KRAS is the most frequently mutated oncogene in cancer, yet there is little understanding of how specific KRAS amino acid changes affect tumor initiation, progression, or therapy response. Using high-fidelity CRISPR-based engineering, we created an allelic series of new LSL-Kras mutant mice, reflecting codon 12 and 13 mutations that are highly prevalent in lung (KRASG12C), pancreas (KRASG12R), and colon (KRASG13D) cancers. Induction of each allele in either the murine colon or pancreas revealed striking quantitative and qualitative differences between KRAS mutants in driving the early stages of transformation. Furthermore, using pancreatic organoid models, we show that KRASG13D mutants are sensitive to EGFR inhibition, whereas KRASG12C-mutant organoids are selectively responsive to covalent G12C inhibitors only when EGFR is suppressed. Together, these new mouse strains provide an ideal platform for investigating KRAS biology in vivo and for developing preclinical precision oncology models of KRAS-mutant pancreas, colon, and lung cancers.
KRAS is the most frequently mutated oncogene. Here, we describe new preclinical models that mimic tissue-selective KRAS mutations and show that each mutation has distinct cellular consequences in vivo and carries differential sensitivity to targeted therapeutic agents.
See related commentary by Kostyrko and Sweet-Cordero, p. 1626.
This article is highlighted in the In This Issue feature, p. 1611
KRAS is the most frequently mutated oncogene in human cancers and considered a key early driver of many tumors. Specific cancer types show a clear bias in the types and frequency of KRAS alterations (1, 2), and although carcinogen-specific mutational signatures define a subset of tissue-selective KRAS changes, they do not account for the majority of tissue-selective KRAS alterations (3, 4). Biochemically, oncogenic KRAS mutations increase the abundance of GTP-bound “active” KRAS protein, but different amino acid changes can significantly alter the kinetics of GDP/GTP exchange and GTP hydrolysis (5). Such changes may have implications for signaling dynamics in different cell or tissue contexts. Finally, mounting clinical and preclinical evidence suggests that tumors carrying distinct KRAS variants are differentially sensitive to targeted therapies (6–8). Thus, despite genetic and epidemiologic evidence that differences between KRAS mutations are functionally important, we still do not have a clear understanding of how distinct KRAS alterations dictate tumor initiation, disease progression, or response to therapy.
Conditional animal models, such as KrasLox-Stop-Lox (LSL)-G12D and KrasLSL-G12Vgeo mice, developed almost 20 years ago (9, 10), have been critical tools for dissecting the role of KRAS mutations in tumor development. However, these models alone do not recapitulate the spectrum of KRAS alterations in human cancer. Here, we describe an efficient pipeline for engineering an allelic series of conditional alleles that significantly expands the repertoire of preclinical KRAS-driven cancer models. Using high-fidelity CRISPR targeting in embryonic stem cell (ESC)–based mouse models (11–14), we engineered six new KrasLSL-mutant alleles (G12V, G12C, G13D, G12R, G12A, and G12S) that represent the most frequent mutations at the G12/G13 hotspot, after G12D. Guided by clinical data, we generated conditional mice representing three tissue-selective alterations observed in colorectal (KRASG13D), pancreatic (KRASG12R), and lung cancers (KRASG12C) and show that even subtle mutational changes in KRAS affect premalignant changes in the colon and pancreas.
In line with these diverse biological outputs in vivo, pancreatic organoids derived from these models uncovered KRAS variant–specific vulnerabilities. Specifically, we show that KRASG13D-mutant cells are sensitive to inhibition of the EGFR, and that combined treatment with RTK and covalent KRASG12C inhibitors is required for potent growth suppression in KRASG12C/p53-mutant organoids. Thus, these new animal models serve as a powerful preclinical resource to interrogate KRAS biology in vivo and develop rational strategies to effectively target specific KRAS-mutant cancers.
A CRISPR-Based Pipeline for Engineering Kras Allelic Variants in an Established Conditional Allele
To engineer new KrasLSL-mutant alleles, we used CRISPR-mediated homology-directed repair (HDR) to introduce specific codon 12/13 mutations into the well-characterized KrasLSL-G12D allele. For this, we took advantage of previously derived genetically engineered embryonic stem cells (GEMM-ESC; ref. 14) carrying the endogenous KrasLSL-G12D allele, with a pancreas-specific Cre recombinase (Ptf1a-Cre, also known as p48-Cre) and a far-red fluorescent Cre-reporter (CAGs-LSL-rtTA3-IRES-mKate2, hereafter LSL-mKate2). The GEMM-ESC approach enables the rapid creation of mouse cancer models without the need for extensive intercrossing to generate appropriate genotypes, which is particularly important when modeling multiple new genetic modifications that may take many years to breed (refs. 11, 13, 14; Fig. 1A). Alleles derived from GEMM-ESCs can be outcrossed from the founder generation to establish independent strains.
