npj | imaging A icle
h ps://doi.o g/10.1038/s44303-024-00041-3
MALDI imaging combined wi h wo-
pho on mic oscopy e eals local
di e ences in he he e ogenei y o
colo ec al cance
Check o upda es
A o a Bha i1, Kulka ni Ajinkya1, Ma kus M. And ea1, Ramos-Gomes Fe nanda1, Bohnenbe ge Hanibal2,
S öbel Philipp2, Al es F auke1,3,4,6 & Klein Oli e 5,6
Colo ec al cance (CRC) emains a leading cause o cance - ela ed mo ali y wo ldwide, accen ua ed
by i s he e ogenei y and complex umou mic oen i onmen (TME). The ole o TME on umou
pa hophysiology is pi o al, especially he influence o componen s o he ex acellula ma ix (ECM),
such as collagen. We in oduce a no el mul imodal imaging s a egy o un a el he complex spa ial
he e ogenei y o CRC by in eg a ing he imaging ea u es om wo-pho on lase scanning mic oscopy
(2PLSM) and his ology wi h p o eomics signa u es om ma ix-assis ed lase deso p ion ioniza ion-
mass spec ome y imaging (MALDI MSI). Ou s udy is he fi s o co ela e he s uc u al cohe ence o
collagen fib es and he nuclei dis ibu ion p ofile o umou issue wi h he pep ide signa u es, o e ing
insigh s in o he p o eomic landscape o CRC wi hin egions o high nuclei dis ibu ion (HND), as well
as chao ic and o ganised egions o collagen. We use his app oach o dis inguish he pa ien issues
o igina ing om le -sided colo ec al cance (LSCC) and om igh -sided colo ec al cance (RSCC).
This disc imina i e signa u e highligh s he ole o high nuclei dis ibu ion and collagen a chi ec u e in
umou p og ession. Complemen a y m/z alues o se e al p o eins associa ed o componen s o
ECM, such as plec in, inculin, imen in, and myosin, ha e shown di e en ially in ensi y dis ibu ions
be ween LSCC and RSCC. Ou findings demons a e he po en ial o combining s uc u al in o ma ion
wi h pep ide ea u es o iden i y molecula signa u es in di e en umou egions and e ie e new
insigh s in o CRC pa hophysiology.
The GLOBOSCAN 2020 epo by he Wo ld Heal h O ganiza ion calls
a en ion o he inc easing numbe o colo ec al cance (CRC) cases in he
wo ld, wi h he hi d-highes incidence a e and he second-highes mo ali y
a e in he wo ld1. These cases ha e been p edic ed o inc ease by 71.5% by
20352. E en hough he ole o he umou mic oen i onmen (TME) in
influencing cance p og ession, me as asis, and esponse o chemo -
adio he apy is inc easingly ecognised, he spa ial he e ogenei y o he TME
s ill poses a p oblem o bo h sub yping and he apy3,4.Theinfluence o he
ex acellula ma ix (ECM), pa icula ly collagen, on he umou pa hophy-
siology is also appa en 5,6. His ology wi h haema oxylin and eosin (H&E)
s aining, along wi h colonoscopy, x- ay imaging, and molecula sc eening
(mu a ions in KRAS o mic osa elli e ins abili y), a e ou inely used me hods
o moni o ing he p ognosis o CRC in clinics7,8. Howe e , hese me hods
canno p o ide he cha ac e isa ion o he spa ial complexi y o umou issues,
such as he in e play o cance p o eome, collagen a chi ec u e, and al e a ions
in he nuclei dis ibu ion ha influence he umou pa hophysiology.
1T ansla ional Molecula Imaging, Max-Planck-Ins i u e o Mul idisciplina y Sciences, He mann Rein ‑S aße 3, 37075 Gö ingen, Ge many. 2Ins i u e o
Pa hology, Uni e si y Medical Cen e Gö ingen, Robe -Koch-S aβe 40, 37075 Gö ingen, Ge many. 3Clinic o Haema ology and Medical Oncology, Ins i u e o
In e en ional and Diagnos ic Radiology, Uni e si y Medical Cen e Gö ingen, Robe -Koch-S aβe 40, 37075 Gö ingen, Ge many. 4Clus e o Excellence
“Mul iscale Bioimaging: om Molecula Machines o Ne wo ks o Exci able Cells”(MBExC), Uni e si y o Goe ingen, Goe ingen, Ge many. 5Be lin Ins i u e o
Heal h a Cha i é –Uni e si ae smedizin Be lin, Co e Uni Imaging Mass Spec ome y, 13353 Be lin, Ge many.
6
These au ho s con ibu ed equally: Al es F auke,
Klein Oli e . e-mail: [email p o ec ed];oli e .klein@bih-cha i e.de
npj Imaging | (2024) 2:35 1
1234567890():,;
1234567890():,;
Two-pho on lase scanning mic oscopy (2PLSM) allows a label- ee
imaging app oach ha cap u es he in insic second ha monic gene a ion
(SHG) signal emi ed by non-cen osymme ic molecula s uc u es,
no ably collagen fib es9. The wo-pho on exci a ion au ofluo escence
(2PEF) signal concu en ly emi ed by he issue gi es in o ma ion on he
issue mo phology. The e o e, 2PLSM enables he isualisa ion o collagen
fib e a chi ec u e wi hin he issues a sub-mic on esolu ion wi hou
necessi a ing ex insic fluo opho es o a s aining p ocess10. Recen in es i-
ga ions ha e unde sco ed he significance o collagen fib e-associa ed
imaging ea u es in a ious clinical aspec s o CRC, such as p edic ing a
pa hological comple e esponse11, p o iding he immunosco e12, an asso-
cia ion o collagen signa u es wi h lymph node me as asis13 and de e mining
he p ognosis o CRC pa ien s. A ecen s udy has shown di e ences in he
collagen mo phology in he ECM o CRC based on he ana omical loca ion
o he umou , dis inguishing be ween LSCC and RSCC by he highe col-
lagen fib e con en ound in LSCC14.
Despi e hese ad ancemen s, a comp ehensi e unde s anding o how
collagen ea u es influence umou pa hophysiology emains elusi e. Spe-
cifically, he e is a no able gap in esea ch ega ding he spa ial co ela ion
be ween collagen-associa ed imaging ea u es and he p o ein landscape in
ECM s uc u es. Thus, o u he unde s and he mechanism by which
specific collagen ea u es a ec umou pa hophysiology, iden i ying he
p o eins in ol ed in he dynamic emodelling o ECM du ing umou
p og ession is o p ime impo ance.
