Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Apr 1.
Published in final edited form as: J Surg Oncol. 2024 Jan 17;129(5):860–868. doi: 10.1002/jso.27582

Myofibroblastic cancer-associated fibroblast subtype heterogeneity in pancreatic cancer

Joseph F Kearney 1, Hannah E Trembath 1, Priscilla S Chan 2, Ashley B Morrison 2, Yi Xu 2, Chang Fei Luan 2, Ian C McCabe 2, Sandra A Zarmer 2, Hong Jin Kim 1, Xianlu L Peng 2, Jen Jen Yeh 1
PMCID: PMC11307498  NIHMSID: NIHMS1957174  PMID: 38233984

Abstract

Background:

Pancreatic ductal adenocarcinoma (PDAC) has a fibrotic stroma that has both tumor-promoting and tumor-restraining properties. Different types of cancer-associated fibroblasts (CAFs) have been described. Here, we investigated whether CAFs within the same subtype exhibit heterogeneous functions.

Methods:

We evaluated the gene and protein expression differences in two myofibroblastic CAF (myCAF) lines using single-cell and bulk RNA-sequencing. We utilized proliferation and migration assays to determine the effect of different CAF lines on a tumor cell line.

Results:

We found that myCAF lines express an activated stroma subtype gene signature, which is associated with a shorter survival in patients. Although both myCAF lines expressed α-smooth muscle actin (α-SMA), platelet-derived growth factor-α (PDGFR-α), fibroblast-activated protein (FAP), and vimentin, we observed heterogeneity between the two lines. Similarly, despite being consistent with myCAF gene expression overall, heterogeneity within specific genes was observed. We found that these differences extended to the functional level where the two myCAF lines had different effects on the same tumor cell line. The myCAF216 line, which had slightly increased inflammatory CAF-like gene expression and higher protein expression of α-SMA, PDGFR-α, and FAP was found to restrain migration of tumor cells.

Conclusions:

We found that two myCAF lines with globally similar expression characteristics had different effects on the same tumor cell line, one promoting and the other restraining migration. Our study highlights that there may be unappreciated heterogeneity within CAF subtypes. Further investigation and attention to specific genes or proteins that may drive this heterogeneity will be important.

Keywords: cancer-associated fibroblasts, pancreatic adenocarcinoma, tumor microenvironment

1 |. INTRODUCTION

Five-year survival for pancreatic ductal adenocarcinoma (PDAC) has doubled in the last decade, but the prognosis for patients remains poor. PDAC is predicted to become the second leading cause of cancer death in the United States by the end of the decade.1 The PDAC tumor microenvironment (TME) remains a challenge for effective treatment.2

The PDAC TME is composed of cancer-associated fibroblasts (CAFs), neurons, immune cells, extracellular matrix, and normal pancreatic tissue. While CAF origins are an area of active research, the general consensus is that most CAFs arise from peritumoral tissues.3 CAFs are believed to alter PDAC phenotype and tumor cell survival by providing growth factors and nutritional support to tumor cells3,4 and modulating the immune environment within the TME.3,5 CAFs have also been implicated in impeding gemcitabine delivery to PDAC tumors secondary to a drug-scavenging mechanism.6

Prior attempts to target the stroma using PEGPH20,7 a hyaluronidase inhibitor, and vismodegib,8 a hedgehog inhibitor, with chemotherapy in patients with metastatic PDAC9 were unsuccessful. Work in genetically engineered mouse models showed that depletion of the stroma accelerates tumor invasion and metastasis, and shortens survival.10 Since these studies, further work in the stroma field has revealed discrete CAF subtypes contributing to CAF heterogeneity. While discriminatory CAF markers for each subtype have not been identified and nomenclature is still in evolution, in general, myofibroblastic CAFs are thought to express α-smooth muscle actin (α-SMA), have contractile properties, and are believed to restrain PDAC tumor growth. Inflammatory CAFs (iCAFs) highly express platelet-derived growth factor-α (PDGFR-α) and secrete chemokines and interleukins (IL-1 and IL-6).11 Antigen-presenting CAFs express major histocompatibility complex-II and bind with CD4+ T cells to modulate their activity.5 There is a paucity of specific markers related to the subtypes, no consensus over which markers are associated with each subtype, and the expression profile of a CAF may have markers from more than one subtype.12

We hypothesize that CAFs of the same subtype may be heterogeneous and may have different effects on PDAC cells. CAF cell lines must be systematically characterized to fully understand and translate subtype-specific laboratory findings to the clinic.