We first designed a single-guide RNA (sgRNA) overlapping codons 12 and 13 of Kras and 140-mer single-stranded DNA (ssDNA) HDR templates carrying specific mutations (Fig. 1B), and introduced them into ESCs by nucleofection. For screening and genotyping purposes, each ssDNA template was designed to carry silent mutations that generate a unique restriction site adjacent to the specific codon change (Fig. 1B). Analysis of the bulk transfected population revealed efficient generation of insertions/deletions (indels) in both the LSL-G12D and wild-type (WT) Kras alleles, suggesting that the single mismatch between the sgRNA and target region in WT Kras was not sufficient to dictate allele selectivity (Fig. 1C; Supplementary Fig. S1A). Indeed, every clone we identified that carried the desired HDR event on the LSL allele carried an indel in the WT coding sequence (14/14 clones sequenced). Our follow-up efforts to first introduce silent mutations in the WT allele and then retarget the KrasLSL-G12D site also failed due to the introduction of indels in Nras with the Kras WT–specific sgRNA (Supplementary Fig. S1B–S1D).
During the course of this work, we independently developed an expression-optimized high-fidelity Cas9 variant that enabled selective and potent genome targeting (15, 16). We thus tested whether this would provide the specificity required for selective LSL-Kras targeting. In contrast to what we observed following transfection of WT Cas9, expression of the optimized HF1 enzyme resulted in the generation of both indels and HDR integration in the LSL allele, but induced no detectable modifications in WT Kras (Fig. 1C; Supplementary Fig. S2A). Although we noted an overall decrease in the efficiency of HDR-targeting in comparison with WT Cas9 (5% vs. 21%), the increased specificity allowed the identification of numerous clones carrying the desired targeting event (Supplementary Table S1). Targeted deep sequencing of Kras exon 2 revealed a consistent 5% to 6% HDR frequency, though the number of individual positive clones identified following each transfection varied from 1% to 5% due to random clone selection (Supplementary Table S1). Clones identified to carry integration of the donor template by restriction digest were confirmed by allele-specific PCR and direct Sanger sequencing (Fig. 1D; Supplementary Fig. S2B).
To determine the impact of selective KRAS mutational variants on cell and tumor biology, we chose to generate mice from three LSL-Kras genotypes: KrasLSL-G12C, KrasLSL-G12R, and KrasLSL-G13D, as these mutations represent frequent and tissue-restricted mutational events in human colorectal cancer, pancreatic ductal adenocarcinoma (PDAC), and lung adenocarcinoma, respectively (Fig. 1E). To confirm that CRISPR-mediated HDR targeting had not caused widespread mutagenesis or large-scale chromosome aberrations, we performed whole-exome sequencing (WES) on targeted ESC clones, those transfected with only the expression-optimized HF1 Cas9 variant (no sgRNA), or mock-transfected cells (PAR). Consistent with highly selective targeting of the Kras locus by this sgRNA, we observed no differences in the number of de novo mutations (average 13.5 mutations/clone) in either PAR or HF1-only transfected clones, and those transfected with both HF1, sgRNA, and HDR template (Supplementary Fig. S3A; Supplementary Table S2). In fact, the number of unique single-nucleotide mutations detected in each clone was in line with the expected baseline mutation rate predicted in mouse ESCs (13.5/clone observed vs. 10.4/clone estimated; see Methods for details; ref. 17). Moreover, most mutations were single-nucleotide variants, rather than indels usually associated with Cas9-mediated mutagenesis, further suggesting they arose spontaneously during ESC culture. We did not detect any chromosome copy-number alterations in targeted clones, with the exception of one KrasLSL-G13D clone that showed a small deletion on chromosome 4 (Supplementary Fig. S3B); this clone was not used for mouse generation.