MALDI mass spec ome y imaging (MALDI MSI) combines spa ial
and label- ee molecula p ofiling and his ological e alua ion di ec ly in issue
sec ions15,16. This imaging modali y combines he capabili y o mass spec-
ome y o iden i y molecules (pep ides, lipids, d ugs, and o he compounds)
while p ese ing spa ial in o ma ion17. Recen s udies ha e exploi ed MALDI
MSI in deciphe ing CRC pa hophysiology, including i) he iden ifica ion o
p o eomics signa u es ha can help o p edic dis an me as asis in CRC
pa ien issues18, ii) he co ela ion o p o eomics signa u es o clinical ou -
come in CRC pa ien issue mic oa ays19, iii) he combina ion wi h che-
mome ic analysis o disc imina e CRC umou cell popula ions20,andi )
e en he epo o abno mal phospholipids in CRC li e me as asis21.
Howe e , na owing down he ele an pep ides o in e es om an
ex ensi e a ay o de ec ed p o eomics signa u es poses a challenge. In his
s udy, we p esen a new app oach o cha ac e ise dis inc i e a eas o he
TME o CRC pa ien issues by combining, o he fi s ime, imaging
ea u es de i ed om 2PLSM and his ology wi h p o eomics signa u es
om MALDI MSI.
We fi s de i ed a hea map o he cohe ence in he collagen fib e
o ganisa ion in umou issues be o e pe o ming MALDI MSI, ollowed by
gene a ing hea maps o he nuclei dis ibu ion in he issue by using his-
ology. These imaging bioma ke s we e used o efine he p o eomic sig-
na u es de ec ed by he MALDI MSI. We p esen his app oach o
co ela i e MALDI MSI and 2PLSM exempla y o he examina ion o
di e ences be ween he p o eomics signa u es o he le -sided colo ec al
cance (LSCC) and he igh -sided colo ec al cance (RSCC).
Ma e ial and me hods
Tumou sample collec ion
The s udy coho consis s o human umou issues esec ed om pa ien s
wi h CRC (n= 14). The coho has a balanced dis ibu ion o he ana omical
o igin o he umou , s ages and mo phology o he umou , as de ailed in
he sample in o ma ion able (Table 1:Samplein o ma ion).Thesamples
we e collec ed and used acco ding o he app o al and egula ions o he
E hics Commi ee o he Uni e si y Medical Cen e Goe ingen (E hics
app o al: #5/10/17). All pa ien s p o ided w i en in o med consen o
using pa hology specimens o esea ch pu poses. The s udy was pe o med
in acco dance wi h he Decla a ion o Helsinki.
Label- ee imaging
2PLSM images we e acqui ed using an up igh T iM Scope II mul ipho on
mic oscope (Mil enyi Bio ec, Biele eld, Ge many) equipped wi h a unable
em osecond lase (C onus 2 P, Ligh Con e sion, Vilnius, Li huania). The
lase was uned o 870 nm a 15% o i s maximal ou pu powe (1.0 W) o
SHG and 2PEF exci a ion. Images we e acqui ed using an Olympus
XLPLAN 25x (NA 1.05) wa e imme sion objec i e. The backsca e ed
emi ed ligh was spli by 495 nm and 560 nm long pass dich oic mi o s
(Sem ock) and de ec ed h ough he objec i e lens a pho omul iplie ube
(PMT) and GaAsP de ec o s (Hamama su). A PMT in he ansmission
posi ion collec ed he o wa d sca e ed emi ed ligh ga he ed by a 1.4 NA
condense lens unde he s age. SHG and2PEFsignalswe ecollec edusing
he fil e se ings 434 ± 20 nm and 525 ± 50 nm, espec i ely (B igh Line
HD fil e s, Sem ock, AHF Analysen echnik Tübingen, Ge many). Collagen
fib e signals we e collec ed as o wa d sca e ed SHG (F-SHG). The 2PEF
signal was collec ed om backsca e ed pho ons. 2D mosaics o en i e
sec ions we e acqui ed wi h indi idual image sizes o 393 × 393 µm wi h
1024 × 1024 pixels, a equency o 600 Hz wi h 2- old a e aging, and 10%
o e lap wi hin each ile o he mosaic.
MALDI IMS sample p epa a ion
All issue samples we e fixed in 4% pa a o maldehyde and embedded in
pa a fin. 5 µm hick sec ions we e cu om he pa a finblocksusinga
ib a ome (VT1000 S; Leica Biosys ems) and moun ed on conduc i e glass
slides coa ed in indium in oxide (B uke Dal onik GmbH, B emen, Ge -
many). The sec ions on he slides we e co e ed wi h a co e slip be o e
imaging by 2PLSM. A e 2PLSM imaging, he co e slip was emo ed, and
he samples we e p epa ed o MALDI MSI as de ailed by Wu e al.24.
Sec ions we e p ehea ed o 80 °C o 15 min. Depa a finisa ion was pe -
o med by successi e imme sion in xylene, 100% isop opanol, and suc-
cessi e hyd a ion s eps o 100%, 96%, 70%, and 50% e hanol o 5 minu es
each. Hea -induced an igen e ie al was pe o med in MilliQ-wa e in a
s eame o 20 min. A e d ying he slides o 10 min, yp ic diges ion was
pe o med using an au oma ed sp aying de ice (HTX TM-Sp aye , HTX
Technologies LLC, ERC GmbH, Rieme ling, Ge many). 16 laye s o yp ic
solu ion (20 µg P omega®Sequencing G ade Modified Po cine T ypsin in
800 µL diges ion bu e -20 mM ammonium bica bona e wi h 0.01% gly-
ce ol, flow a e = 0.015 ml/min eloci y = 750 mm/min, ack spacing:
2 mm, concen a ion = 0.025 mg/ml, pa e n CC) we e sp ayed on o each
sec ion a 30 °C. Tissue sec ions we e hen incuba ed o 2 h a 50 °C in a
humidi y chambe sa u a ed wi h po assium sulpha e solu ion. A e
incuba ion, he HTX TM Sp aye applied 4 laye s o he ma ix solu ion
(7 g/L a-cyano-4-hyd oxycinnamic acid in 70% ace oni ile and 1% i-
fluo oace ic acid, flow a e = 0.012 ml/min eloci y = 1200 mm/min, ack
spacing: 3 mm, concen a ion = 7 mg/ml, pa e n HH) a 75 °C.