2 |. METHODS

2.1 |. Patient samples

Deidentified PDAC tumors were collected from the University of North Carolina Lineberger Comprehensive Cancer Center Tissue Procurement Facility under IRB exemption.

2.2 |. RNA-sequencing (RNAseq)

RNA was extracted from fresh frozen tissue using the AllPrep DNA/RNA Mini Kit (Qiagen). Libraries were prepared using the TruSeq Stranded mRNA Library Prep (Illumina). The final complementary DNA (cDNA) libraries were pooled and diluted to 1.65 pM before being sequenced on a NextSeq 500 using the NextSeq 500/550 Mid Output Kit v2.5 (150 cycles) (Illumina). Using the bcl2fastq2 Conversion software 2.20.0 we converted BCL files to FASTQ files and then collapsed the lanes into one file. Total expected read counts were quantified using Salmon 1.9.013 using arguments “--gcBias -- seqBias.” The UCSC human reference genome GRCh38.p14 and RefSeq transcripts assembly (GCF_000001405.40) were used as references for the quantification of reads to derive the transcripts per million values.

2.3 |. Single-cell RNA-sequencing (scRNAseq)

Myofibroblastic CAF216 (myCAF216) cell count was determined using an automated cell counter (Bio-Rad) and viability was determined through Trypan blue (Gibco) staining. Ten thousand live cells were encapsulated for droplet-based single-cell RNAseq using Chromium Single Cell 3′ v3 reagents (10x Genomics), according to the manufacturer’s protocols. cDNA library quality was assessed with the 4150 Tapestation System and D5000 double-stranded DNA screen tapes (Agilent). Sequencing was performed on a NextSeq 500 (Illumina) using NextSeq 500/550 Output Kit v2.5 (150 cycles) (Illumina) at 200M reads.

Cell Ranger 6.1.2 was used for preprocessing of the raw files. Briefly, BCL image files were converted to fastq files using “cellranger mkfastq,” followed by unique molecular identifier quantification using “cellranger count” based on the human GRCh38 genome.14 SCISSORS (Sub-Cluster Identification through Semi-Supervised Optimization of Rare-cell Silhouettes),15 which is a wrapper of the Seurat package (v4), was employed for downstream analysis on the count matrix (filtered_feature_bc_matrix). The quality control steps include (1) the inclusion of genes expressed in more than two cells, (2) the inclusion of cells that have the number of genes captured within two times absolute deviation (+−2 SD) from the median (1500 < nFeatures < 4700), and (3) the inclusion of cells that have less than three times SD (+3 SD) from the median of the percent of mitochondrial reads (percent_MT < 10). The filtered data were processed by SCISSORS to derive robust clusters as described before.15 Resultant clusters were annotated using SingleR.16

2.4 |. Exemplar genes of CAF subtypes

SCISSORS was used to derive myCAF- and iCAF-like exemplar genes from the Elyada data set5 as previously described.15 panCAF exemplar genes were defined by combining myCAF-, iCAF-, and antigen-presenting CAF-like clusters and compared to perivascular fibroblast and endothelial cell clusters to derive a candidate gene list (p < 0.05, Wilcoxon’s rank-sum test). This gene list was filtered by excluding the genes that were highly expressed (top 10%) in other cell clusters. The top 10 SCISSORS myCAF, iCAF, and panCAF genes ranked by fold change were used for variance-adjusted Mahalanobis (VAM) score estimation in scRNAseq.17 In addition, the top 25 genes ranked by fold change for myCAF- and iCAF-like subtypes were used for gene expression estimation in bulk RNAseq.

2.5 |. Cell lines

The P411T1 patient-derived xenograft (PDX) PDAC cell line was previously established as described.18 Cells were stably infected using pLenti-CMV-mCherry-Neo (Addgene).

CAF lines were previously established using the outgrowth method as described.19 Briefly, deidentified primary PDAC tumors were mechanically dissociated and plated in Advanced Dulbecco’s modified Eagle’s medium/nutrient mixture F-12 (DMEM/F-12) (Gibco) media supplemented with 15% fetal bovine serum (FBS). Deposited fibroblast cells were isolated from epithelial cells via differential trypsinization and immortalized with pBABE-puro-hTERT (Addgene).

