For each major or critical comment surfaced by the pre-submission reviewer report (5.17), here is a draft response paragraph the author can adapt for the rebuttal letter when the journal returns a major-revision decision. Each response is grounded in the same verbatim manuscript evidence the reviewer cited.
Response to Comment 0: 'Orchestration' claim lacks progenitor-specific functional evidence
Severity: critical | Stance: partially agrees revises scope
Reviewer Comment:
Author Response:
We thank the reviewer for raising this important concern about the causal interpretation of the progenitor-like compartment. We agree that the MRTX1133 and p53 perturbations do not isolate progenitor-like cells from other KRAS-dependent epithelial states, and therefore they cannot by themselves establish progenitor-specific sufficiency or necessity for niche assembly. A lineage-restricted ablation or transplantation experiment would be a valuable next step, however available Hmga2- or Nes-based strategies are not fully specific to this state in pancreas and would introduce injury or developmental confounders that are difficult to resolve within the present revision. We will therefore revise the manuscript to distinguish clearly between epithelial state association, KRAS-dependent niche remodeling, and progenitor-specific causality. We will replace broad uses of "orchestrate" and "architect" with more precise language such as "progenitor-associated niche organization" unless the sentence refers specifically to the inferred spatial association. We will also add a quantitative covariate analysis using the existing spatial data to test whether progenitor-like abundance remains associated with Tnc+ myCAF and Itgax+/Arg1+ macrophage enrichment after accounting for lesion area, treatment group, and mouse-level effects. This change will make clear what the current data demonstrate and what remains to be tested by progenitor-selective perturbation.
Manuscript Change:
We will revise the Abstract, Results section "Progenitor-associated niches emerge during early tumorigenesis," Discussion, and model schematic to remove causal overstatement [revised Fig. 7f]. We will add the covariate analysis in [new Fig. S12] and describe it in [Methods, new subsection "Spatial covariate modeling of niche composition"].
Response to Comment 1: Underpowered and unbalanced spatial transcriptomics cohorts for the central claims
Severity: critical | Stance: partially agrees revises scope
Reviewer Comment:
Author Response:
We appreciate the reviewer’s focus on animal-level replication and statistical calibration, which are central for interpreting spatial omics experiments. We agree that cell number cannot substitute for mouse-level replication, and the revised manuscript will state explicitly that the animal, not the cell, is the primary unit of biological replication. Rather than expanding cohorts in an unplanned manner, which would not address calibration of the existing spatial test, we will add effect sizes with 95% confidence intervals for all principal comparisons, including progenitor-like abundance, niche composition, and spatial-Milo differential abundance estimates. We will also add post hoc detectable-effect calculations for each cohort and permutation-based simulations at the actual sample sizes and group balance used in the study to evaluate empirical FDR control. To address sensitivity to individual animals, we will report leave-one-mouse-out analyses for the MRTX1133, p53 knockdown, and injury time-course comparisons. These additions will allow readers to assess robustness, effect magnitude, and statistical uncertainty without implying that millions of cells provide independent biological replicates.
Manuscript Change:
We will add effect-size and confidence-interval reporting to the main figure legends and [new Table S6]. We will add the power, permutation, and leave-one-out analyses in [new Fig. S13] and describe them in [Methods, new subsection "Power, effect-size, and robustness analyses"].
Response to Comment 2: No data deposition statement and no public release of Calligraphy or spatial-Milo code
Severity: critical | Stance: agrees and commits
Reviewer Comment:
Author Response:
We agree fully with the reviewer that public data and code availability are essential for a study of this scale. The absence of accession numbers and repository information reflects an omission in the presubmission version rather than an intention to withhold these resources. We will deposit raw and processed scRNA-seq and Xenium data, including count matrices, cell metadata, segmentation outputs, and processed analysis objects, in an appropriate public repository such as GEO or ArrayExpress before resubmission. We will release the Calligraphy analysis code and the spatial-Milo implementation as version-tagged repositories with Zenodo DOIs, pinned software environments, and scripts corresponding to the analyses in the manuscript. We will also provide the full 480-gene Xenium panel, including probe identifiers and selection annotations, as a supplementary table. Finally, we will include a minimal reproducible example that regenerates one representative niche-composition or spatial-Milo panel from deposited count data. These changes will make the empirical and computational basis of the study independently inspectable and reproducible.