MALDI MSI
MALDI imaging was pe o med on he apifleX®MALDI Tissue ype ®
(B uke Dal onik GmbH, B emen, Ge many) in eflec o mode wi h a
de ec ion ange o 800–3200 m/z,500lase sho spe spo ,a1.25GS/s
sampling a e and as e wid h o 50 µm. FlexImaging 5.1 and flexCon ol
3.0 so wa e (B uke Dal onik GmbH) we e used in coo dina ion. Ex e nal
calib a ion was pe o med using a pep ide calib a ion s anda d (B uke
Dal onik GmbH).
P o ein iden ifica ion
Iden ifica ion o p o eins based on m/z alue was conduc ed on adjacen
issue sec ions u ilising a bo om-up app oach in ol ing nano-liquid
ch oma og aphy elec osp ay ioniza ion andem mass-spec ome y, as
ou lined in p e ious wo k22. Fi s ly, he samples we e p ehea ed o 15 min
a 80 °C, ollowed by depa a finiza ion, an igen e ie al, and yp ic
diges ion. The samples we e hen incuba ed o 2 h a 50 °C in a humidi y
chambe filled wi h a po assium sulpha e solu ion. Pep ides we e sepa a ely
ex ac ed om each issue a e 2 h in o 40 µL o 0.1% ifluo oace ic acid
(TFA) and hen incuba ed o 15 min a oom empe a u e. As pe he
manu ac u e ’s guidelines, he diges s we e fil e ed h ough a ZipTip®C18,
and he fil a es we e concen a ed using a acuum e apo a o (Eppendo ®
Concen a o 5301, Eppendo AG, Ge many) and e-dissol ed in 0.1%
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 2
TFA. 2 µL o he pep ide mix was loaded o an Acclaim PepMap™100 C18
ap column (100 µm × 2 cm, PN 164564, The mo Fishe Scien ific, USA)
and p e- ea ed wi h 10 mM sodium hypofluo i e (flow a e 20 µL/h) be o e
being sepa a ed on an Acclaim PepMap™RSLC C18 column (75 µm ×
50 cm, PN 164942, The mo Fishe Scien ific, USA) using a 2–35% ace -
oni ile g adien in 0.1% o mic acid (flow a e 400 nL/min, p essu e ange
10–800 ba ) o e 90 min a 60 °C. The andem mass spec ome e (Impac
II, ESI QTOF MS, B uke Dal onik GmbH, B emen, Ge many) de ec ed he
ionized pep ides h ough a ull-mass scan (150–2200 m/z) a a 50,000
FWHM esolu ion. The au oMS/MS Insan Expe ise ea u e selec ed peaks
o agmen a ion ia collision-induced dissocia ion. Fo pep ide iden ifi-
ca ion, he peak lis s we e sea ched in he human Swiss-P o da abase using
PEAKS-s udio-p o eomics so wa e (Bioin o ma ics Solu ions, Ve sion
11.6) using PEAKS-DB and PEAKS-de-no o sequencing. Fo MALDI MSI
and LC–MS/MS m/z alue compa isons, iden ifica ion equi ed mo e han
one pep ide wi h mass di e ences <0.15 Da. Pep ides ha had he highes
sco e (-10lolgP, p o ein confidence sco e) and he smalles mass di e ences
be ween he MALDI MSI and LC-MS/MS da ase s we e assumed as iden-
ified. Deiso oping was pe o med.
His ology
Following MALDI -MSI, he ma ix was emo ed om he issue sec ions
wi h 70% e hanol, and he sec ions we e hen washed in dis illed wa e o
7 min. They we e s ained wi h haema oxylin (Himedia) o 10 min and
washed wi h wa m wa e wice o 10 min each. This was ollowed by
washing wi h dis illed wa e . The slides we e hen s ained wi h eosin
(Himedia) o 60 s and washed wice wi h dis illed wa e . Se ial dehyd a ion
was pe o med by imme sing he slides in 80%, 96%, 100% e hanol, and
xylol, ollowed by moun ing hem wi h a co e slip. The slides we e hen
scanned wi h a whole-sec ion scanne (NanoZoome , Hamama su). These
haema oxylin and eosin (H&E) images we e used o anno a e umou
egions in Qupa h SW23 by a pa hologis be o e ans e ing hem in o SCiLS
Lab so wa e e sion 2024a P o (12.00.15110).
Image p ocessing
The F-SHG signal om 2PLSM images was ex ac ed and denoised using
he global O su adap i e h esholding. Tex u e analysis was pe o med on
he SHG signal as explained be o e14 o quan i y he cohe ence o collagen
fib e. The local cohe ence hea map is bound be ween alues 0 and 1, whe e 1
indica es highly aligned s uc u es, and 0 indica es chao ic and diso ganised
fib es. This cohe ence map was o e laid on he o iginal 2PLSM image in FIJI
by me ging images o s ack, ollowed by aking a maximum in ensi y p o-
jec ion o he me ged s ack. The cohe ence egions we e classified as chao ic
(local cohe ence alues om 0 o 0.5) and o ganised (local cohe ence alues
om 0.5 o 1). The pe cen age a ea occupied by chao ic and o ganised
egions was quan ified.
The nuclei we e segmen ed om he H&E images using he
“2D_ e sa ile_he”p e- ained model om S a Dis 24. Du ing pos nuclei
segmen a ion, he nuclei dis ibu ion was quan ified by iden i ying he
numbe o nuclei pe uni a ea. A hea map o nuclei dis ibu ion was
o med, bound be ween 0 and 1, whe e 1 indica es densely packed nuclei,
and 0 indica es nuclei sca e ed la gely in space. This nuclei dis ibu ion map
was hen o e laid on he o iginal H&E image in FIJI by me ging images o
s ack, ollowed by aking a maximum in ensi y p ojec ion o he me ged
s ack. The nuclei dis ibu ion egions we e classified as low nuclei dis-
ibu ion ( alues om 0 o 0.5) and high nuclei dis ibu ion ( alues om 0.5
o 1). The pe cen age a ea occupied by low and high nuclei dis ibu ion
egions was quan ified. The nuclei segmen a ion and ex u e analysis sou ce
codes a e a ailable a dedica ed Gi Hub eposi o ies25,26.