2.6 |. Western blot

Cells were grown to 80% confluence in a 150 mm plate in DMEM/F-12 with 10% FBS and lysed using the RIPA (25 mM Tris-HCl, 150 mM NaCl, 0.1% sodium dodecyl sulfate [SDS], 1% NP-40, 1% deoxycholate) lysis buffer with 1× protease and phosphatase inhibitors (Complete Protease Inhibitor Cocktail Sigma and Phosphatase Inhibitor Cocktails 2 and 3; Sigma) incubated on ice for 15 min, and centrifuged at 20 000 g for 10 min in a cold room. Thirty micrograms of protein was loaded onto a 16% SDS-polyacrylamide gel electrophoresis gradient gel. Proteins were transferred to a polyvinylidene fluoride membrane and probed with antibodies. All measurements were performed in the linear range without saturation and were normalized to β-actin as the loading control.

Antibodies used for Western blot:

  • α-SMA: Invitrogen, MA5-11547, lot #YE3836061, 1:500.

  • Vimentin: Abcam, ab16700, lot #GR204647, 1:500.

  • PDGFR-α: Cell Signaling, 3174S, lot #8, 1:500.

  • Fibroblast-activated protein (FAP): Cell Signaling, 66562S, lot #3, 1:500.

  • β-Actin: Santa Cruz, 4777B, lot #H122, 1:500.

  • E-cadherin: Cell Signaling, 3195, lot #15, 1:500.

2.7 |. CAF conditioned media

CAFs were plated at a density of two million cells in a 150 mm TC-treated dish in DMEM/F-12 with 10% FBS for 24 h. After 24 h, the media were removed, and the plates were washed with 5 mL of phosphate-buffered saline four times. Twenty milliliters of DMEM/F-12 (phenol-free; Gibco) was added to each plate and incubated for 48 h. The conditioned media (CM) were then harvested, passed through a 0.45-μm PES filter, aliquoted, and snap-frozen. On the day of each experiment, CM were thawed, and FBS was added.

2.8 |. Proliferation

P411T1 cells were plated in a 96-well TC-coated plate at a density of 1000 cells per well in DMEM/F-12 with 10% FBS. For each replicate, conditions were plated in quadruplicate, and each plate contained a dilution curve to assess the dynamic range. Plates were incubated overnight. Media were removed the next day, and the wells were washed with 100 μL of PBS for four washes. CM aliquots were thawed, and FBS was added to a final concentration of 1%. DMEM/F-12 with 1% FBS was used as the control media. Cells were incubated with the media at 37°C for 72 h. The Cell-Titer Glo 2.0 Cell Viability Assay (Promega) was performed to quantify tumor cell growth. Results from two biological replicates were used for analysis.

2.9 |. Migration

P411T1 cells were plated in a 24-well TC-coated plate at a density of 100 000 cells per well in DMEM/F-12 with 10% FBS. Cells were incubated overnight, and then a cruciate scratch was made using a sterile 1 mL pipette tip. The wells were then washed four times with PBS. CM aliquots were thawed, and FBS was added to a final concentration of 2.5%. DMEM/F-12 with 2.5% FBS was used for the control media. The wells were imaged at 4× using the Keyence fluorescent microscope with a Texas Red Filter (585 nm; Keyence BZ-X Series; #OP-87765) using the same XY coordinates at 0, 24, 48, and 72 h. Scratches were quantified using imageJ20 using the wound healing size tool macro.21 Results from two biological replicates were used for analysis.

2.10 |. Statistics

The statistical program Tidyverse22 in R was used for data analysis.23 Overall survival was analyzed using the Kaplan–Meier product-limit method. For proliferation analysis, a two-sided t-test was used to determine significance. For migration, a two-sided t-test was used to determine the significance of percent closure at 72 h.

3 |. RESULTS

3.1. |. Stroma differences in patients

We evaluated a previously defined stroma signature19 in 102 treatment-naïve PDAC tumors (Figure 1A). In agreement with findings from our previously published independent cohort of 145 PDAC patient samples with microarray data, we found that patients with an activated stroma signature had a hazard ratio of 2.20 with a 95% confidence interval of [1.01–4.82] (Figure 1B). We previously showed that CAFs are the contributory cell type to the activated stroma signature.19 To identify models that recapitulate the activated stroma signature, we derived two CAF lines from primary PDAC tumors (CAF216 and CAF227). Compared to a PDX-derived tumor cell line, P411T1, we found that both CAF216 and CAF227 had strong expression of the activated stroma signature (Figure 1C). Neither P411T1 nor the CAF lines expressed Moffitt’s normal stroma signature genes.19 In contrast, P411T1 was characterized by largely Moffitt basal-like signature genes (Figure 1D). Both CAF lines had an absence of basal-like and classical gene signatures.

