Basal Marker ΔNp63α Is Enriched in Pancreatic Cancer Metastases but Not in Primary Tumors Post‐Chemotherapy Authors: Daniel H. Ryan, Yuuki Ohara, Michael J. Kruhlak, Xianyu Zhang, Miao Lin, Chin‐Hsien Tai, Christine Alewine Published: 2026 ABSTRACT ABSTRACTCancer cells within a single pancreatic ductal adenocarcinoma (PDAC) are most often heterogeneous in their expression of the basal or classical transcriptional program. Predominance of the basal subtype has been associated with poorer patient outcomes and increased resistance to chemotherapy. Based on this, we predicted that basal subtype would be enriched in tumors previously treated with chemotherapy, and in metastases derived from primary PDAC tumors with a mixed basal/classical population. To create matched classical and basal cell lines, we stably transduced two classical PDAC cell lines (AsPC1 and HPAFII) with the basal program transcriptional driver ∆Np63α or an mCherry control and assessed their growth, metastasis, and response to chemotherapy both in vitro and in vivo. Forced expression of ∆Np63α successfully converted both cell lines to basal transcriptomic subtype, yet the basal program failed to drive increased cell growth or chemoresistance in culture and did not affect tumor growth of AsPC1 in the murine pancreas. In mice orthotopically engrafted with ∆Np63α‐expressing cells, no statistically significant increase in spontaneous pulmonary metastases was observed for AsPC1 or HPAFII, however, pulmonary metastases were enriched for ∆Np63 expression as compared to established primary tumors, including expression of endogenous ∆Np63 in the mCherry control. Increased resistance to irinotecan chemotherapy was observed in ∆Np63α + HPAFII, yet treatment selected for ∆Np63‐negative cells in both HPAFII and ASPC1. In summary, our data show that PDAC metastases are enriched for basal subtype, but chemotherapy treatment does not preferentially select for basal‐type cells. Introduction Pancreatic cancer is the third‐leading cause of cancer related deaths in the United States, and is projected to reach second by 2030 due to rising incidence and a dearth of highly effective therapeutic options [1]. By far the most common histology of pancreatic cancer is pancreatic ductal carcinoma (PDAC). PDAC tumors partition into two transcriptomic clusters, named by Moffit and colleagues as classical or basal [2]. The classical transcriptomic subtype of PDAC displays a glandular/progenitor phenotype and is enriched for GATA6 expression, whereas basal PDAC is frequently associated with pockets of squamous histologic phenotype, carries a poorer clinical outcome, and increases resistance to some types of standard chemotherapy [3, 4]. PDAC is transcriptionally heterogeneous, containing geographic subareas of both subtypes within a single tumor and also cells with an intermediate program [5, 6, 7, 8]. Other studies found that cancer cells have sufficient plasticity for tumors to rapidly evolve from one transcriptomic subtype into the other based on external cues from the microenvironment [9]. Conditions which control this evolution are still being elucidated. One molecular driver of the PDAC basal subtype is posited to be ∆Np63 (also called p40), a splicing isoform of the TP63 gene that is both highly expressed in normal squamous tissue and overexpressed in squamous carcinomas [10]. Alternative promoters in the TP63 gene result in production of TAp63 and ∆Np63 transcripts which each contain a unique N‐terminus [11]. Both N‐terminal variants can be coupled with various C‐terminal splice variants (indicated as α, β, γ etc.), resulting in more than 6 unique isoforms which are differentially expressed across tissue type and developmental stage. While the longer TAp63 acts as a tumor suppressor and is expressed at a low level in most non‐stressed tissues, the shorter ∆Np63α variant is the primary form found in epithelial tumors. ∆Np63α controls distinct transcriptional programs and behaves as an oncogene [12]. Forced overexpression of ∆Np63α in human PDAC cells confers transition to the basal transcriptomic subtype, augments colony formation, motility, invasion, and resistance to platinum chemotherapy in vitro, and increases the size of pulmonary metastatic deposits in a murine tail vein injection model [13, 14, 15]. In patients, higher tumor expression of ∆Np63 correlates with decreased survival [14]. Overall, ∆Np63α promotes more aggressive cancer cell behavior and increased metastatic potential. We