STRING analysis
P o ein-p o ein associa ions we e de i ed om he STRING da abase
(h ps://s ing-db.o g/). The lis o UniP o IDs o pep ides iden ified by
MALDIMSIwasaddedin he‘Mul iple p o eins’sea ch o he o ganism
Homo sapiens and he confidence in e al sco e o 0.4 o gene a e a ull
s ing ne wo k indica ing bo h unc ional and physical edges. The hicke
edge indica es he s eng h o he da a suppo . Func ional en ichmen
analysis was pe o med by applying he ollowing se ings: Maximum
FDR < 0.01, s eng h >0.01 and minimum coun in ne wo k.
MALDI MSI da a p ocessing o s a is ics
MALDI MSI aw da a we e impo ed in o he SCiLS Lab so wa e e sion
2024a P o (B uke Dal onik GmbH) using se ings o p ese e he o al ion
coun wi hou baseline emo al and con e ed in o he SCiLS base da a .sbd
and .slx file. An a ibu e able was made, which consis s o anno a ed
egions o LSCC, RSCC, high collagen cohe ence, low collagen cohe ence,
high nuclei dis ibu ion and low nuclei dis ibu ion. These a ibu es we e
used o di ide he da ase in o espec i e spa ial g oups o compa e he
di e ences in he spec a in pa icula egions o he issues. Peak finding
and alignmen we e conduc ed ac oss a da ase using a s anda d
Table 1 | Sample in o ma ion
S.No. TNM classifica ion Gende Age TNM s age G ade (G2/G3) LSCC/RSCC
1 pT3 N0 (0/19) L0 V0 Pn0 G2 R0 M 83 T3 G2 RSCC
2 pT3 N1a (1/38) L1 V0 Pn0 G2 R0 F 88 T3 G2 RSCC
3 pT2 N0 (0/12) L1 V0 Pn0 G3 R0 M 81 T2 G3 RSCC
4 pT1 N0 (0/28) G2 R0 F 73 T1 G2 RSCC
5 pT3 N0 (0/25) M1 (PER) L1 V0 Pn0 G3 R0 M 49 T3 G3 RSCC
6 pT2 N1b (3/88) L0 V0 G2 R0 F 49 T2 G2 RSCC
7 pT3 N2a (6/47) L1 V0 G3 R0 F 75 T3 G3 LSCC
8 pT3 N0 (0/15) G2 R0 M 72 T3 G2 LSCC
9 pT3 N0 (0/15) L0 V1 Pn1 G2 R0 M 66 T3 G2 LSCC
10 pT2 N0 (0/12) L0 V0 Pn0 G2 R0 M 73 T2 G2 LSCC
11 pT3 N2b (8/25) M1a (HEP) L1 V1 Pn1 R0 G2 M 81 T3 G2 LSCC
12 pT3 N0 (0/16) L0 V0 G2 R0 M 59 T3 G2 LSCC
13 pT2 N0 (0/19) MX L0 V0 Pn0 G2 R0 F 66 T2 G2 LSCC
14 pT3 N1a (1/17) L0 V0 Pn0 R0 G2 M 95 T3 G2 LSCC
The umou s ha e been ca ego ised by he TNM s aging, whe e T deno es he s age ( om T1 o T4 in he ascending o de o agg essi eness), N indica es lymph node in ol emen , and M indica es he
me as asis s a us. The g ade gi es in o ma ion on he appea ance o umou cells, whe e G1 indica es well-di e en ia ed issues, G2 is mode a ely di e en ia ed, and G3 is poo ly di e en ia ed issues. All
he umou s a e mic osa elli e s able (MSS).
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 3
segmen a ion pipeline (SciLS Lab so wa e) in maximal in e al p ocessing
mode wi h TIC no malisa ion, medium noise educ ion and no smoo hing.
Anno a ed egions we e ans e ed om QuPa h as se .file in o SCilS Lab.
S a is ics e alua ion
Disc imina i e MALDI MSI m/z alues om anno a ed egions o each
g oup (LSCC, RSCC, high collagen cohe ence, low collagen cohe ence,
high nuclei dis ibu ion and low nuclei dis ibu ion) we e iden ified using
supe ised ecei e ope a ing cha ac e is ic (ROC) analysis. The a ea
unde hecu e (AUC) a ies be ween 0 and 1, whe e alues close o ei he
0 o 1 indica e disc imina o y pep ides and alues close o 0.5 indica e ha
he pep ides ha e a simila dis ibu ion in he g oups. The p esence o
mul iple pep ides o he same p o ein wi h a simila ROC alue (all
pep ides wi h AUC ≥0.6 ≤0.4) was conside ed a good indica o o he
p esence o p o ein. The pep ides wi h an AUC ≥0.6 ≤0.4 we e selec ed as
candida e ma ke s o p incipal componen analysis (PCA) and seg-
men a ion using bisec ing k-means clus e ing analysis on he espec i e
g oups.
A wo-way ANOVA was pe o med o compa e he pe cen age a ea
occupied by chao ic egions agains o ganised egions and simila ly o
compa e he pe cen age a ea occupied by he high agains low nuclei dis-
ibu ion egions in he LSCC /s RSCC. S a is ical analysis was pe o med
wi h G aph Pad P ism 9. A p- alue o 0.05 was conside ed as a ma gin o
s a is ical significance. Da a is p esen ed as mean ± SD, *indica es p≤0.05,
** indica es p≤0.01, *** indica es p≤0.001, **** indica es p≤0.0001.
Figu es we e c ea ed using he SCiLS Lab so wa e (B uke , B emen,
Ge many, 2024a P o), bio ende , and Inkscape, and g aphs we e c ea ed
using G aphPad P ism 9.
