Single-Cell Cross-Species Profiling identifies Conserved Transcriptional Networks in Early Pancreatic Tumourigenesis
This article has been Reviewed by the following groups
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
- Evaluated articles (Review Commons)
Abstract
Pancreatic ductal adenocarcinoma (PDAC) is the most common form of pancreatic cancer and carries the poorest prognosis among all cancers, largely because it is frequently diagnosed at metastatic stages. It is therefore critical to identify reliable markers of preinvasive stages and to decipher the network driving preinvasive lesions to invasive carcinoma.
Here, we generated a zebrafish model in which KRAS G12D is specifically expressed in pancreatic acinar cells, inducing acinar-to-ductal metaplasia that faithfully mirrors mammalian tumorigenesis. Single cell RNA-seq allowed us to capture transcriptional changes occurring at early stages of the disease. Cross-species comparison with mouse and human scRNAseq transcriptomes revealed a striking conservation of the genes upregulated during metaplasia, triggering common signalling pathways and regulatory programs. Notably, metaplastic cells reactivate a broad set of developmental genes expressed in multipotent pancreatic progenitors.
Mapping the acinar-to-cancer trajectories revealed a set of cytoskeletal and migration-related genes specifically upregulated during the late phase of metaplasia, immediately prior to malignant transformation, likely conferring invasive potential to these cells. SCENIC analysis further identified regulatory networks that become progressively activated as cells transition toward cancer, suggesting their involvement in the acquisition of malignant traits.
In conclusion, our cross-species comparison demonstrates a high degree of conservation in the molecular mechanisms driving pancreatic cancer progression from early to late stages across evolutionarily distant species, including zebrafish, mouse, and human, highlighting critical pathways that should be targeted to prevent cancer progression.
To allow researchers to easily explore gene expression profiles during pancreatic cancer progression across all three species, the datasets are publicly accessible via a user-friendly web platform ( https://www.zddm.page.gd/ )
Article activity feed
-
Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.
Learn more at Review Commons
Reply to the reviewers
The reponse to the reviewer comments are provided as a pdf file
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #3
Evidence, reproducibility and clarity
Summary:
Pancreatic ductal adenocarcinoma (PDAC) has poor prognosis because it is usually diagnosed when already metastasized. Therefore it is important to identify early markers of disease. In this work, the authors developed a zebrafish model of PDAC which shows similarity to human tumor formation. Single cell RNA sequencing was used to identify transcriptional changes associated with early disease and acinar-to-ductal metaplasia, which precedes cancer formation. Cross-species comparison revealed transcriptional changes that were highly conserved between zebrafish and mouse cancer models, and human disease. These include …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #3
Evidence, reproducibility and clarity
Summary:
Pancreatic ductal adenocarcinoma (PDAC) has poor prognosis because it is usually diagnosed when already metastasized. Therefore it is important to identify early markers of disease. In this work, the authors developed a zebrafish model of PDAC which shows similarity to human tumor formation. Single cell RNA sequencing was used to identify transcriptional changes associated with early disease and acinar-to-ductal metaplasia, which precedes cancer formation. Cross-species comparison revealed transcriptional changes that were highly conserved between zebrafish and mouse cancer models, and human disease. These include reactivation of developmental genes and increased expression of genes related to cytoskeleton and cell migration.
Major comments:
The authors' conclusions are well supported through comparison of newly generated zebrafish datasets with available published data from a mouse model and human patient samples. Providing an accessible website for further data exploration will be a valuable service to the research community (this website is not yet available).
Optional additional experiments:
The claims of this work would be strengthened by experiments to detect the identified upregulated pathways in the zebrafish cancer model at the tissue level. This could be performed by RNA in situ hybridization or antibody staining, if reagents for zebrafish are available. This would substantiate the likely biological relevance of changes detected in the single cell datasets. Correlation of marker expression with regions of varying differentiation or tumor progression may implicate mechanistic importance. qPCR would be an alternative method to confirm gene expression changes, but this does not provide the possibility for correlation of gene expression with histology.
There is extensive speculation in the results section relating to cellular acquisition of invasive potential, which is more suitable for the discussion. Focal adhesions, integrin-mediated signaling and actin-ECM coupling are implicated but not examined. Ultrastructure of cytoskeletal elements and focal adhesions could be examined in tissue samples by electron microscopy. These structures could also be studied with immunofluorescence labeling of individual components and phalloidin staining. Such experiments may show regional variations within tumor samples and reveal changes specific to regions of invasion.
The SCENIC analysis was performed only using mouse and human datasets. Factors that are part of conserved regulons relevant for tumor progression could be examined in the zebrafish model - at a minimum in the scRNAseq dataset, or better by using RNA in situ hybridization or antibody approaches.
Minor comments:
Methods - Differential expressed genes: the justification for use of Seurat v4 versus v5 is confusing. Figure 3 F-H - Alcian Blue staining is hard to interpret. It would be helpful to include a negative control (no staining) and/or a positive control (normal tissue), and close-up views. (Possibly the images provided for review are of low quality and therefore difficult to appreciate.) p. 16 To clarify the statement "Early metaplastic cells were excluded" it would be helpful to refer to Figure 5A p. 21 "Visible" WISH is an odd term, better to instead say 'chromogenic' WISH p. 25 Refers to Figure 11. This should be Figure 8?