hypothesized that in tumors with mixed classical/basal subtype (estimated to be 90% of all PDACs [7]), the basal component would be more likely to establish distant metastases and more likely to survive and grow out following treatment with chemotherapy. To test this hypothesis, ∆Np63α was stably expressed in two human PDAC cell lines of classical transcriptomic subtype using lentiviral transduction. Both cell lines, HPAFII and AsPC1, contain a KRAS G12D activating mutation and also an inactivating mutation of the tumor suppressor TP53 as occurs commonly in human patients with PDAC. Rates of spontaneous metastasis and the ability of the tumors to grow through chemotherapy treatment were assessed in +∆Np63α cells as compared to control cells expressing mCherry. We predicted that +∆Np63α should predominate following chemo treatment and in metastatic deposits. Materials and Methods See Supporting Doc S1 for detailed methods for standard assays. Cell Culture and Counting AsPC1 and HPAFII human PDAC cell lines were purchased from ATCC (Manassas, VA, USA). KLM1 was originally obtained from Riken (Saitama, Japan) for use at our institution. Cell line identity was authenticated through short‐tandem repeat testing and confirmed free of mycoplasma. All cell culture reagents were purchased from Gibco/ThermoFisher (Waltham, MA, USA) unless otherwise denoted. Cells were grown at 37°C/5% CO2 in RPMI 1640 medium supplemented with L‐glutamine (2 mmol/L), penicillin (100 U), streptomycin (100 μg), and FBS (10% final volume, Hyclone/ThermoFisher). For proliferation assays, triplicate wells of cells were detached with trypsin for each timepoint, resuspended in complete medium, and counted using a Cellometer Auto T4. Lentiviral Transduction AsPC1 or HPAFII cells were treated with mCherry (Takara Biosciences, San Jose, CA, USA) or ∆Np63α (Genscript, Piscataway, NJ and VectorBuilder, Chicago, IL, USA) lentivirus. Cells were selected using 2 μg/mL of puromycin for 2 weeks. Pooled fractions were propagated to make the stably transduced cell lines. Antibodies for Immunoblot, Immunocytochemistry, Flow Cytometry, and Immunohistochemistry (IHC) For a complete list of antibodies used see Table S1. Flow Cytometry 5 × 105 fixed cells per tube were analyzed using a Sony ID7000 Spectral Cell Analyzer and results were visualized using FloJo. Cell Viability Assays 5 × 103 cells/well were plated on Day 0 then treated with oxaliplatin (Sigma Aldrich, St. Louis, MO or Teva Pharmaceuticals, Cambridge, MA, USA), 7‐ethyl‐10‐hydroxy‐camptothecin (SN38; Sigma Aldrich), or vehicle 24 h later. Relative cell viability was measured using the Cell Counting Kit‐8 WST assay (Vita Scientific, Airmont, NY). Values were normalized between 0% viability (through treatment with 1 mM staurosporine which produced complete cell killing; Sigma Aldrich) and 100% viability for addition of medium alone. Murine Tumor Studies Female athymic nude mice were purchased from Charles River, Frederick, MD, USA. For orthotopic tumor studies, 1 × 106 cells (AsPC1 lines) or 2 × 106 cells (HPAFII lines) in RPMI 1640 media with 3% methylcellulose and no additional additives were orthotopically implanted into the pancreas of 7–11 week‐old mice. Mice were monitored 3 times weekly. For subcutaneous tumor studies, 6–8‐week‐old mice were inoculated subcutaneously with 2 × 106 cells in 4.0 mg/mL Matrigel (Corning, Tewksbury, MA, USA) in RPMI 1640 with no additives. The +mCherry cells were implanted into the right flank and + ∆Np63α cells were implanted into the left flank of each mouse. Mice were monitored 3 times per week for animal weight and tumor volume. When average tumor size in a cohort reached 100 mm3, mice were randomized to receive either vehicle (PBS, Gibco/ThermoFisher, weekly), irinotecan (Camptosar, Pfizer Injectables, 40 mg/kg biweekly), or oxaliplatin (Teva Pharmaceuticals, 10 mg/kg weekly) via intraperitoneal injection. Mice were euthanized when tumors showed signs of ulceration or approached 1000 mm3. IHC Fixed primary tumor tissue was processed by Histoserv (Germantown, MD, USA) and fixed liver and lung tissue by NIH Molecular Histopathology Laboratories (MHL; Frederick, MD, USA) who provided paraffin blocks, unstained slides, and H&E‐stained slides for further analysis. Analysis of H&E‐stained tissue for changes in morphology and for counting of metastatic deposits in liver and lungs was performed by a trained clinical pathologist (Y.O.). Anti‐p40 staining of HPAFII primary tumors and triple‐labeling of