FIGURE 1.

FIGURE 1

Open in a new tab

(A) Unsupervised hierarchical clustering of treatment-naïve resected primary pancreatic ductal adenocarcinoma (PDAC) samples using the Moffitt stroma signature. (B) Kaplan–Meier plot of patients with treatment-naïve resected primary PDAC stratified by the Moffitt stroma signature. (C) Moffitt stroma gene expression in tumor and cancer-associated fibroblast (CAF) lines. (D) Moffit tumor gene expression in tumor and CAF lines.

3.2 |. CAF marker expression

Rapid advances in CAF research have highlighted their heterogeneity with multiple subpopulations including myCAFs, which are largely but not completely characterized by α-SMA expression and other subpopulations including some by FAP expression.24

To characterize our CAF lines, we first performed scRNAseq of CAF216 and found that single R identifies all cells as activated stellate cells (Figure 2A).16 As there is no single panCAF marker, we used our previously described method SCISSORS15 to derive exemplar panCAF marker genes (Figure 2B,C). Briefly, SCISSORS is a computational method that identifies rare cell clusters and cell-type specific exemplar genes in scRNAseq.5,11 We confirm that all CAF216 cells express panCAF marker genes and express myCAF-like genes.

FIGURE 2.

FIGURE 2

Open in a new tab

(A) Single-cell RNA-sequencing of cancer-associated fibroblast 216 (CAF216) clustering (left) and annotation (right). (B) List of top 10 SCISSORS (Sub-Cluster Identification through Semi-Supervised Optimization of Rare-cell Silhouettes) CAF marker genes. (C) Uniform Manifold Approximation and Projection plots showing variance-adjusted Mahalanobis score of the SCISSORS CAF marker genes.

General mesenchymal markers used to characterize CAFs include PDGFR-α and vimentin. To determine the expression of these markers in CAF216 and CAF227, we evaluated the expression of α-SMA, FAP, PDGFR-α, and vimentin using Western blot and immunofluorescence (Figure 3). We found that both CAF216 and CAF227 expressed all four CAF markers but to variable degrees. CAF227 had much lower FAP and PDGFR-α expression. PDGFR-α was 9.2-fold higher in CAF216 compared to CAF227. FAP was 4.7-fold higher in CAF216 compared to CAF227 (Figure 3B). In contrast, neither CAF line expressed the epithelial marker E-cadherin, which was only expressed in the P411T1 tumor cell line.

FI GURE 3.

FI GURE 3

Open in a new tab

(A) α-smooth muscle actin, fibroblast-activated protein, platelet-derived growth factor-α, vimentin, E-cadherin, and β-actin expression. (B) Western blot quantitated band intensity normalized to β-actin loading control. (C) Immunofluorescence for each antibody of interest, single channel on left in green, with merged image on right with F-actin (red), and 4′,6-diamidino-2-phenylindole (blue) for each myCAF line.

Next, we determined the expression of previously published Elyada myCAF and iCAF gene signature genes in both CAF216 and CAF227 lines.5 We find that CAF216 and CAF227 are similar, with myCAF genes strongly expressed and iCAF genes less expressed (Figure 4A,C). To determine gene expression using more discriminatory genes, we next compared CAF subtype exemplar genes that we identified through SCISSORS.15 We previously derived discriminatory exemplar genes for CAFs, based on the previously described iCAF and myCAF subtypes (SCISSORS iCAF-like and myCAF-like). Both CAF lines were consistent with the myCAF-like signature (Figure 4B) and are hereafter referred to as myCAF216 and myCAF227. Within the two myCAF lines, we found heterogeneity in specific signature genes. CAF216 had higher expression of SCISSORS myCAF-like genes, Wnt family member 5A (WNT5A) and Eva-1 homolog A (EVA1A), and SCISSORS iCAF-like genes hyaluronan synthase 2 (HAS2) and secreted frizzled-related protein 1 (SFRP1); CAF227 expressed slightly more SCISSORS iCAF-like genes, specifically mesenteric estrogen-dependent adipogenesis (MEDAG) and fibulin 2 (FBLN2) (Figure 4B,D). In contrast, the tumor cell line P411T1 had minimal expression of these exemplar CAF genes.