Resul s
Imaging p o ocol o co ela ion o 2PLSM, MALDI MSI and
his ology
We de eloped an imaging and analysis pipeline o in eg a e he imaging
ea u es om 2PLSM and his ology wi h he p o eomics signa u es om
MALDI MSI. Ini ially, we scanned he issue sec ions wi h 2PLSM o cap u e
he SHG and 2PEF signals. Subsequen ly, we pe o med MALDI MSI on
hese sec ions, ollowed by H&E s aining (Fig. 1). We used he SHG signal
emi ed by collagen fib es o o m a hea map. This hea map isually
ep esen s he deg ee o o ganisa ion o collagen fib es wi hin he issue,
dis inguishing a eas based on he cohe ence o collagen fib es. The egions
inside each issue we e ca ego ised in o wo g oups: egions wi h high
cohe ence o collagen fib es, called ‘o ganised egions’, and egions wi h low
cohe ence in collagen fib es, called ‘chao ic egions. Concu en ly, we seg-
men ed he nuclei om H&E images and o med a hea map indica ing he
dis ibu ion o nuclei in he issue. This allowed us o seg ega e he issue
egions in o wo g oups based on nuclei dis ibu ion — egions wi h high
nuclei dis ibu ion and egions wi h low nuclei dis ibu ion. Finally, we
compa ed he p o eomics signa u es ac oss hese ou mo phologically
dis inc ca ego ies — egions wi h high and low collagen fib e cohe ence and
egions wi h high and low nuclei dis ibu ion.
Fig. 1 | Imaging p o ocol o co ela ion o 2PLSM, MALDI MSI and his ology.
The umou issues we e fi s scanned by label- ee 2PLSM. The SHG signal gen-
e a ed by he collagen was p ocessed o gene a e a hea map depic ing he deg ee o
cohe ence in he collagen fib es. Following 2PLSM, MALDI MSI was pe o med on
he umou issues. Spa ial hea maps we e gene a ed, depic ing he deg ee o
exp ession o each pep ide in he issue. A e MALDI MSI, H&E s aining was
pe o med on he umou issues. The nuclei segmen ed om he H&E image we e
used o gene a e a hea map. The pep ide exp ession in he egions o high nuclei
dis ibu ion and high and low cohe ence we e analysed.
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 4
The ollowing image analysis pipeline was in oduced o o m he
hea maps and anno a e he egions (Fig. 2). We fi s segmen ed om H&E
images (Fig. 2A) he nuclei (Fig. 2B). The numbe o nuclei pe uni a ea o
he issue was used o o m a hea map, whe e ed ep esen s he egions wi h
he highes nuclei dis ibu ion, and blue he egions wi h he lowes nuclei
dis ibu ion (Fig. 2C). The hea map was hen o e laid on he H&E image
(Fig. 2D).
This me ged file was used o anno a e he high and low nuclei dis-
ibu ion egions in he o iginal H&E files in QuPa h, which we e sub-
sequen ly impo ed in o he SCiLS lab so wa e as classes using he
QuPa h-SCiLS plugin. Simila ly, o o m he hea map o he cohe ence o
collagen fib es, he SHG signal (Fig. 2F) was fi s ex ac ed om he
2PLSM image (Fig. 2E), whe e g een ep esen s 2PEF o au ofluo escence
and whi e ep esen s SHG. The au ofluo escence signal emi ed om he
issue gi es a issue mo phology simila o ha o he H&E s aining.
Tex u e analysis was pe o med on he SHG signal, and a hea map
indica ing he cohe ence in collagen fib es was o med (Fig. 2G), whe e
yellow ep esen s egions wi h high cohe ence in collagen fib es and iole
chao ic egions. The hea map was hen o e laid on he 2PLSM image
(Fig. 2H). This me ged file was used o anno a e he high and low
cohe ence egions in he o iginal H&E files in QuPa h, which was sub-
sequen ly impo ed in o he SCiLS lab so wa e as he classes using he
QuPa h-SCiLS plugin.
Disc imina i e p o eomic signa u es in LSCC in compa ison o
RSCC issue specimens
Uni a ia e analysis o MALDI MSI da a esul ed in single pep ides which a e
di e en ially spa ially dis ibu ed be ween LSCC and RSCC umou issue.
520 aligned m/z peaks we e de e mined. F om hese, 203 m/z alues could
be assigned o 83 p o eins om he LC-MS analysis (Supplemen a y Table 1
and Da a 1). Based on hese assigned and aligned m/z alues ecei e
ope a o cha ac e is ics analysis (ROC) we e pe o med and esul ed in
disc imina i e 76 m/z alues (AUC ≥0.6 ≤0.4; p< 0.01) om 31 co e-
sponding p o eins be ween LSCC and RSCC (Supplemen a y Table 2).
Among he pep ides iden ified as disc imina i e, se e al showed pa icula ly
high ROC alues, indica ing an inc eased spa ial in ensi y dis ibu ion in
RSCC umou issues compa ed o LSCC.
Co esponding pep ides om ATPase sa coplasmic/endoplasmic
e iculum Ca2+ anspo ing 2 (ATP2A2, m/z 1477, AUC = 0.25), alpha-
enolase (ENO1, m/z 1962, AUC = 0.26) and inculin (VLC, m/z 1493,
AUC = 0.27) showed dec eased spa ial in ensi y dis ibu ion in LSCC in
compa ison o RSCC. In con as , pep ides om euka yo ic ansla ion
elonga ion ac o 1 alpha 1 (EEF1A1, 1 m/z 1588, AUC = 0.66) showed
inc eased spa ial in ensi y dis ibu ion (selec ion is shown in Fig. 3).
Subsequen ly, unc ional anno a ion was pe o med o anno a e he
highly ep esen ed gene on ology (GO) in o he subon ology “molecula
unc ion”(Supplemen a y Da a 2, and Supplemen a y Figu e 1). The
Fig. 2 | Image analysis pipeline. A H&E image o a ep esen a i e CRC pa a fin
sec ion depic ing issue mo phology. BThe image shows nuclei segmen ed om he
H&E image. CA hea map is gene a ed om he segmen ed nuclei and depic s he
dis ibu ion o nuclei. DO e lay o nuclei dis ibu ion hea map and H&E image o
anno a e he high nuclei dis ibu ion egions. E2PLSM image depic ing he issue
mo phology by he au ofluo escence signal (in g een) and he collagen-de i ed SHG
signal (in whi e). FCollagen fib e isualisa ion a e ex ac ing he SHG signal.