lung metastases with p40, Nuclear Mitotic Apparatus (NuMA), and DAPI probes was performed by NIH MHL. Immunofluorescence (IF) Analysis of Tumor Metastases Confocal images were acquired using a Leica Stellaris 5 microscope equipped with a 20× plan‐apochromat objective lens (N.A. 0.75), white light laser and HyD S and HyD X detectors (Leica Microsystems, Deerfield, IL, USA). Images were collected with 0.142 μm X‐Y pixel size and optical section thickness of 3.96 μm. The confocal images were exported as OME‐TIFF files and imported into the Nikon Elements image analysis software (V. 5.4.1) (Nikon Instruments, Melville, NY, USA) where they were processed using the Denoise.ai algorithm. The denoised images were imported into ArivisPro (V.4.3) image analysis software (Carl Zeiss Microscopy, Thornwood, NY, USA) for cell segmentation and quantitation of fluorescence intensity from the individual fluorescent labels. Briefly, a base Cellpose model [16] was updated using human‐in‐the‐loop retraining of the model to segment specifically cells expressing NuMA. The cell mask was dilated by a maximum distance of 2 μm and the mean fluorescence intensity (MFI) of the remaining channels was measured for each cell. A negative labeling control was used to measure background fluorescence intensity. The background subtracted MFI measurements were used to distinguish between NuMA‐positive cells that were either p40 positive or negative. The percentage of p40 positive cells of all NuMA‐positive cells per lesion in each sample was calculated. Analysis of RNA Sequencing Data RNA samples (n = 5 from each condition) were sequenced by the NCI CCR Genomics Core (Bethesda, MD, USA) with Illumina NextSeq 2000 (serial number VH01090). Sequence data was processed by standard methods (Supporting Doc S1) and deposited in GEO (GSE307048). DESeq2 [17] and GSEA [18, 19] were used for gene differential analysis and gene set enrichment analysis with Hallmark Gene Set and the basal and classical gene sets defined by Moffitt and colleagues [2]. Statistical Analyses GraphPad Prism (GraphPad Software, Boston, MA, USA) was used for graphing, statistical calculations, and curve fitting unless otherwise indicated. Data are presented as mean ± SD. Comparisons of > 2 groups were assessed by ANOVA. For post hoc analysis and two group comparisons, Mann–Whitney nonparametric test or two‐tailed Student t‐test with Welch correction were used. No significant difference is denoted by ns. All experiments were confirmed by repeat. Results Generation of ∆Np63α‐Expressing Cell Lines Expression of p63α was observed in AsPC1 and HPAFII cells transduced to stably express ∆Np63α as compared to the parent cell lines. Although levels of the transgene were lower than endogenous expression in the basal cell line KLM1, sufficient ∆Np63α was available to induce expression of basal transcriptomic program markers KRT17 and S100A2 (Figure 1A,B; Figure S1A). Of note, IF and flow cytometry analyses demonstrated heterogeneous expression of the ∆Np63α transgene in +∆Np63α cells (Figure 1C,D; Figure S1B), but relatively universal and uniform expression of mCherry in the transduced control cell lines (Figure S1B). Several reports have suggested that overexpression or knockdown of ∆Np63α in PDAC alters tumor cell growth in culture, although these reports differ in the directionality [13, 14]. Here, no significant difference in doubling time was observed for +∆Np63α cells as compared to +mCherry cells when grown on tissue culture plastic (Figure 1E,F). Forced expression of mCherry and ∆Np63α in classical cell lines does not change cell growth in culture. HPAFII and AsPC1 pancreatic cancer cells were transduced with lentivirus to stably express ∆Np63α (experimental) or the fluorophore mCherry (control). The basal PDAC cell line KLM1 was used as a positive control for ∆Np63 expression. (A, B) Western blot showing expression detected by anti‐p63antibody or of GAPDH loading control. (C, D) Immunofluorescence using anti‐p63 antibody or Hoechst nuclear stain. (E) Representative proliferation curves of the cell lines. (F) Doubling time as measured in triplicate cell counting experiments. Evaluating the Effect of ∆Np63α Expression on Tumor Growth and Architecture In Vivo To assess effects of ∆Np63α in vivo, the +∆Np63α and + mCherry cell lines or a 1:1 suspension of both +∆Np63α and + mCherry cells together (1:1 Mixed) were orthotopically engrafted into the pancreas of athymic nude mice. Tumors were permitted to mature for 