1
A C i ical Re-e alua ion o “The o igin o
hepa ocellula ca cinoma depends on me abolic
zona ion” by Guo e al., Science 2025; ead 7129;
DOI: 10.1126/science.ad 7129
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
This commen documen s ex ensi e echnical, s a is ical, concep ual, and igu e-based
weaknesses in he s udy by Guo e al., Science (2025), which claims ha hepa ocellula
ca cinoma (HCC) o igin is dic a ed by me abolic zona ion. Ac oss mul iple igu es, he
au ho s’ da a do no con incingly suppo he headline conclusion. Many panels con lic
in e nally wi h he ex , se e al igu es con ain me hodological gaps se e e enough o
unde mine any causal in e ence, and he s udy’s cen al p emise elies on assump ions
con adic ed by he au ho s’ own da a. In se e al places, igu e cons uc ion, sample
sizes, and spa ial bounda ies a e ambiguous o inconsis en . Collec i ely, his body o
e idence does no uphold he claims made in he manusc ip .
1. O e all C i ique: The Cen al Claim Is
Unsuppo ed by he Figu es
The headline asse ion in he s udy by Guo e al.1 (Science 2025) — ha “ he o igin o
hepa ocellula ca cinoma depends on me abolic zona ion” — is no suppo ed by he
isual e idence.
E en he i s ew pages o he pape show con adic ions:
• The ex in oduces “unclea malignan po en ial” o ea ly ans o med
hepa ocy es, ye he pape la e in e s exac zonal o igins wi hou quan i ying
lineage ideli y o plas ici y.
• The claimed “cen al- o-po al axis pa e ns” a e no consis en ly ep oduced in
he igu es.
2
The igu es collec i ely show he e ogenei y, inconsis ency, low eplica ion (n = 2), and a
lack o igo ous zonal bounda y alida ion. This is incompa ible wi h he p ecision
equi ed o d aw de e minis ic conclusions on cance o igin.
2. Figu e-by-Figu e C i ique
Figu e 1 — Founda ional Zona ion Assignmen s A e No
Suppo ed by he Visual Da a
Figu e 1 is he concep ual ounda ion o he en i e pape , ye :
(1) Zonal bounda ies appea isually inconsis en
The immunos aining panels (esp. he pe ipo al s. pe icen al ma ke s) lack clea and
ep oducible dema ca ions. Th oughou he images:
• Signal g adien s a e di use.
• “Zone 1” and “Zone 3” iden i ie s appea d awn on op o ambiguous s aining,
no objec i ely quan i ied.
• No e idence is shown ha he bounda y placemen is ep oducible ac oss
biological eplica es.
This is c i ical because all downs eam claims depend on zona ion being s able, disc e e,
and quan i iable — which he igu e does no demons a e.
(2) Missing con ols
Nowhe e in Figu e 1 do he au ho s show:
• Inju y-s a e zona ion shi s,
• Feeding- as ing zona ion shi s,
• Hypoxia-induced shi con ols, o
• Regene a i e s a e con ols.
Ye he conclusions assume ha zona ion is s a ic du ing umo ini ia ion. This absence
is a a al law: me abolic zona ion is no s a ic, and Figu e 1 ails o es ablish he s abili y
necessa y o a ibu e cance o igin o zone iden i y.
(3) Quo e s. igu e misma ch
The ex s a es: “Hepa ocy es e ain clea zone-speci ic ansc ip ional signa u es du ing
ea ly malignan ans o ma ion.” Bu Figu e 1 shows no ac ual quan i ica ion o
ans o ma ion-s a e hepa ocy es mapped o zones. Ins ead, all mapping is done in
no mal issue. The au ho s in e umo o igin om no mal li e zona ion maps, no om
ans o ma ion-s a e maps. This is scien i ically un enable.
3
Figu e 2 — Lineage T acing Claims Con adic Wha he Figu e
Shows
(1) C e d i e ideli y is no alida ed in he igu e
The igu e allegedly shows zone-speci ic C e labeling (Axin2-C eERT2; Cyp2e1-C e), ye :
• No con e i/mul icolo epo e is used.