To produce mice, targeted KrasLSL-mut ESC clones were injected into host albino C57BL/6J blastocysts, creating a range of high-contribution chimeras (Supplementary Fig. S3C); we further bred the founders to C57BL/6N mice to increase the number of animals for this study. To first confirm that each of the new strains showed equivalent expression of the Kras-mutant allele, we generated multiple independent murine embryonic fibroblast (MEF) cultures from each KrasLSL-mut line (n = 3) and immortalized the cells by disruption of p53 with CRISPR. Delivery of Cre recombinase on the same vector as Cas9 (Cas9-P2A-Cre) enabled simultaneous induction of each Krasmut allele. As expected, all mutant alleles were expressed similarly, as measured by allele-specific transcript reads counts (Fig. 1F) or total KRAS protein (Supplementary Fig. S3D). Consistent with previous analysis of KrasG12D MEFs (18), induction of endogenous Kras mutations in Trp53 WT cells led to a proliferative advantage, but there was no consistent difference in proliferation over 3 weeks between MEFs carrying each different Kras-mutant allele (Fig. 1G). RNA-sequencing (RNA-seq) analysis in Krasmut/Trp53KO MEFs (hereafter KrasMUT), revealed a range of transcriptional changes between WT and mutant cells, including the upregulation and downregulation of genes previously linked to KRAS-driven transformation in murine fibroblasts (Fig. 1H; Supplementary Table S3; ref. 19). Notably, although each of the Kras variants carried the mutant transcriptional signature, the magnitude of the effect was markedly reduced in KrasG12R and KrasG13D cells (Fig. 1H). RAS-GTP levels were increased in all KRAS-mutant MEFs, most elevated in KrasG12D and KrasG12C, and lower in KrasG12R and KrasG13D cells (Fig. 1I). In low serum culture conditions, only KrasG12D MEFs displayed clear activation of the downstream signaling effectors MEK/ERK/AKT, whereas KrasG12C and KrasG13D cells showed a lower level of phospho-MEK (pMEK)/pERK than KrasG12D, but elevated above KrasWT cells. KrasG12R MEFs showed pMEK/pERK levels similar to KrasWT cultures. All genotypes showed similar phosphorylation of AKT and S6, with the exception of KrasG12R-expressing MEFs, which had reduced pAKT and pS6, even below that of KrasWT cells. All MEFs remained responsive to upstream mitogenic signals, as acute EGF stimulation increased MEK, ERK, and AKT phosphorylation in all genotypes, including KrasG12R (Fig. 1I). Together, these data show that each KrasLSL-mut strain enables comparable induction of endogenous KrasLSL-mut alleles, and that although each mutation drives KRAS-associated phenotypes, the downstream effect of individual codon 12/13 Kras mutations are not identical.
In Vivo Consequences of Different Kras Mutations in the Colon and Pancreas
To determine how distinct oncogenic alterations in codons 12/13 of Kras affect tumor initiation in vivo, we built mouse models expressing each oncogenic Kras allele in the epithelium of the colon or pancreas. We chose these two organs because KRAS mutations occur in 40% to 50% of colorectal cancer cases and are a near universal feature of PDAC, where KRAS mutations are considered the key disease-initiating event (14, 20–22).
Previous work has demonstrated that induction of KrasG12D mutations in the mouse colon epithelium drives widespread hyperplasia, characterized by lengthening of the crypts (23). To assess the impact of non-KrasG12D mutations on homeostasis of the colonic epithelium, we generated mice carrying each KrasLSL-mut allele and Fabp1-Cre (24). In this model, Cre recombinase is expressed mid-gestation and activates KrasMUT in the colon and distal ileum (24). Consistent with published data, by 8 weeks of age colons expressing KrasG12D displayed widespread hyperplasia, showing significant thickening of the mucosa (Fig 2A–C, top; refs. 23, 25). KrasG12C mutation induced a dramatic hyperplastic response, identical to that seen in KrasG12D colons, whereas colons expressing KrasG13D showed only moderate hyperplasia, as recently described (26). Strikingly, KrasG12R mutation had no obvious effect on crypt length or tissue structure, closely resembling KrasWT tissue (Fig. 2A–C). It is notable that of each of the four Kras mutations assessed here, KRASG12R is the least frequently observed in colon cancer. Despite the clear impact on crypt height (Fig. 2C), none of the Kras mutations induced significant expansion the proliferative zone, quantified as the height of the region of BrdU incorporation (Fig. 2D). In contrast, G12D-, G12C-, and G13D-mutant colons all showed a significant expansion of KRT20-positive differentiated cells (Fig. 2E), together suggesting that increased crypt height may in part be due to persistence of committed epithelial cells. In all, these data show that just as mutation of different RAS family members has different effects on the colonic epithelium (23), subtle mutational changes within the same RAS gene (Kras) can dramatically alter the acute physiologic response in an in vivo setting.