4–5 weeks until some mice in the cohort exhibited signs of illness, then all mice were euthanized. No difference in the size of primary tumors was noted amongst the transduced AsPC1 cells, however, increased primary tumor size at end point was noted for HPAFII +∆Np63α and 1:1 Mix tumors (Figure 2A). Interestingly, this difference was large enough to reach statistical significance in the 1:1Mix but not +∆Np63α, although 1:1 Mix growth was not statistically different from +∆Np63α growth. Notably, although ∆Np63 expression was easily detected in AsPC1 and HPAFII +∆Np63α and 1:1 Mix tumors, ∆Np63α expression was variable in HPAFII such that +∆Np63α cells did not necessarily generate increased ∆Np63 expression compared to 1:1Mix (Figure S2B,C). No significant difference in primary tumor morphology was noted in +∆Np63α or 1:1 Mix tumors on H&E (Figure 2B, Figure S2D,E), although forced ∆Np63 expression was previously reported to result in more poorly differentiated tumors [14]. Bulk RNA deep sequencing analysis of tumor tissue from the mice confirmed a shift from the classical transcriptomic subtype in tumors derived from +mCherry control cell lines to a basal transcriptomic subtype in tumors from +∆Np63α cell lines (Figure 2C). Increased expression of TP63 and basal markers KRT17 and S100A2 were observed in all mice bearing +∆Np63α tumors (Figure 2D,E). In addition, a trend toward decreased expression of classical transcriptomic marker GATA6 was noted at the RNA level in ∆Np63α expressing tumors for both AsPC1 and HPAFII expression, however, decreased GATA6 protein upon ∆Np63α expression was only seen in AsPC1 (Figure 2E, Figure S2B,C). As expected, forced expression of ∆Np63α converted classical PDAC cells and tumors to express the basal program. Forced expression of mCherry and ∆Np63α in classical cell lines induces basal transcriptomic signature. AsPC1 or HPAFII +mCherry, + ∆Np63α or a 1:1 Mix of both cell types were orthotopically implanted into the pancreas of athymic nude mice and allowed to mature for 4–5 weeks before tumor harvest. (A) Tumor volume was assessed at end point. For AsPC1 (top), mass of primary tumor after dissection from normal pancreas is reported. For HPAFII (bottom), 3 mice in +mCherry group had no visible tumor. Mass of pancreas+tumor is reported following subtraction of estimated normal pancreas weight. (B) Immunohistochemistry of mouse tumors showing representative hematoxylin and eosin (H&E), and ∆Np63 (anti‐p40 which recognizes all isoforms) staining. Scale bars =50 μm. Inset shows full tumor spread with orange dot indicating location of high‐power image. (C) Bulk RNA sequencing was performed on murine tumors (n = 5/group). GSEA analysis was used to assess basal versus classical transcriptomic subtype in +∆Np63 tumors as compared to +mCherry control. Enrichment plots of three gene sets: Basal‐like UP Moffitt, Basal‐like DOWN Moffitt and Hallmark EMT. (D, E) TPM data from RNA seq for specified genes of interest. Each marker represents value for 1 mouse. *indicates p < 0.05. Effect of ∆Np63α on Formation of Spontaneous Metastases Despite the switch to basal subtype, the rates of spontaneous lung and liver metastases trended higher in mice bearing AsPC1 + mCherry control tumors than in those expressing ∆Np63α (Figure 3A). Conversely, increases in spontaneous lung and liver metastases were observed in HPAFII+∆Np63α and 1:1 Mix, although this did not reach statistical significance (Figure 3B). Determining the exact cause for this cell line‐specific difference is beyond the scope of the present work; however, it is interesting to note that GSEA analysis identified enrichment of the epithelial‐to‐mesenchymal transition (EMT) pathway in primary HPAFII+∆Np63α tumors as compared to +mCherry, while this pathway was attenuated by expression of ∆Np63α in AsPC1 primary tumors (Figure 2C). Forced ∆Np63α expression can increase spontaneous metastases. The indicated cell lines or a 1:1 Mix of both were orthotopically implanted into the pancreas of athymic nude mice and allowed to mature for 4–5 weeks before lungs and liver were harvested. (A, B) Spontaneous metastases to liver and lung identified on H&E by a trained pathologist were assessed in AsPC1 (A) and HPAFII (B) models. Left‐ Legend indicates number of pulmonary metastases found in each mouse. Right‐ Black bar shows number of mice with any hepatic metastases. (C) Representative immunofluorescence images of lung tissue from mice bearing HPAFII +mCherry and + 1:1Mix