Manuscript Change:
We will add a complete Data Availability and Code Availability section with accession numbers, repository links, software versions, and Zenodo DOIs. The full Xenium panel will be provided in [Table S4], and the worked example will be referenced in [Methods, new subsection "Reproducible analysis resources"].
Response to Comment 3: Chimeric ES-cell KPLOH model may not represent sporadic somatic TP53 LOH
Severity: critical | Stance: partially agrees revises scope
Reviewer Comment:
Author Response:
We thank the reviewer for highlighting the interpretive boundary of the KPLOH system. We agree that this model is not equivalent to sporadic adult-onset TP53 LOH arising from a single pancreatic epithelial cell, and we should not imply that the observed frequency of p53-LOH cells directly estimates the frequency of such events in human initiation. At the same time, the model was designed to resolve rare p53-status transitions within a KRAS-mutant epithelial field, and that feature remains useful for studying state associations at early stages. A full adult tamoxifen-inducible somatic validation would require a separate breeding and induction strategy and is beyond the scope of the present revision. We will instead add an explicit limitations paragraph comparing the chimeric ES-cell platform with adult somatic induction models and with sporadic human PDAC initiation. Where available, we will add a targeted comparison to published adult inducible Kras/Trp53 and human PanIN or early PDAC datasets to assess whether the progenitor-like and p53-target signatures are observed outside the chimeric context. This revision will preserve the value of the KPLOH system while making its generalizability and limitations transparent.
Manuscript Change:
We will add the model-scope discussion to [Discussion, new subsection "Model context and limits of generalization"]. Any cross-dataset signature comparison will be included in [new Fig. S14] and described in [Methods, new subsection "External dataset comparison"].
Response to Comment 4: 'Senescence-like' designation rests on transcriptional signatures alone
Severity: major | Stance: agrees and commits
Reviewer Comment:
Author Response:
We appreciate the reviewer’s point that transcriptional similarity to OIS is not equivalent to a functional senescence state. We agree that the current manuscript should not conflate p53/CDKN2A/SASP-like transcriptional programs with stable proliferative arrest. Because this distinction cannot be resolved by text revision and citations alone, we will add a bounded validation using existing model conditions: HMGA2 co-staining with proliferation and arrest markers, including EdU or Ki67 and p21 or p16, on pancreatic sections from KC and KPLOH samples. Where suitable fresh or frozen tissue is available, we will also perform SA-β-gal staining and relate the signal to HMGA2-enriched epithelial regions. If progenitor-like cells show continued cycling, we will reframe the state as "senescence-associated" or "SASP-like without stable arrest" rather than OIS. We will also revise the Discussion to distinguish stable OIS, partial senescence programs, and stress-associated secretory phenotypes in KRAS-mutant epithelium. These additions will clarify whether the progenitor-like state reflects arrest, slow cycling, or a proliferative stress state with senescence-associated features.
Manuscript Change:
We will add the marker co-staining and, where feasible, SA-β-gal analysis in [new Fig. 5g and Fig. S15]. We will revise the Results section describing OIS signatures and the Discussion section on senescence to use "senescence-associated" terminology unless functional arrest is supported.