Significance
The zebrafish cancer model used in this work, which combines acinar cell expression of KrasG12D with a tp53-mutant background, is superior to previously reported models which showed lower frequency of tumor formation. Here, the authors report tumor development detectable at 3 months of age which affects 100% of animals by one year.
Cross-species scRNAseq analysis comparing zebrafish to mouse and human tumor samples showed conserved activation of progenitor factors, and cytoskeletal and migration related genes. The involvement of such pathways in tumorigenesis is not novel. However, the demonstration of similarity of tumor-promoting pathways between zebrafish and rodent models, and human patient samples, is an important reinforcement of the value of the zebrafish model for preclinical investigations. Additional experiments as outlined above would substantiate functional relevance for identified pathways.
One issue warranting clarification is the onset of tumor formation in the reported model. Since KrasG12D expression is detectable starting at 5dpf, neoplastic changes may initiate already at larval and juvenile stages. A further issue is that KrasG12D expression decreases as cells shift away from the acinar fate, and this may limit the progression to advanced cancer stages. An inducible model, for example using a Cre/loxP system with the ela3l promoter driving Cre, could provide adult-onset tumor initiation and would maintain oncogene expression independent of changes in the cells' differentiation state. Using fish that are transparent as adults could allow detection of tumor growth before a visible mass is detectable externally. Subsequent decrease of GFP expression could indicate progression to more advanced tumor stages.
Authors should acknowledge that transcriptional changes may not directly reflect protein abundance and activity, and therefore functional biological relevance is only suggested.
This paper will be of interest to scientists interested in mechanisms of cancer formation and in the development of cancer therapies. The datasets that will be made accessible to the research community will be a valuable resource for researchers studying development and diseases of the pancreas.
My background includes medical training, research in developmental and cancer biology, and ongoing research using zebrafish for modeling disease processes.
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #2
Evidence, reproducibility and clarity
This manuscript is an interesting comparative biology resource. It describes the development of a novel model of pancreatic metaplasia and tumorigenesis in Danio rerio. Authors characterize histopathological features of pancreatic tumors. In a more molecular-oriented approach, they perform single-cell RNA sequencing of KRASG12D-expressing pancreata at different timepoints. Comparing both mouse and human datasets, they describe a conserved cross-species signaling program in PDAC carcinogenesis, which includes the reactivation of developmental genes.
The sequencing approach chosen suffers from several technical hindrances that limits, …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #2
Evidence, reproducibility and clarity
This manuscript is an interesting comparative biology resource. It describes the development of a novel model of pancreatic metaplasia and tumorigenesis in Danio rerio. Authors characterize histopathological features of pancreatic tumors. In a more molecular-oriented approach, they perform single-cell RNA sequencing of KRASG12D-expressing pancreata at different timepoints. Comparing both mouse and human datasets, they describe a conserved cross-species signaling program in PDAC carcinogenesis, which includes the reactivation of developmental genes.
The sequencing approach chosen suffers from several technical hindrances that limits, in the opinion of this reviewer, the quality of the results.
Following is listed a brief summary of the points that we would suggest in the revision of this work:
- The sequencing of the zebrafish samples involves heavy steps of pancreatic tissue enzymatic dissociation, and, as the authors also acknowledge, this could lead to major contamination of ambient RNA. The authors state the application of functions coming from the CellBender and DoubletFinder packages to improve the quality of the results. We realize that, while it would be beneficial, the employment of less aggressive dissociation techniques (if any) is unfeasible in this stage of the work, but it would be useful for the authors to comment and describe how much the samples were initially contaminated by ambient RNA and to provide parameters to show the improvements obtained with their pre-processing pathway.
- The choice to perform sample integration just on the zebrafish samples appears to insert an additional difference in the processing of datasets coming from different species. In our opinion, since this is a complete reanalysis of published datasets, the analysis pipeline employed should be the same for all the datasets.
- Figures 1K-L does not appear of sufficient quality to assess tissue morphology.
- Authors state that the reduction in GFP signal in staining of tumoral samples is probably due to a downregulation of ela3l promoter in tumoral cells; however, from the provided images, it is not clear if the assessed area is indeed cellularized, or if it may match to a fibrotic acellular region, consequently hampering the validity of the other immunostaining findings. Clearer pictures (probably including a nuclear/cellular staining should be provided to ensure the cellularity of the tissue portion.
- In multiple points of the manuscript, the authors refer to S and G2/M phase signatures, which are not described in the methods section. Additionally, if these signatures derive from the CellCycleScoring framework from the Seurat package, the authors should show the pattern of expression of at least some relevant genes, since it is debated if the small signatures used by Seurat might be sufficient to infer cell cycle phases.
- Authors emphasize a possible role of OLFM4 and ribosomal/translation proteins upregulation as putative mechanisms of metaplastic progression to overt tumors; however, these reports are purely speculative and based on reports performed in other cancer context, or on hypothesis on the possible functions of these factors in metaplastic cells. The authors should provide experimental validation of their claims on the role of these factors.