In the mouse pancreas, induction of KrasG12D or KrasG12V mutations in the developing epithelium drives transdifferentiation of the acinar compartment [acinar to ductal metaplasia (ADM)] and the development of premalignant pancreatic intraepithelial neoplasias (PanIN; refs. 21, 22, 27). We analyzed the impact of each different Krasmut allele in a KrasLSL-mut/p48-Cre (hereafter, KC-MUT) model whereby Cre is expressed mid-gestation and activates KrasMUT expression in almost all epithelial cells of the pancreas (14, 28). These mice also carried a far-red mKate2 fluorescent reporter (29) to monitor Cre-mediated recombination (Supplementary Fig. S4A). By 4 weeks of age, changes were apparent in the pancreas of all genotypes, with evidence of ADM and early PanIN development (Supplementary Fig. S4A). Transcriptome analysis of whole pancreatic tissue at this time revealed gene signatures consistent with KRAS activation in the pancreas, including upregulation of genes involved in epithelial-to-mesenchymal transition (EMT), KRAS signaling, and inflammatory responses (Supplementary Fig. S4B; Supplementary Table S4). Overall, altered gene sets were similar between all mutants at this early time point (Supplementary Fig. S4B–S4C), although the increase in ductal (epithelial) and fibroblast markers, and corresponding decrease in the expression of genes marking acinar cells, was higher in KC-G12D and KC-G12C pancreata relative to KC-G12R and KC-G13D tissue (Supplementary Fig. S4D).
By 12 weeks of age, KC-G12D mice showed the expected appearance of PanIN lesions with loss of the acinar marker Carboxypeptidase A1 (CPA1), SOX9 induction (30), expression of the ductal lineage cytokeratin KRT19, and the production of mucins (Alcian Blue; Fig. 3A; Supplementary Fig. S5A and S5B). Accompanying this change was the infiltration of alpha smooth muscle actin (αSMA)–positive stromal cells (Fig. 3A) and ectopic deposition of extracellular matrix (Fig. 3A; Masson Trichrome, blue staining). We quantified phenotypic change by measuring the relative area of normal acini, ADM, and PanINs across each genotype at 12 and 50 to 60 weeks of age (Fig. 3B; Supplementary Fig. S5B–S5D). As expected, KC-G12D pancreata showed predominantly PanINs at 12 weeks (>90%), with little normal acinar tissue remaining. The overall PanIN burden was lower in KC-G12C mice (∼50%), but the remainder of the pancreas was ADM, with less than 10% of pancreatic area containing normal acini (Fig. 3A and B). In contrast, at 12 weeks, KC-G13D pancreata appeared predominantly histologically normal, with less than one third of the pancreas showing evidence of ADM or PanIN transition (Fig. 3B); however, those regions containing PanINs closely resembled KC-G12D or KC-G12C lesions, containing increased Alcian blue and KRT19 staining (Fig. 3A). Similar to KC-G12D epithelium, KC-G12R lesions displayed early responses to KRAS activation, including elevated SOX9 expression (Supplementary Fig. S5A), but in contrast showed almost no progression to PanIN (Fig. 3B; Supplementary Fig. S5C and S5D); KC-G12R pancreata had no evidence of Alcian Blue staining, whereas the epithelium showed expression of both acinar (CPA1) and ductal (KRT19) markers, suggesting a stalled progression at ADM (Fig. 3A and B). Lack of phenotypic progression was also evident in KC-G12R mice aged to 24 weeks and one year of age (Fig. 3B; Supplementary Fig. S6A–S6C), suggesting that the differences observed were not simply due to a moderate slowing of disease course. In contrast, KC-G13D mice showed more obvious time-dependent disease progression, and by 1 year, more than 50% of the pancreas was comprised of PanIN lesions.