Response to Comment 5: CNA inference from scRNA-seq used as a genomic-instability timestamp without orthogonal validation
Severity: major | Stance: respectfully disagrees
Reviewer Comment:
Author Response:
We acknowledge the reviewer’s concern about using scRNA-seq-based CNA inference in stress- and EMT-like epithelial states. We respectfully agree that such calls should not be treated as direct genomic measurements, but we do not intend the CNA analysis to serve as the sole foundation for the progenitor-like state or niche conclusions, which are supported independently by transcriptional, spatial, perturbational, and histological analyses. We will revise the manuscript to present inferred CNAs as supportive evidence for possible genomic instability rather than as a timestamp that defines lineage progression. We will remove or soften chromosome-specific language unless it is directly supported by orthogonal data in the relevant cells. To bound false-positive risk without adding a separate rare-cell genomics campaign, we will benchmark the CNA inference against expected diploid stromal and immune compartments and against stress-high populations where copy-number neutrality is expected. We will also compare broad-arm patterns to published scDNA-seq from the same model only as external context, not as validation of individual cells profiled here. These revisions will prevent overinterpretation while retaining the CNA analysis as a hypothesis-generating layer.
Manuscript Change:
We will revise the Results section on CNA inference and the corresponding figure legend to state explicitly that these are RNA-derived inferences [revised Fig. 3 and Fig. S7]. Benchmarking and caveats will be added in [new Fig. S16] and [Discussion, paragraph "Limits of RNA-based CNA inference"].
Response to Comment 6: p53/CDKN2A/SMAD4 'convergence' framing ignores known pathway epistasis
Severity: major | Stance: partially agrees revises scope
Reviewer Comment:
Author Response:
We thank the reviewer for identifying an important distinction between co-engagement and pathway convergence. We agree that the present data do not establish epistasis among p53, CDKN2A, and SMAD4 programs, nor do they distinguish linear ARF-to-p53 or TGFβ-to-CDKN2A relationships from parallel tumor-suppressive inputs. We will therefore revise the language from "convergence" of independent barriers to "co-engagement" of tumor-suppressive programs within the progenitor-like state. We will also add a discussion of known pathway architecture, including p19ARF-mediated p53 activation and TGFβ/SMAD4 regulation of CDKN2A, to clarify that co-induction may arise from oncogenic stress rather than independent pathway activation. Additional genetic epistasis experiments with p19ARF-null or Smad4-conditional alleles would be informative, but they would constitute a new mechanistic study beyond the current scope. This reframing will align the interpretation with established PDAC tumor-suppressor biology while preserving the observation that these programs are enriched in the same early epithelial state.
Manuscript Change:
We will replace "convergence" language with "co-engagement" in the Abstract, Results, Discussion, and relevant figure titles [revised Fig. 4]. We will add pathway-hierarchy caveats and citations in [Discussion, new paragraph "Tumor-suppressor pathway architecture"].
Response to Comment 7: Custom 480-gene Xenium panel: selection criteria and validation not provided
Severity: major | Stance: agrees and commits
Reviewer Comment:
Author Response:
We agree with the reviewer that the custom Xenium panel is a central analytic filter and must be fully transparent. We will provide the complete gene and probe list, the rationale for each gene category, and the inclusion and exclusion criteria used during panel design. We will also describe how the panel was benchmarked against the scRNA-seq reference, including coverage of epithelial states, fibroblast subsets, macrophage and immune populations, endothelial cells, and key signaling modules. To assess whether the gastric-to-progenitor axis and niche composition are artifacts of panel ascertainment, we will add sensitivity analyses using randomly down-sampled panel gene sets and using the matched 480-gene subset within the whole-transcriptome scRNA-seq data. We will report which cell states or ligand-receptor modules are well represented and which are not, rather than implying uniform sensitivity across all compartments. These additions will allow readers to evaluate whether the observed progenitor-associated niche is robust to panel composition.
Manuscript Change:
We will expand [Table S4] to include gene symbols, probe identifiers, design category, and selection rationale. Panel design and sensitivity analyses will be added to [Methods, new subsection "Xenium panel design and validation"] and [new Fig. S17].