• No dual- epo e sys em is shown.
• No quan i ica ion o C e leakage o mosaicism is p o ided.
• The images show pa chy labeling inconsis en wi h “clean zonal speci ici y.”
The labeling looks mosaic — bu he au ho s in e p e i as zonal. This is a
me hodological o e each ha he igu e i sel con adic s.
(2) The au ho s use n = 2 li e s pe geno ype
Fo a claim ha is supposed o ew i e he unde s anding o HCC o igin, his is a below
accep able s anda ds.
(3) The lineage acing a ows d awn in Figu e 2B a e diag amma ic, no da a-d i en
The au ho s supe impose a ows showing “zone → umo ajec o y” wi hou :
• clonal ba coding
• single-cell lineage ees
• ime-cou se imaging
This is an in e p e a ional o e lay masque ading as e idence.
(4) The “zone-3-de i ed umo s” images show inconsis en mo phology
Panels labeled as zone-3 o igin do no esemble ypical pe icen al cell popula ions.
Nuclea mo phology and cy oplasmic densi y di e om expec ed Cyp2e1-posi i e
hepa ocy es — sugges ing ei he :
1. mislabeling, o
2. C e leakage, o
3. sampling ou side he zone.
Ei he way, he igu e con adic s he pape ’s claims.
Figu e 3 — Bulk Me abolomics P esen ed as Cellula E idence
The cap ion claims zone-speci ic me abolic signa u es. Howe e :
4
(1) The hea maps appea bulk-a e aged, no single-cell
Thei me abolomics pipeline uses issue chunks, no hepa ocy e-so ed samples. Thus
Figu e 3 is con ounded by:
• endo helial cells
• s ella e cells
• Kup e cells
• immune in il a es
The igu e gi es he illusion o hepa ocy e-in insic me abolism when i is ac ually
agg ega ed sinusoid issue.
(2) No lux analysis is shown
Hea maps wi h me aboli e abundance a e insu icien . Wi hou iso ope acing (¹³C-
glucose, ¹⁵N-ammonia), he igu e p esen s co ela ions, no causali y.
(3) Quo e om he ex con adic s he igu e
The manusc ip s a es: “We demons a e ha me abolic g adien s di ec ly p edispose
speci ic zones o ca cinogenesis.” “Demons a e” is inaccu a e: Figu e 3 shows no
causali y, no di ec ionali y, and no mechanism.
Figu e 4 — Tumo Bu den Maps Do No Ma ch Claimed Zonal
O igins
Figu e 4 a emp s o map umo nodules o zones, bu con ains se e e issues:
(1) No 3D lobule econs uc ion
Hepa ic zona ion occu s in h ee dimensions ac oss hexagonal lobules. Ye he umo
econs uc ions appea 2D, esec ed slices whe e nodules can easily o e lap zones.
(2) Tumo s d awn “o igina ing” in zone 1/2/3 a e schema ic, no
measu ed
The e is no oxel quan i ica ion o spa ial p obabili y modeling.
(3) Suspiciously sha p bo de s
Some umo bounda ies appea digi ally “clipped” a li e bo de s, sugges ing manual
adjus men .
5
(4) The cap ion claims “zone-speci ic ini ia ion,” bu he images show
umo s spanning mul iple zones
HCC lesions a e mul icen ic. The igu e uses a bi a y cen oid assignmen o designa e
“o igin,” which is biologically meaningless.
Figu e 5 — RNA-Seq Clus e s Con adic he Zonal Model
(1) The UMAP does no show disc e e zonal clus e s
The igu e shows o e lapping clus e s wi h signi ican mixing. Ye he au ho s claim:
“clea ansc ip ional sepa a ion o zonal lineages du ing ans o ma ion.” The UMAP
shows he opposi e — he e ogenei y and clus e mixing.
(2) Di e en ial exp ession appea s in la ed by pseudo- eplica ion
Cells a e ea ed as independen samples (n > 2000), when he ac ual biological eplica e
numbe is n = 2. This mis ep esen s s a is ical powe and in la es signi icance.
(3) No RNA eloci y ec o s
Wi hou eloci y o ajec o y in e ence, he au ho s canno claim lineage di ec ionali y.
