ge numbe o isoba ic species
con ibu ing o each mass channel. Likewise, PC-
O/P and PE-O/P a e p esen ed as PC-O. *p< .05;
**p< .01; ***p< .001 e sus ne us.
720 HUERGO-BAÑOS ET AL.
10970215, 2024, 4, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1002/ijc.34800 by Uni e sidad Del Pais Vasco, Wiley Online Lib a y on [02/05/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License