- The authors state that reactivation of a pancreatic developmental program is a central feature of metaplastic transition. However, this is based on developmental genes being an (albeit significant) minor fraction (~10%) of the total regulated genes. Moreover, many of the cited pathways are already widely known regulator of pancreatic metaplasia and early tumorigenesis.
- Authors state that transcriptional program activated at the terminal stages of metaplasia in zebrafish closely mirrors that observed in mammals. This however is based on sequencing of pre-neoplastic samples in zebrafish, and in an overlap of just the 25% of the DEGs in mouse and human samples.
Significance
The newly developed zebrafish model appears sound and well-characterized, and could definitely provide a useful platform to perform preliminary in vivo experiments for scientists in the field. It is really difficult to grasp the carcinogenic stage and authors might want to do a better job in characterizing the multi-step evolution of PDAC in their model. Terminology and histo-pathological accuracy should be improved. While the use of trajectory analysis and GRNs assessment with SCENIC is a very appreciated and elegant approach, the major novelty of this paper is related to the description of a correspondence between zebrafish models and other organisms in the molecular mechanisms subtending the early phases of pancreatic carcinogenesis, but the pathways described are almost all already known and linked to tumor development in this context. Yet, very informative from a comparative biology perspective.
The opinion of this reviewer is that this paper constitutes a very interesting study harnessing a novel zebrafish model which could be quite useful to researchers in the field, given its similarity to more advanced model organisms, but lacks major molecular findings describing the early phases of tumorigenesis in the pancreas. It should benefit from a stronger/more robust histological characterization - where possible.
My expertise: pancreatic carcinogenesis
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #1
Evidence, reproducibility and clarity
Major points
The manuscript is very descriptive, and very long on the molecular analyses. The work would gain impact from a clearer focus on what is learned about pancreatic tumour biology and metaplastic processes, rather than on exhaustive dataset description.
Tumour latency is very long (up to one year in a 2-5 year lifespan ), with relatively low penetrance and few tumours progressing to PDAC. These features raise concerns about the robustness and translational relevance of the model as a PDAC system.
All experiments involve male fish only, despite known sex differences in PDAC biology. Either inclusion of sex stratified data or …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #1
Evidence, reproducibility and clarity
Major points
The manuscript is very descriptive, and very long on the molecular analyses. The work would gain impact from a clearer focus on what is learned about pancreatic tumour biology and metaplastic processes, rather than on exhaustive dataset description.
Tumour latency is very long (up to one year in a 2-5 year lifespan ), with relatively low penetrance and few tumours progressing to PDAC. These features raise concerns about the robustness and translational relevance of the model as a PDAC system.
All experiments involve male fish only, despite known sex differences in PDAC biology. Either inclusion of sex stratified data or a clear rationale for restricting the study to males will be important for assessing generalisability.
The behaviour of non progressing tumours, and its relationship to oncogene status (such as possible Kras silencing), is mentioned but not mechanistically explored. A more explicit analysis or discussion of why some lesions fail to progress would add valuable biological insight.
Molecular profiling is based on a small number of tumours sampled at different time points, capturing inter tumour heterogeneity rather than true biological replication. As a result, statements about conserved pathways and transcriptional programmes are supported by limited replication and would benefit from more cautious wording.
Functional validation of key regulators or pathways is lacking. Even limited functional work-such as manipulation of one candidate gene, or derivation of a transplantable line to enable testing of tumour responses to standard of care treatment-would considerably enhance the biological and translational impact of the study; in the absence of such data, the limitations of purely descriptive findings warrant clearer acknowledgement.
Pathology assessment is incompletely documented. It is not clear whether ductal adenocarcinoma was actually recognised in the zebrafish lesions (differentiation grade was scored).
Minor comments
SCENIC is mentioned in the abstract without explanation. A brief indication of its role (for example, inference of transcription factor regulons associated with specific tumour states) would aid readers unfamiliar with the method.
Tumour assessment is described as occurring "as soon as they were visible," but the practical detection method, and surveillance intervals are not specified. A clearer description of how tumours were detected and monitored would make the methodology more transparent.
The Methods section, particularly the bioinformatics part, is disproportionately long and includes interpretative elements that read like Results. Relocating detailed computational pipelines and secondary explanations to the Supplementary Methods could improve flow and readability. The Discussion is lengthy (around seven pages) with extensive molecular detail. A more concise Discussion that emphasises main conclusions, limitations, and translational implications would likely improve readability without loss of content.
Wording about invasion and metastasis sometimes appears stronger than the data support, given that only a minority of tumours show metastatic behaviour. Statements about invasive potential would benefit from closer alignment with observed frequencies.
The number of figures, particularly supplemental ones, is high, and some the main figures appear to be included twice. Streamlining the figure set and removing redundancies would make the presentation more focused.
Significance
The manuscript is very descriptive, and very long on the molecular analyses. The work would gain impact from a clearer focus on what is learned about pancreatic tumour biology and metaplastic processes, rather than on exhaustive dataset description.
-