FIGURE 4.

FIGURE 4

Open in a new tab

Expression of Elyada and SCISSORS (Sub-Cluster Identification through Semi-Supervised Optimization of Rare-cell Silhouettes) cancer-associated fibroblast (CAF) genes. Heatmap showing (A) Elyada iCAF and myCAF genes, and (B) SCISSORS iCAF-like and myCAF-like genes. Boxplot showing (C) Elyada iCAF and myCAF genes, and (D) SCISSORS iCAF- and myCAF-like genes.

3.3 |. Paracrine signaling differences within myCAFs

We hypothesized that the heterogeneity in gene and protein expression within myCAF lines may also be associated with differences in biological function. To address this, we isolated CM from each myCAF line. First, we determined the paracrine effects of the myCAFs on tumor cell (P411T1-mCh) proliferation. Compared to control media, myCAF227-CM significantly increased P411T1-mCh proliferation (p = 0.008), while myCAF216-CM had no effect (Figure 5A). Next, we determined the effect of the different myCAF-CM on tumor cell migration. While the myCAF227-CM was able to promote proliferation in P411T1-mCh cells, it had no effect on cell migration. In contrast, the myCAF216-CM significantly decreased PDAC migration (p < 0.001) (Figure 5B,C), despite having no effect on proliferation.

FIGURE 5.

FIGURE 5

Open in a new tab

(A) Proliferation of P411T1-mCh grown in different myCAF CM compared to control media measured by raw luminescence using the CellTiter-Glo assay. (B) Representative photomicrographs of scratch assays of P411T1-mCh in the different myCAF CM compared to control media. (C) Migration quantitated by percent closure at 72 h using ImageJ. NS, nonsignificant.

4 |. DISCUSSION

CAF heterogeneity presents a challenge in studying the TME. We found that CAF markers by either gene or protein expression alone may not be sufficient to discriminate between fibroblasts that are either tumor-promoting or tumor-restraining. Prior failures in translating stroma-based therapies were likely due to heterogeneity within the TME. To develop effective CAF-based therapies to cooperatively target the stroma, accurate preclinical models are needed. While CAF subtypes have been a promising development,5,15 our results show that CAFs are very heterogeneous even within a specific subtype such as myCAFs. Therefore, preclinical results derived from CAF subtypes will require more investigation for successful translation to the clinic.

This study had several limitations. First, only one PDAC and two myCAF lines were evaluated. However, our observations of heterogeneity within myCAF lines support that future work on PDAC and CAF interactions will need to incorporate more models to account for this heterogeneity. Systematic characterization of individual CAF lines in the laboratory will be needed to fully understand how to translate results into effective therapies. Future work will be needed to further classify CAF subtypes in terms of their effect on different tumor cell lines.

5 |. CONCLUSION

Despite having similar gene and protein expression characteristics, myCAF lines had different effects on the tumor cells. Improved classification strategies and the incorporation of more CAF cell lines within studies will be important to better understand and translate these findings.

ACKNOWLEDGMENTS

This work was supported by the National Cancer Institute (T32CA244125 to J. F. K. and H. E. T.; P50CA257911 and U01CA274298 to J. J. Y. and X. L. P.; F31CA257224 to S. A. Z.).

Funding information

National Cancer Institute, Grant/Award Numbers: T32CA244125, P50CA257911, U01CA274298, F31CA257224