Response to Comment 8: Spatial adaptation of Milo not described in sufficient detail to reproduce
Severity: major | Stance: agrees and commits
Reviewer Comment:
Author Response:
We agree that the spatial-Milo adaptation requires a much fuller methodological description. We will revise the Methods to specify whether neighborhoods are defined in transcriptomic, spatial, or joint space; the neighborhood size or radius; the graph construction parameters; the abundance model; the test statistic; and the multiple-testing correction procedure. We will also state explicitly how mouse, section, and treatment structure are handled so that cells are not treated as independent biological replicates. To evaluate robustness, we will add sensitivity analyses across neighborhood sizes and report whether the principal Tnc+ myCAF and Itgax+/Arg1+ macrophage depletion calls are stable. The implementation will be released with the public code repository and a worked example using deposited data. These changes will make the analysis reproducible and will allow readers to judge whether the reported niche-depletion results are robust to spatial modeling choices.
Manuscript Change:
We will add a detailed [Methods, new subsection "Spatial differential abundance analysis"] and include parameter tables in [Table S7]. Sensitivity analyses and the worked example will be referenced in [new Fig. S18] and the Code Availability section.
Response to Comment 9: Blinding for histological scoring and cell-state annotation not reported
Severity: major | Stance: agrees and commits
Reviewer Comment:
Author Response:
We thank the reviewer for raising this important reporting and rigor issue. We will explicitly state the blinding status for each quantitative histological, immunofluorescence, and annotation step in the Methods and reporting checklist. For endpoints that were not originally scored under blinded conditions, we will re-score the key histological endpoints, including HMGA2+ cell counts, lesion morphology, and p-ERK quantification, using two independent observers blinded to genotype and treatment. We will report inter-rater agreement using Cohen’s kappa for categorical lesion calls and intraclass correlation coefficients for continuous measurements. For cell-state annotation, we will clarify which steps were unsupervised, which marker panels were used for label assignment, and whether treatment or genotype labels were hidden during annotation. This addition will document the safeguards against observer bias and provide a quantitative measure of scoring reproducibility.
Manuscript Change:
We will add blinding and inter-rater agreement details to [Methods, new subsection "Blinding and histological scoring"] and [Nature Reporting Summary]. Re-scoring results will be reported in [new Table S8] and referenced in the relevant figure legends.
Response to Comment 10: Spatial niche-pseudotime framework not benchmarked against existing spatial-trajectory methods
Severity: major | Stance: respectfully disagrees
Reviewer Comment:
Author Response:
We appreciate the reviewer’s point that the methodological positioning should be better calibrated against the existing spatial-analysis literature. We respectfully view the pseudotime analysis in this manuscript as a domain-specific organizing strategy for linking epithelial state transitions to local niche composition, not as a comprehensive new spatial-trajectory method. Accordingly, a full benchmark against Banksy, NCEM, niche-DE, MISTy, SpatialDM, and related tools would be disproportionate to the biological scope of this revision. We will remove or soften statements describing the approach as a broadly generalizable framework unless they are explicitly framed as a conceptual extension rather than a methods advance. We will add citations to relevant spatial niche and trajectory methods and explain the specific conceptual delta: anchoring stromal and immune neighborhood changes to an epithelial differentiation axis in early pancreatic tumorigenesis. If space permits, we will include a limited qualitative comparison of the assumptions and outputs of these methods rather than a quantitative performance benchmark. This revision will prevent overclaiming novelty while preserving the biological interpretation of niche-state ordering.
Manuscript Change:
We will revise the Results and Discussion language around niche pseudotime and add a short positioning paragraph with citations to existing methods [Discussion, new paragraph "Relation to spatial trajectory and niche-inference methods"]. The phrase "generalizable framework" will be replaced or qualified in the relevant Results section and figure legend [revised Fig. 6].
Notes for the author
- Only one response commits to new wet-lab validation, the senescence-marker analysis. The remaining commitments are text, reporting, computational robustness, data deposition, code release, or blinded re-scoring.
- Before submission, replace placeholder figure, table, and section labels with final manuscript numbering and ensure that all committed analyses are feasible with the available data.
- For Issue 4, if EdU or suitable SA-β-gal material is not available, revise that response to commit only to Ki67 or pHH3 and p21 or p16 co-staining on existing sections, and strengthen the language that the manuscript will use "senescence-associated" rather than functional OIS.