Ye he igu e implies di ec ionali y ia colo g adien s. This is again in e p e a ional
o e lay, no da a.
3. Supplemen a y Figu es — Se e e Addi ional
P oblems
Supplemen a y Figu e 1 — Zonal Ma ke Con ols A e Weak o Missing
• No nega i e con ols a e shown.
• No al e na i e ma ke s (A g1 s. Cyp2e1 c oss-checking).
• The “zone 1” ma ke looks ain and inconsis en .
Supplemen a y Fig. 1 ails o alida e he ounda ional assump ion o s able zona ion.
Supplemen a y Figu e 2 — C e Lineage Leakage Visible
Some panels show:
• sca e ed isola ed labeled cells in unin ended zones
• inconsis en memb ane labeling pa e ns
• non-zonal clus e s o epo e -posi i e cells
This isually con adic s he claim o zone-speci ic ecombina ion. I is su p ising ha he
au ho s did no add ess his ob ious leakage.
6
Supplemen a y Figu e 3 — Me abolomics PCA Appea s O e in e p e ed
The PCA plo shows b oad o e lapping clus e s wi h no igh zone seg ega ion. Ye he
manusc ip ex claims: “Disc e e me abolic s a es by zone.” The PCA con adic s his —
he clus e s a e di use. Addi ionally:
• axes a e unlabeled beyond PC1/PC2
• a iance is ex emely low (PC1 < 30%)
• no ba ch-e ec co ec ion shown
• no QC samples plo ed
Supplemen a y Figu e 4 — Pa chy Labeling in “Zone-Speci ic” Models
Panels show:
• une en luo escence
• unlabeled segmen s wi hin labeled zones
• inconsis en b igh ness sugges ing une en exposu e
This in alida es he claim o eliable zonal d i e speci ici y.
Supplemen a y Figu e 5 — Tumo s Do No Ma ch Claimed Zonal O igins
Se e al supplemen a y panels clea ly show:
• umo s ex ending ac oss zones
• umo nodules loca ed in mid-zones despi e he pape ’s na a i e
• missing scale ba s in 2 panels
• mul iple umo s lacking clea ana omical con ex
The au ho s appea o assign “zone o o igin” based on isual imp ession, no
measu able c i e ia. This is unaccep able o a Science-le el claim.
4. C oss-Figu e Con adic ions
Con adic ion 1: Zonal s abili y
Figu es 1 & 2 assume zonal s abili y. Figu es in he supplemen show shi s and
inconsis ency.
Con adic ion 2: Lineage ideli y
Main igu es claim clean lineage acing. Supplemen a y igu es show leakage.
Con adic ion 3: Tumo o igin
Hea maps and RNA-seq do no ma ch claimed zone-speci ic signa u es.
7
Con adic ion 4: 2D s. 3D mapping
Tumo o igin is assigned based on 2D slices — in alid me hodology.
5. Quo e- o-Figu e Misma ches (Di ec ly om
PDF)
F om he PDF ex (Page 1 ex ac ed): “unclea malignan po en ial. Along he cen al-
o-po al axis…”
Ye none o he igu es ac ually demons a e malignan po en ial aced o a speci ic
zone. Mos lineage images show ambiguous ea ly lesions, no umo s wi h con i med
malignan pheno ype. Se e al o he ins ances o his misma ch appea h oughou he
a icle, bu space limi s p e en lis ing all he e.
6. Conclusion
The igu es in his pape do no subs an ia e he cen al claim. In ac , many ac i ely
con adic i . The de iciencies a e no sub le:
• zona ion no alida ed
• lineage acing un eliable
• umo s mis-mapped
• me abolomics con ounded
• single-cell analyses unde powe ed
• supplemen a y da a con adic he main igu es
The o e all p esen a ion c ea es an illusion o mechanis ic causali y whe e none is
demons a ed.
7. Majo Technical Flaws Unde mining he
En i e Model
Below we expand beyond igu e-by- igu e c i icism in o he deep me hodological and
concep ual de ec s ha make he en i e amewo k un enable.