Footnotes

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

REFERENCES

  • 1.Park W, Chawla A, O’Reilly EM. Pancreatic cancer: a review. JAMA. 2021;326(9):851–862. doi: 10.1001/jama.2021.13027 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Perez VM, Kearney JF, Yeh JJ. The PDAC extracellular matrix: a review of the ECM protein composition, tumor cell interaction, and therapeutic strategies. Front Oncol. 2021;11:751311. doi: 10.3389/fonc.2021.751311 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Sahai E, Astsaturov I, Cukierman E, et al. A framework for advancing our understanding of cancer-associated fibroblasts. Nat Rev Cancer. 2020;20(3):174–186. doi: 10.1038/s41568-019-0238-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Sousa CM, Biancur DE, Wang X, et al. Pancreatic stellate cells support tumour metabolism through autophagic alanine secretion. Nature. 2016;536(7617):479–483. doi: 10.1038/nature19084 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Elyada E, Bolisetty M, Laise P, et al. Cross-species single-cell analysis of pancreatic ductal adenocarcinoma reveals antigen-presenting cancer-associated fibroblasts. Cancer Discov. 2019;9(8):1102–1123. doi: 10.1158/2159-8290.CD-19-0094 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Hessmann E, Patzak MS, Klein L, et al. Fibroblast drug scavenging increases intratumoural gemcitabine accumulation in murine pancreas cancer. Gut. 2018;67(3):497–507. doi: 10.1136/gutjnl-2016-311954 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Van Cutsem E, Tempero MA, Sigal D, et al. Randomized phase III trial of pegvorhyaluronidase alfa with Nab-paclitaxel plus gemcitabine for patients with hyaluronan-high metastatic pancreatic adenocarcinoma. J Clin Oncol. 2020;38(27):3185–3194. doi: 10.1200/JCO.20.00590 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Benson AB, Abrams TA, Ben-Josef E, et al. Hepatobiliary cancers. J Natl Compr Cancer Netw. 2009;7(4):350–391. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Catenacci DVT, Junttila MR, Karrison T, et al. Randomized phase Ib/II study of gemcitabine plus placebo or vismodegib, a hedgehog pathway inhibitor, in patients with metastatic pancreatic cancer. J Clin Oncol. 2015;33(36):4284–4292. doi: 10.1200/JCO.2015.62.8719 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Özdemir BC, Pentcheva-Hoang T, Carstens JL, et al. Depletion of carcinoma-associated fibroblasts and fibrosis induces immuno-suppression and accelerates pancreas cancer with reduced survival. Cancer Cell. 2014;25(6):719–734. doi: 10.1016/j.ccr.2014.04.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Öhlund D, Handly-Santana A, Biffi G, et al. Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer. J Exp Med. 2017;214(3):579–596. doi: 10.1084/jem.20162024 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Caligiuri G, Tuveson DA. Activated fibroblasts in cancer: perspectives and challenges. Cancer Cell. 2023;41(3):434–449. doi: 10.1016/j.ccell.2023.02.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods. 2017;14(4):417–419. doi: 10.1038/nmeth.4197 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Hao Y, Hao S, Andersen-Nissen E, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573–3587.e293587.e29. doi: 10.1016/j.cell.2021.04.048 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Leary JR, Xu Y, Morrison AB, et al. Sub-Cluster Identification through Semi-Supervised Optimization of Rare-cell Silhouettes (SCISSORS) in single-cell RNA-sequencing. Bioinformatics. 2023;39(8):btad449. doi: 10.1093/bioinformatics/btad449 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Aran D, Looney AP, Liu L, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol. 2019;20(2):163–172. doi: 10.1038/s41590-018-0276-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Frost HR. Variance-adjusted Mahalanobis (VAM): a fast and accurate method for cell-specific gene set scoring. Nucleic Acids Res. 2020;48(16):e94. doi: 10.1093/nar/gkaa582 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Lipner MB, Peng XL, Jin C, et al. Irreversible JNK1-JUN inhibition by JNK-IN-8 sensitizes pancreatic cancer to 5-FU/FOLFOX chemotherapy. JCI Insight. 2020;5(8):e129905. doi: 10.1172/jci.insight.129905 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Moffitt RA, Marayati R, Flate EL, et al. Virtual microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal adenocarcinoma. Nature Genet. 2015;47(10):1168–1178. doi: 10.1038/ng.3398 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nat Methods. 2012;9(7):671–675. doi: 10.1038/nmeth.2089 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Suarez-Arnedo A, Torres Figueroa F, Clavijo C, Arbeláez P, Cruz JC, Muñoz-Camargo C. An Image J plugin for the high throughput image analysis of in vitro scratch wound healing assays. PLoS One. 2020;15(7):e0232565. doi: 10.1371/journal.pone.0232565 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wickham H, Averick M, Bryan J, et al. Welcome to the Tidyverse. J Open Source Softw. 2019;4(43):1686. doi: 10.21105/joss.01686 [DOI] [Google Scholar]
  • 23.R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing; 2022. https://www.R-project.org/ [Google Scholar]
  • 24.Chen Y, McAndrews KM, Kalluri R. Clinical and therapeutic relevance of cancer-associated fibroblasts. Nat Rev Clin Oncol. 2021;18(12):792–804. doi: 10.1038/s41571-021-00546-5 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

RESOURCES