tumors. Anti‐NuMA marks human cells (green), anti‐p40 marks ΔNp63 expression (white), and DAPI shows cell nuclei (blue). Scale bars show 50 μm (D) Quantitation of stained cells from C with calculated percentage of Numa and ΔNp63 (p40) double‐positive cells. Graphed in Microsoft Excel. (E) FFPE blocks of corresponding primary tumors were stained with p40 and percentage of positive tumor cells was quantified by a trained pathologist. We hypothesized that metastases from HPAFII 1:1Mix tumors should preferentially form from ∆Np63α‐expressing cells, since basal subtype is known to be more aggressive. Therefore, we stained lung tissue from mice bearing these tumors for NuMA protein to identify deposits of metastasized human cancer cells and co‐labeled with p40 (which recognizes all ∆Np63 isoforms) to mark cells expressing ∆Np63. ∆Np63+ cells represented 26%–79% of metastatic cancer cells in mice inoculated with 1:1Mix (Figure 3C,D). Interestingly, although 50% of cells inoculated to form primary tumors were HPAFII‐ΔNp63α, only 6%–30% of tumor cells in established primary tumors stained positive for p40 (Figure 3E), percentages lower than those observed in spontaneous metastasis, demonstrating that ∆Np63‐expressing cells are enriched in metastasis. In addition, 59%–96% of spontaneous lung metastasis derived from mCherry control cells stained positive for p40 (Figure 3C,D), although no p40 expression was identified in primary tumors (Figure 3E). Control cells with sufficient plasticity to upregulate endogenous ∆Np63 were also more likely to establish metastases and thrive in the pulmonary niche. In summary, spontaneous pulmonary metastases were enriched for a ∆Np63‐expressing cell population. Effect of ∆Np63α on Chemosensitivity The FOLFIRINOX chemotherapy regimen, consisting of 5‐fluorouricil, leucovorin, irinotecan, and oxaliplatin, is widely considered the most active drug combination in PDAC and has produced the best overall survival when given in the adjuvant setting to early‐stage patients following surgical resection [20]. Several groups have reported increased resistance of basal‐type PDAC to chemotherapy, especially to platinum drugs like oxaliplatin, and ∆Np63α expression has also been implicated in chemotherapy resistance [4, 13, 21]. By contrast, we detected no significant difference in IC50 for oxaliplatin or SN38 (the metabolized form of irinotecan) in +∆Np63α and + mCherry cells grown on tissue culture plastic (Figure 4A,B). Influence of ∆Np63α expression on chemotherapy sensitivity in culture. (A, B) Cells were plated on tissue culture plastic and treated with the indicated chemotherapy agent. Left‐ Representative viability curves. Right‐ Summary of IC50 values for each chemotherapy treatment. To determine whether the effect of +∆Np63α on chemotherapy resistance might only occur in the in vivo setting, mice were inoculated subcutaneously with +∆Np63α tumor cells in one flank and + mCherry tumor cells in the opposite flank then treated with chemotherapy or vehicle and tumor volume was serially measured (Figure 5A,B). With the chemotherapy doses used, mice maintained their body weight throughout the experiment (Figure S3A,B), and no significant difference in tumor growth of +∆Np63α versus +Cherry tumors was observed in the vehicle treated mice (Figure 5A). These data demonstrate that expression of ∆Np63α does not confer a growth advantage to subcutaneous PDAC deposits, in contrast to the growth advantage we observed in HPAFII+∆Np63α cells grown orthotopically in the pancreas. Influence of ∆Np63α expression on chemotherapy sensitivity in vivo. (A, B) Athymic nude mice were inoculated subcutaneously with +mCherry cells in one flank and + ∆Np63α cells in the opposite flank, then treated with vehicle (once weekly PBS), irinotecan (40 mg/kg biweekly) or oxaliplatin (10 mg/kg once weekly). Mice were euthanized if either tumor became necrotic. (A) Subcutaneous tumor volume was serially measured for each mouse. (B) Tumor volume of +mCherry versus +∆Np63α at natural death or euthanasia. Bars connect the two flank tumors in each mouse. * indicates p < 0.05. Expression of ∆Np63α also failed to enhance resistance to irinotecan or oxaliplatin in AsPC1 tumors (Figure 5A,B). However, HPAFII+∆Np63α tumors displayed increased resistance to irinotecan, as measured by increased size of these tumors at time of euthanasia as compared to +mCherry tumors grown in the opposite flank (Figure 5B). Irinotecan‐treated HPAFII+∆Np63α tumors also had a higher percentage