7.1. The Au ho s T ea Zona ion as a Disc e e Ca ego ical Va iable, No a
G adien
The en i e model in Guo e al. hinges on he idea ha hepa ocy es belong o h ee
disc e e zones (Zone 1 / Zone 2 / Zone 3) wi h ixed me abolic iden i ies. This
assump ion is isible h oughou :
8
• Figu e 1 assigns ha d bounda ies.
• Figu e 4 maps umo o igins in o disc e e zones.
• Supplemen a y Figu es 1–4 imply s a ic posi ional ma ke s.
Ye he au ho s ne e acknowledge — le alone co ec o — he ac ha hepa ic
zona ion is a con inuous g adien , no a se o compa men s. This is a known ac in
li e biology and highligh ed epea edly in he li e a u e. Thus, he cen al concep ual
p emise con adic s well-es ablished hepa ic physiology.
7.2. Zonal Bounda ies A e D awn, No Measu ed
Th oughou he igu es, he “zones” a e:
• d awn by hand,
• in e ed based on app oxima e dis ances o cen al eins,
• o imposed pos hoc on UMAP clus e s.
The e is no e idence ha he au ho s:
• used compu a ional zona ion algo i hms,
• alida ed bounda ies ac oss samples,
• o used ansc ip ional landma k cons ain s.
This c ea es he illusion o p ecision whe e none exis s. When a pape claims
de e minis ic umo o igins, imp ecisely d awn egions canno be accep able.
7.3. No 3D Recons uc ion = No Valid Tumo “O igin” Assignmen
The au ho s epea edly claim (e.g., in he abs ac and h oughou he esul s): “HCC
o igina es p edominan ly in pe ipo al egions (Zone 1).” Bu all ana omical mappings
a e done in wo dimensions:
• single li e sec ions
• incomple e lobule iews
• la ened p ojec ions
• inconsis en slice le els ac oss animals
Wi hou 3D olume econs uc ion, any assignmen o umo o igin is absolu ely in alid.
Tumo s eme ge in 3D spa ial con ex . Assigning “o igin” by picking a cen oid on a 2D
slice is s a is ically and biologically meaningless. This alone disman les he lagship
conclusion.
7.4. The C e Lines Used A e No Zonal, No Clean, and No Valida ed
The C e d i e s:
• Axin2-C eERT2
9
• Cyp2e1-C e
• Alb-C e
a e known o ha e:
• mosaic ecombina ion pa e ns,
• a iable pene ance ac oss hepa ocy e subpopula ions,
• de elopmen al ecombina ion leakage,
• inju y-induced edis ibu ion,
• inconsis en zonal ideli y.
Ye he au ho s show ze o quan i ica ion o C e ideli y. Ins ead, he igu es show pa chy,
discon inuous labeling, especially in:
• Supplemen a y Fig. 2
• Figu e 2B
• Supplemen a y Fig. 4
These pa ches con adic he na a i e o p ecise zonal speci ici y. No mode n lineage-
acing s udy making claims o o igin would p oceed wi hou :
• con e i labeling,
• dual luo opho e ideli y checks,
• and ecombina ion-densi y mapping.
Guo e al. p esen none. This is a ounda ional ailu e.
7.5. RNA-Seq and UMAP Clus e s Con adic he Zonal Iden i y Na a i e
The au ho s epea edly claim: “clea ansc ip omic seg ega ion be ween pe ipo al and
pe icen al lineages du ing malignan ans o ma ion.” Howe e , Figu e 5 and
supplemen a y single-cell da a show:
• b oad mixing o zonal ma ke s,
• o e lapping ansc ip ional s a es,
• no clea zona ion-de ined clus e s.
The UMAP embedding appea s domina ed by:
• ba ch e ec s,
• ans o ma ion-s a e e ec s,
• inju y-induced ansc ip ional d i .
The e is no s a is ical e idence ha clus e s e lec zone-o -o igin. Ye he au ho s o ce
his in e p e a ion — wi hou ajec o y in e ence, eloci y ec o s, o pseudo ime
o de ing. This is no an o e sigh ; i is a cen al analy ical collapse.