GHea map depic ing he deg ee o cohe ence in he collagen fib es gene a ed a e
ex u e analysis o he SHG signal. HO e lay o a cohe ence hea map and a 2PLSM
au ofluo escence image o anno a e he high and low cohe ence egions. Scale ba s in
all images ep esen 1000 µm.
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 5
p o eins plec in (PLC) and inculin (VCL) could be de e mined wi h a
highe in ensi y dis ibu ion in RSCC in compa ison o LSCC. These p o-
eins a e in ol ed in dys oglycan binding (GO:0002162) which is highly
en iched in he da ase . Mo eo e , he p o eins alin-1 (TLN1) and synemin
(SYNM) showed a s ong in e ac ion in inculin binding (GO:0017166).
Co esponding pep ides om inculin (VLNC) also showed an inc eased
in ensi y dis ibu ion in RSCC in con as o LSCC.
In o de o explo e he po en ial o hese 76 disc imina i e m/z alues,
p incipal componen analysis (PCA) was pe o med (AUC > 0.6 o <0.4).
This esul ed in an inc eased in ensi y dis ibu ion o PC1 in he umou
egionso RSCCincompa ison oLSCC(selec ionisshowninFig.4A). The
loadings plo shows di e ences be ween LSCC (Fig. 4B, shown in ed) and
RSCC (Fig. 4B, shown in black) in he plo o PC1 (x-axis) agains PC3 (y-
axis). The fi s h ee p incipal componen s accoun ed o 80.34% o he
a iabili y in he da a (Fig. 4C). Subsequen ly, he disc imina i e pep ides
we e used o gene a e p o eomic clus e s by he bisec ing k-means me hod.
The dis ibu ion o he gene a ed p o eomic clus e s (segmen s indica ed by
espec i e colou s) showed ma ked di e ences be ween LSCC and RSCC
(Fig. 5). The ed segmen is highly ep esen ed in LSCC (64%) in con as o
RSCC issue (27%). Mo eo e , he RSCC umou egion includes a highe
numbe o u he segmen s (e.g. blue 27.18%) (Supplemen a y Table 7).
High nuclei dis ibu ion egions ha e di e en ially exp essed
pep ide p ofiles in LSCC compa ed o RSCC
Based on 2PLSM, we ound ha he high nuclei dis ibu ion (HND) egions
occupied a significan ly lowe a ea (p< 0.0001) compa ed o he low nuclei
dis ibu ion (LND) egions in bo h LSCC and RSCC issues (Fig. 6A).
Howe e , he pe cen age a ea occupied by he high and low nuclei dis-
ibu ion egions did no di e be ween LSCC and RSCC. In o de o
de e mine di e ences in p o eomics composi ion o umou cell- ich, HND
egions agains he LND, a ROC analysis was pe o med based on he
MALDI MSI da a. The pai ed compa ison o he HND agains he LND
egions de e mined 14 disc imina i e m/z alues (Supplemen a y Table 4)
which co esponded o 7 p o eins. Co esponding m/z om collagen ype I
alpha 2 (COL1A2), ca hepsin (CTSG), elonga ion ac o 1-alpha 1
(EEF1A1), EH domain-con aining p o ein 2 (EHD2), epiplakin 1 (EPPK1),
p elamin-A/C (LMNA), and p ola gin (PRELP) showed highe in ensi y
dis ibu ion in HND in compa ison o LND umou a eas.
Mo eo e , we pe o med a pai ed compa ison o HND egions in
LSCC agains hose in RSCC. Ou o he 240 aligned and assigned m/z
alues, we de e mined 65 m/z alues ha we e disc imina i e and iden ified
in he HND egions o LSCC compa ed o RSCC (Supplemen a y Table 3).
65 disc imina i e m/z alues could be assigned o 25 p o eins. Th ee p o-
eins, caldesmon (CALD), he e ogeneous nuclea ibonucleop o ein M
(HNRNPM) and opomyosin 1 (TPM1) we e unique disc imina i e
be ween LSCC and RSCC umou cell- ich HND egions, howe e could
no be iden ified as disc imina i e when he whole LSCC and RSCC umou
egion we e analysed.
In o de o de e mine he disc imina i e alue o hese 65 m/z alues
we gene a ed p o eomic clus e s by he bisec ing k-means me hod. The
dis ibu ion o he gene a ed p o eomic clus e s (segmen s indica ed by
espec i e colou s) showed ma ked di e ences be ween he high nuclei
dis ibu ion egions o LSCC and RSCC (selec ion is shown in Fig. 6). We
quan ified he pe cen age a ea occupied by he high and low nuclei dis-
ibu ion egions (Fig. 6B, C; Supplemen a y Table 7).
Chao ic and o ganised collagen egions ha e disc imina i e
pep ides in LSCC e sus RSCC
The examina ion o cohe ence in he collagen fib e o ganisa ion in umou
issues based on 2PLSM esul ed in significan inc eased pe cen ages o
a eas wi h chao ic collagen compa ed o o ganised egions wi hin LSCC
(p= 0.0293) as well as RSCC umou egions (p= 0.0025) (Fig. 7E).
We did no find any di e en ial m/z alues be ween high and low
collagen cohe ence egions (da a no shown). Using MALDI MSI, he pai ed
compa ison o chao ic egions (HC) in LSCC agains RSCC by ROC analysis
e ealed 43 di e en ial m/z alues (assigned o 19 p o eins) (Supplemen a y
Table 5). Among he pep ides iden ified as disc imina i e we e hea shock
p o ein 90 be a amily membe 1 (HSP90B1, m/z 1515, AUC = 0.27),
Fig. 3 | MALDI-MSI disc imina i e p o eomic signa u es o LSCC in compa ison
o RSCC. ROC analysis iden ified di e ences in p o eomic signa u es be ween
LSCC (le -panel) and RSCC ( igh -panel). AO iginal H&E images o issues.
BInc eased in ensi y dis ibu ion o m/z co esponding o ATP2A2 in RSCC
compa ed o LSCC. CInc eased m/z alues co esponding o EEF1A1 in LSCC
compa ed o RSCC shown in he hi d ow. Scale ba s ep esen 6 mm.
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 6
Fig. 4 | P inciple componen analysis (PCA) e ealed di e ences in p o eomic
landscape be ween LSCC and RSCC. A PCA e eals a highe in ensi y dis ibu ion
o PC1 in RSCC ( igh -panel) han ha o LSCC (le -panel). BThe da a poin s in
LSCC, shown in ed, ha e a di e en da a poin cloud compa ed o RSCC, shown in
black. CFi s h ee p inciple componen s explain 80% o he a iabili y in he wo
g oups.