of cancer area within each tumor as compared to cognate +mCherry tumors (Figure S3C). In addition, there was a trend toward increased resistance to oxaliplatin in HPAFII+∆Np63α, but this did not reach statistical significance (Figure 5B). In summary, enhanced chemoresistance of ∆Np63α‐expressing tumor cells was cell‐line dependent. Chemotherapy Selects for ∆Np63α Non‐Expressors Since ∆Np63α expression did affect chemotherapy sensitivity in HPAFII cells, we predicted that chemotherapy should select for ∆Np63α expression, so that ∆Np63α and basal markers such as SOX2 should be enriched in HPAFII post‐treatment. However, we found that ∆Np63α expression was diminished in tumors from mice treated with irinotecan in both the HPAFII and AsPC1 models as compared to vehicle (Figure 6A,B). In the AsPC1 model, oxaliplatin treatment also diminished ∆Np63α expression, further emphasizing that this change in expression did not correlate with therapy resistance. Endogenous SOX2 expression mirrored that of ∆Np63α, demonstrating the change was not specific to the transgene alone but was reflected in other proteins associated with basal subtype. Similarly, decreases in basal transcriptomic markers KRT17 and S100A2 were seen when AsPC1 + ∆Np63α and HPAFII + ∆Np63α grown in culture were exposed to 2 days of SN38 or oxaliplatin (Figure 6C). Decreases in basal markers were not as robust after 7 days of chemotherapy treatment, but no increase was observed. In summary, chemotherapy treatment appeared to select against ∆Np63α expression in direct contradiction to our hypothesis. Irinotecan selects against ΔNp63α expression. (A, B) Western blot of ΔNp63α ‐expressing AsPC1 (A) and HPAFII (B) subcutaneous tumor samples from athymic nude mice treated with vehicle, irinotecan, or oxaliplatin. Samples were probed for p63, ΔNp63α, and SOX2 (basal marker), and GAPDH (loading control). (C) Quantitative RT‐PCR was performed using lysates from ΔNp63α expressing cell lines grown in culture following 2 or 7 days of treatment with DMSO (control), SN38, or oxaliplatin. Chemotherapy concentrations were chosen to be IC75 for each drug/cell type (HPAFII: SN38 = 175 nM, oxaliplatin = 250 nM; AsPC1: SN38 = 50 nM, oxaliplatin = 200 nM). The 2^ΔCt value for control was normalized to 1.0 for each gene and condition. Mean value of triplicate biologic replicates is plotted with error bars showing SEM. *indicates p < 0.05. Discussion Clinical data from multiple cohorts support claims that basal PDAC has poorer overall survival whether basal status was assessed by the gold standard of performing RNA sequencing to classify cases [4, 7], or through use of biomarker surrogates such as low GATA6 or high p63 expression [22, 23, 24]. Differences in patient overall survival may reflect enhanced intrinsic aggressiveness of these tumors, a poorer response to available therapies, or both. Data suggesting enhanced aggressiveness include findings that basal tumors are more poorly differentiated [5, 22], and that basal subtype is under‐represented in early‐stage disease and over‐represented in newly diagnosed metastases [7, 8]. Based on these clinical data, we anticipated that a forced switch to a basal subtype through genetic manipulation would enhance the tumor growth and metastasis of PDAC and produce a highly controlled model system to address whether basal subtype cells are more successful at establishing metastases. To implement the subtype switch, we utilized forced expression of ∆Np63α which was previously found sufficient for this purpose [14, 15]. We anticipated that this change would increase cell growth in culture, tumor growth in vivo, and spontaneous metastasis, as has been seen previously with overexpression of ∆Np63α in SUIT‐2 [14] and for the in vitro measures in Panc‐1 [13]. Surprisingly, basal subtype switch through forced expression of ∆Np63α did not affect in vitro cell growth, and had mixed results in promoting tumor growth and metastasis in the AsPC1 and HPAFII models that we tested despite transforming the cells to basal subtype. Our data confirm that a ∆Np63α‐mediated transition to basal subtype can enhance tumor growth and metastasis in some contexts, but also illuminated that it is not universally sufficient to do so. Our data show that the phenotype spectrum caused by forced ∆Np63α expression is cell‐line dependent. By contrast, attenuating or knocking out ∆Np63α expression in basal cell lines appears to universally impair tumor growth [13, 14, 15]. As described by Somerville and colleagues [14], PDAC cells that have evolved in the presence of an active ∆Np63α transcription program have developed an addiction to these drivers. Our present research study did not assess the effect of ∆Np63α loss or knock‐down in a basal cell line, which is an important limitation of our study. Nevertheless, our data support the concept that transcriptomic subtype alone need not dictate phenotype. Consistent with this, Kloesch and colleagues found that overexpression of the classical subtype driver GATA6 in basal cell lines (or knock‐down in classical cells) successfully switched their subtype, but did not alter proliferation, migration, or invasion [25]. Perhaps for some cell lines, gain of drivers other than ∆Np63α are more important in conveying aggressive phenotype. For instance, it has been shown that promotion of a basal subtype in AsPC1 through overexpression of basal marker SERPINB3 was sufficient to promote spontaneous metastasis [26]. Taken together, these data suggest that the basal transcriptomic profile is a strong marker for a more aggressive phenotype but not necessarily the sole cause of it, and that other factors contribute to driving aggressive phenotypes frequently associated with a basal profile. Here, we found that metastatic proficiency was better associated with activation of the EMT program, while others have demonstrated concordance with MYC expression [27], or chromosomal imbalance [8]. Although a clear increase in the number of metastases was not observed in cells genetically engineered to express ∆Np63α, we observed that cells successful at spontaneously metastasizing to the lung were more likely to express ∆Np63 than the corresponding primary tumor. In fact, expression of ∆Np63 in mCherry control metastases occurred in more than 50% of cells, presumably due to spontaneous upregulation of the endogenous gene which was not detectable in corresponding primary tumors. Similarly, pulmonary metastases from 1:1 Mix were also more likely to express ∆Np63 than primary tumors. While the number of samples assessed in the analysis was small, and analyses were performed using different methods (IF versus IHC), these data suggest that the basal expression pattern makes cells more fit to metastasize and/or survive in the metastatic niche. These data are consistent with clinical data reporting that basal subtype is over represented in metastases. Basal PDAC is widely believed to respond more poorly to FOLFIRINOX chemotherapy than the classical subtype based on the COMPASS study cohort [4], and additional circumstantial data supports this [28, 29]. Based on this, we hypothesized that treatment with irinotecan or oxaliplatin, key components of FOLFIRINOX, should select for basal subtype cells. In fact, we found the opposite. ∆Np63α expression was decreased in tumors following chemotherapy treatment in both AsPC1 (which showed no evidence of primary chemo resistance) and in HPAFII (which did appear partially resistant). Few assessments of subtype switching following chemo have been published. Consistent with our results, a recent single‐cell RNA analysis of 6 human PDAC tumors serially acquired pre‐ and post‐chemotherapy similarly found no enrichment for basal subtype after treatment, and even noted increased classical subtype score for treated samples [6]. Another study identified a post‐chemo switch from basal to classical in the single case analyzed [8]. Kokumai and colleagues more comprehensively evaluated phenotype switching after neoadjuvant chemo using a dual IHC biomarker system and noted that 55.2% of basal‐like tumors (defined as GATA6‐/CK5+, n = 29) continued to classify as basal‐like post‐treatment, 31% became more classical (GATA6+), and only 13.8% became more poorly differentiated (double‐negative). Similarly, 43% (nearly half) of hybrid and classical tumors (n = 49 total) were more likely to become more basal‐like or undifferentiated post‐treatment [22], a ratio about as predictive as a coin toss. Our study is unable to add any evidence that chemotherapy selects for the basal subtype, or that basal subtype confers resistance to FOLFIRINOX components. Instead, it calls attention to clinical studies unable to find an association between transcriptomic subtype and response to specific chemotherapy regimens [7, 22]. Our study has several limitations which could impact the generalizability of our results. We have focused our efforts on two standard PDAC cell lines; representing the overall heterogeneity