Fig. 5 | Di e ences in p o eomic clus e s be ween LSCC and RSCC. A The p o-
eomic clus e s gene a ed by bisec ing k-means di e be ween LSCC (le -panel) and
RSCC ( igh -panel). The spec a and dis ance o he pep ide clus e s a e shown on
ex eme le . BQuan ifica ion o p o eomic clus e s showed a di e ence be ween
pe cen age a ea occupied by clus e 1 in LSCC and RSCC.
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 7
ATPase sa coplasmic/endoplasmic e iculum Ca2+ anspo ing 2
(ATP2A2, m/z 1477, AUC = 0.27) and non-pou domain con aining oc a-
me binding (NONO, m/z 1554, AUC = 0.29), indica ing a highe in ensi y
dis ibu ion o hese pep ides in RSCC compa ed o LSCC.
The compa ison o o ganised o low chao ic egions (LC) in LSCC
agains RSCC by ROC analysis esul ed in 63 di e en ial m/z alues (25
p o eins, Supplemen a y Table 6). Among he pep ides iden ified as dis-
c imina i e, we ound ac inin alpha 1 (ACTN1, m/z 1961, AUC = 0.19),
enolase 1 (ENO1, m/z 1962, AUC = 0.21), ATPase sa coplasmic/endo-
plasmic e iculum Ca2+ anspo ing 2 (ATP2A2, m/z 1477, AUC = 0.22),
hea shock p o ein 90 be a amily membe 1 (HSP90B1, m/z 1515,
AUC = 0.23) indica ing a highe in ensi y dis ibu ion o hese pep ides in
RSCC compa ed o LSCC. In o de o in es iga e he impac o collagen
o ganisa ion a umou he e ogenei y segmen a ion was pe o med based
on di e en ial m/z alues o high and low collagen cohe ence. P o eomic
clus e s (segmen s indica ed by espec i e colou s) showed ma ked di e -
ences bo h in he o ganised (selec ion is shown in Fig. 7A) and chao ic
egions (selec ion is shown in Fig. 7B) be ween LSCC and RSCC. The
pe cen age a ea occupied by each segmen indica ed ha he p o eomic
clus e s in he o ganised (Fig. 7C) and chao ic (Fig. 7D) egions o RSCC a e
composed di e en ly in compa ison o LSCC (Supplemen a y Table 7). Fo
ins ance, he segmen 1 (colou ed in ed) is much mo e p onounced in
chao ic egions o LSCC in compa ison o RSCC.
Discussion
This s udy p esen s a new mul imodal imaging app oach o cha ac e ise he
umou mic oen i onmen (TME) in CRC pa ien issues. We show he
po en ial o combining s uc u al in o ma ion and he nuclei dis ibu ion
p ofile wi h pep ide ea u es o iden i y molecula signa u es in di e en
umou egions o un a el he complex spa ial he e ogenei y o CRC. MALDI
MSI has p e iously been coupled wi h a ious o he imaging modali ies.
Pa e son e al. co- egis e ed he au ofluo escence signal wi h MALDI MSI o
analyse lipid p ofiles in mu ine issue27. In ano he s udy combining his ology
wi h MALDI MSI, H&E images we e ained o di e en ia e umou om
non- umou a eas and au oma ed anno a ions we e used o compa e
me abolomic p ofiles in u achal adenoca cinoma, a a e ype o non-
u o helial malignancy, wi h colo ec al adenoca cinoma28.MALDIMSIsig-
na u es we e also co- egis e ed wi h con ocal immunofluo escence signals
o single-cell analysis by Niki ina e al.29. Fu he mo e, esea che s ha e
co ela ed MALDI MSI wi h o he imaging echnologies such as Magne ic
Resonance Imaging (MRI)30, Fou ie T ans o m In a ed Spec oscopy (FT-
IR)31 and Raman Spec oscopy32. Non-linea image egis a ion has been
used be o e o egis e adjacen his ology sec ions o MALDI MSI o apid
analysis33. He e, we demons a e, o he fi s ime, a wo kflow o co ela i e
2PLSM, his ology and MALDI and hei po en ial o ob aining pep ide
ea u es (m/z alues) o in e es om dis inc umou egions.
Fi s ly, we in es iga ed he di e ences in he m/z alues (pep ides)
be ween LSCC and RSCC using uni a ia e es ing (ROC analyses). We
iden ified ha p o eins co esponding o he m/z alues we e associa ed o
molecula unc ions o p o eins, e.g. dys oglycan binding (GO:0002162)
and inculin binding (GO:0017166). Dys oglycan is in ol ed in he cell’s
a achmen o ECM34 and i s educed exp ession is a p edic o o poo
ou come in CRC pa ien s35. Simila ly, loss o cell-cell adhesion due o loss o
inculin and memb ane-bound ca enin has been shown o p omo e
me as asis and is a p edic o o poo p ognosis in CRC36. Mul i a ia e es ing
using PCA based on he disc imina i e iden ified m/z alues demons a e
he high disc imina i e capaci y o hese pep ide signa u es. The po en ial o
hese m/z alues in un a elling issue he e ogenei y is shown by using he
segmen a ion. In LSCC, he fi s componen o segmen a ion is significan ly
mo e p onounced han in RSCC.
Spa ial he e ogenei y in colo ec al cance s, e.g. due o he p esence o
gene ically di e en umou cells o di e en immune cell dis ibu ion can
Fig. 6 | MALDI-MSI iden ified di e ences in p o eomic signa u es in he high
nuclei densi y (HND) egions o LSCC agains RSCC. A The pe cen age a ea
occupied by he HND and LND egions in he LSCC and RSCC is shown.
BRep esen a i e images o di e en ial p o eomic clus e s in he HND egions o
LSCC and RSCC, segmen ed by bisec ing k-means a e p esen ed. The spec a and
dis ance o he pep ide clus e s a e shown abo e. CQuan ifica ion o p o eomic
clus e s shown in Bhighligh s di e ences in pe cen age a ea occupied by all fi e
clus e s be ween HND egions o LSCC and RSCC. N= 14, **** signifies p< 0.0001.
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 8
influence umou p og ession and he apy esponse37,38. In o de o de e -
mine egion specific al e a ions be ween LSCC and RSCC in he s udy, we
classified he issues using 2PLSM and his ology in o ou spa ially dis inc
ca ego ies: low and high collagen cohe ence egions as well as low and high
nuclei dis ibu ion egions. P e ious s udies ha e highligh ed he impo -
ance o analysing he nuclea mo phology abno mali ies o cance p og-
nosis. Nuclea mo phome y was ound o be a p ognos ic de e minan in
CRC ca cinoma39–41. Mo eo e , benign and CRC issues ha e been classified
using a andom o es algo i hm on he ea u es ob ained om cell nuclei
segmen a ion on H&E images42. Väy ynen e al. iden ified ha a high nuclei
dis ibu ion o s omal lymphocy es and eosinophils is co ela ed o a be e
cance -specific su i al43. Resea che s ha e also used nuclea mo phome y
o ca ego ise high and low- isk b eas cance g oups44 o he iden ifica ion
o umou ma gins45 and classifica ion in o no mal, benign, in-si u ca ci-
noma and in asi e ca cinoma using con olu ional neu al ne wo ks46.These
s udies ad oca e o he clinical significance o looking mo e closely a he
high nuclei dis ibu ion a eas o umou . In ou s udy, we ound m/z alues
co esponding o 6 p o eins wi h a highe in ensi y dis ibu ion in he high
nuclei dis ibu ion egions compa ed o low. These a e COL1A2, CTSG,
EEF1A1, EHD2, EPPK1, LMNA, and PRELP. Some o hese, such as col-
lagen, laminin, p ola gin, and epiplakin a e s uc u al p o eins and o he s
such as elonga ion ac o 1 alpha 1 a e mainly p esen in he nucleus.
Laminin has been ound o p omo e umou budding in CRC by in e ac ing
wi h o he ex acellula ma ix (ECM) p o eins47 and o p omo e
me as asis48. EEF1A1 has also been shown o p omo e CRC p og ession49.
We u he analysed he di e ences in he m/z alues co esponding o
31 p o eins in he high nuclei dis ibu ion (HND) a eas o LSCC compa ed
o RSCC. The co esponding m/z alues o Non-POU domain-con aining
oc ame -binding (NONO) p o ein was ound wi h dec eased in ensi y
dis ibu ion in he HND o LSCC compa ed o RSCC. Mo eo e , exclusi e
in he HND egions co esponding o m/z alues om caldesmon 1
(CALD1) and opomyosin 1 (TPM1) showed highe in ensi y dis ibu ion
in RSCC han in LSCC. CALDI1 is a p ognos ic ma ke and a po en ial
he apeu ic a ge o s age III/IV pMMR (misma ch epai ) CRCs50.TPM1
ac s as a umou supp esso and i s loss p omo es cell p oli e a ion and
me as asis by inducing epi helial o mesenchymal ansi ion and egula ing
cy oskele al emodelling in CRC51.
In a nex s ep we selec ed “collagen o ganisa ion”as he second imaging
ea u e o his s udy. Mul iple s udies ha e unde sco ed he ole o collagen
fib e o ganisa ion in cance 52, including ou own, which showed significan
di e ences in he collagen fib eamoun ,wa inessandcohe encein umou s
o igina ing om LSCC agains RSCC14. S udies ha e shown ha collagen
fib es a e s aigh e in cance issues53, a e mo phologically dis inc in
adio he apy ea ed /s un ea ed pa ien issues54, and hese mo phological
ea u es o collagen can e en be use ul in p edic ing umou ecu ence in
CRC pa ien s55.Thesefindings unde line he need o in es iga e i p o eins
a e di e en ially exp essed in such spa ially dis inc egions o collagen fib e
o ganisa ion. Al hough no significan di e ences in he p o eomic sig-
na u es be ween chao ic and o ganised collagen fib e egions we e p esen ,
by compa ing highly and chao ic (low) o ganised egions o LSCC and
RSCC, we ound di e en ially exp essed m/z alues be ween he wo ana-
omical loca ions o umou .
The p o eins o co esponding m/z alues, namely alpha-ac inin-1
(ACTN1), So bin and SH3 domain-con aining p o ein 1 (SORBS1), alin-1
(TLN1) and plec in (PLEC) a e associa ed wi h a a ie y o biological
unc ions such as cell-subs a e junc ion assembly and ac in-filamen
o ganisa ion. Vimen in (VIM) was ound o be in ol ed in he caspase-
media ed clea age o cy oskele al p o eins, which may di ec ly con ibu e o
he apop o ic changes in cell shape56.
The p ognosis and esponse o ea men in CRC pa ien s a y
depending on he ana omical o igin o he p ima y umou , i.e., whe he i
de i ed om he le side o he igh side o he colon57.Since heECM,in
pa icula collagen o ganisa ion, plays a c i ical ole in umou
pa hophysiology58, analysing he p o eomic signa u es in highly diso ganised
Fig. 7 | Combina ion o MALDI-MSI wi h 2PLSM iden ifies di e en ial com-
posi ion o p o eomic clus e s in chao ic and o ganised egions o LSCC com-
pa ed o RSCC. A Segmen a ion by using k-means esul s in di e en ial p o eomic
clus e s in he o ganised egions o LSCC compa ed o RSCC. The spec a and
dis ance o pep ide clus e s a e shown beside he segmen a ion maps.
BSegmen a ion by using k-means esul s in di e en ial p o eomic clus e s in he
chao ic egions o LSCC compa ed o RSCC. The spec a and dis ance o pep ide
clus e s a e shown beside he segmen a ion maps. CQuan ifica ion o p o eomic
clus e s in he o ganised egions highligh s he di e ences in pe cen age a eas
occupied by each clus e be ween LSCC and RSCC. DQuan ifica ion o p o eomic
clus e s in he chao ic egions highligh s he di e ences in pe cen age a eas occupied
by each clus e be ween LSCC and RSCC. EThe pe cen age a ea occupied by he
o ganised and chao ic egions in he LSCC compa ed o RSCC is shown. N= 14,
*signifies p< 0.05, **signifies p< 0.01.
h ps://doi.o g/10.1038/s44303-024-00041-3 A icle
npj Imaging | (2024) 2:35 9