of PDAC would require many more unique lines. As these are human cell lines, this required the use of immunosuppressed mice for in vivo experiments, so we are unable to evaluate how the full immune milieu contributes to ∆Np63α + tumor behavior. In addition, we did not engineer conversion of a basal cell line to classical subtype through knockdown of ∆Np63α. Nevertheless, our data suggest that conversion to basal transcriptomic subtype through expression of ∆Np63α is insufficient to universally drive more aggressive tumor behavior such as growth, metastasis, and chemoresistance, nor did chemotherapy treatment select for outgrowth of more basal‐like clones. Author Contributions Daniel H. Ryan: investigation, formal analysis, methodology, writing – original draft, writing – review and editing. Xianyu Zhang: investigation, writing – review and editing. Christine Alewine: conceptualization, resources, funding acquisition, supervision, methodology, writing – original draft, writing – review and editing. Miao Lin: investigation. Chin‐Hsien Tai: formal analysis, methodology, writing – original draft, writing – review and editing. Yuuki Ohara: investigation, formal analysis, methodology, writing – review and editing. Michael J. Kruhlak: investigation, formal analysis, writing – original draft, writing – review and editing. Funding This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research (Project No. ZIA bc 011652), and also by funding from Friends of Dartmouth Cancer Center (to C.A.). Computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov), Dartmouth Cancer Center Shared Resources (CCSG: P30CA023108) and Microscopy Shared Resource (RRID: SCR_025077) were utilized in the research. The contributions of the NIH author(s) were made as part of their official duties as NIH federal employees, are in compliance with agency policy requirements, and are considered Works of the United States Government. However, the findings and conclusions presented in this paper are those of the author(s) and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services. Ethics Statement All mouse experiments were approved by the NIH Animal Care and Use Committee. No human research is described in this manuscript. Approval of the research protocol by an Institutional Review Board, informed consent processes, and registry/registration# are not applicable. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Forced expression of mCherry and ΔNp63α in classical cell lines. HPAFII and AsPC1 pancreatic cancer cells were transduced with lentivirus to stably express ΔNp63α (experimental) or the fluorophore mCherry (control). (A) Quantitative RT‐PCR evaluation of ΔNp63 and basal transcriptomic subtype markers S100A2 and KRT17 in mCherry‐ versus ΔNp63α‐transduced cells. Representative run of 5 biologic replicates. (B) Flow cytometry analysis to assess expression of p63 using anti‐p63α antibody conjugated to AlexaFluor488 (AF488) in fixed and permeabilized cells or mCherry in unpermeabilized cells. The basal PDAC cell line KLM1 was used as a positive control for ΔNp63 expression. Figure S2: Orthotopic mouse model. The +mCherry, + ΔNp63α or a 1:1 Mix of both cell types were orthotopically implanted into the pancreas of athymic nude mice and allowed to mature for 4–5 weeks before tumor harvest. (A) Average mouse weights (±standard deviation) during the experiment. The experiment was conducted twice (Expt#1: top, Expt#2: bottom). (B, C) Immunoblot of murine orthotopic primary tumor samples for protein markers of transgene expression (ΔNp63α, p63α, mCherry), and classical subtype (GATA6). Samples from 2 tumors for each condition are shown. Results were confirmed by repeat. (D) GSEA assessment of EMT pathway in +ΔNp63 tumors as compared to +mCherry control using bulk RNA from orthotopic tumors (n = 5/group). (D) Low power view, p40 staining of AsPC1 FFPE tumor samples. All scale bars 1 mm. (E) Low power view, p40 staining of HPAFII FFPE tumor samples. All scale bars 1 mm. Figure S3: Subcutaneous mouse model. Athymic nude mice were inoculated subcutaneously in one flank with +mCherry cells and in the opposite flank with + ΔNp63α cells. Mice were treated with vehicle, irinotecan or oxaliplatin after tumor maturation. (A, B) Individual weights from mice. (C) Histologic analysis of post‐treatment tumors as quantified by a trained pathologist based on H&E stained slides. Data S1: Detailed Methods. Table S1: Identification of antibodies reagents and their use conditions.