At 12 weeks, pERK was robustly detected in KC-G12D and KC-G12C pancreata, but was also more prominently associated with PanIN lesions that were more commonly found in these genotypes. In whole-pancreas lysates at 4 weeks of age, ERK phosphorylation as well as RAS-GTP levels (Fig. 3C) were elevated in most strains compared to KC-WT tissue. In contrast to pERK, S6 phosphorylation was most strongly correlated with Kras-mutant ADM-like lesions, and thus was most apparent in KC-G13D and then in KC-G12R pancreas lysates (Fig. 3A and C). Together, these results show that KrasG12C, KrasG12R, and KrasG13D mutations have reduced disease-initiating capacity compared with the well-studied KrasG12D mouse model in the pancreas, and in some cases (G12R) do not follow the existing acini–ADM–PanIN progression model.
It is worth noting that mice that did not carry the Rosa26-targeted mKate2 reporter allele KrasLSL-mut/p48-Cre showed less penetrant ADM/PanIN phenotypes compared with those with mKate2 (Supplementary Fig. S7). The reason for this is not known, but the data suggest an effect of the rtTA3-IRES-mKate2 reporter or a Rosa26-linked phenotypic modifier. Importantly, this effect was seen across all genotypes (Supplementary Fig. S7), and in the case of KC-G12R and KC-G13D mice, dramatically reduced disease burden, and pancreata more closely resembled KrasWT mice.
Acute Pancreatitis Promotes Progression of KrasG12C and KrasG13D, but Not KrasG12R Preneoplastic Lesions
In patients, chronic pancreatitis substantially increases the risk of developing PDAC (31), and, similarly, induction of acute pancreatitis in mice by high doses of the cholecystokinin analogue cerulein promotes the early progression of disease (27, 32, 33). To determine whether acute pancreatitis would alter the progression of pancreatic precursors carrying different Kras mutations, we treated 9-week-old mice with cerulein (8 doses, 50 μg/kg, 1 hour apart) and assessed pancreatic response 20 days following injury (Fig. 4A). As expected, KrasWT mice showed full recovery of the acinar tissue, whereas KC-G12D mice contained almost no normal acinar tissue, with the majority of the pancreas made up of PanIN lesions (Fig. 4B and C). Both KC-G12C and KC-G13D mice showed a marked progression of disease, with cerulein-treated KC-G12C pancreata containing more than 90% PanINs, whereas KC-G13D animals progressed from >65% normal acinar tissue to >65% PanINs (Fig. 4B and C). Each of the cerulein-treated KC-G12D, KC-G12C, and KC-G13D mice showed a more severe histologic phenotype than untreated mice, with evidence of inflamed areas, atrophy, and the presence of cysts (Fig. 4B). Strikingly, cerulein-treated KC-G12R pancreata appeared similar to untreated KC-G12R mice, with the majority of tissue stalled at ADM, and less than 5% of the pancreas containing PanIN lesions (Fig. 4B and C).
Progression of KrasG12D-mutant preneoplastic lesions in the pancreas both before and following cerulein-induced pancreatitis requires the induction of a quiescent stem cell population marked by expression of Doublecortin-like kinase 1 (DCLK1; refs. 34, 35). To test whether KrasG12R-mutant pancreata had a selective defect in the induction of this stem population, we quantified the frequency of DCLK1-positive cells in untreated and cerulein-treated pancreata. Like KC-G12D, both KC-G12C and KC-G13D pancreata showed relatively abundant DCLK1-positive cells that were increased approximately 2-fold following cerulein treatment (Fig. 4D and E). In contrast, KC-G12R pancreata contained almost no DCLK1-positive cells, and although the number increased following pancreatitis, it remained lower than all other genotypes (Fig. 4D and E). Together, these data show that distinct Kras mutations have both a quantitative and qualitative impact on the premalignant transformation of the pancreatic epithelium. The specific failure of KC-G12R–mutant pancreatic epithelium to transition from metaplastic acini to PanIN lesions is unexpected, given the frequency of KRASG12R mutation in human pancreatic cancer, but maybe linked to a reduction in the DCLK1-positive regenerative stem cells that are important for disease progression in the pancreas (34).
Generation and Analysis of KP Organoid Models
To assess tumor cell–intrinsic differences between Kras mutations and explore the potential for these new strains as effective tools for testing therapeutic interventions, we derived ductal pancreatic organoids from each KrasLSL-mut mouse and induced simultaneous Cre-mediated KRAS activation and p53 disruption by CRISPR (hereafter, KP-MUT; Fig. 5A). Targeted deep sequencing of the Trp53 locus following Nutlin3 selection confirmed frameshift alterations in greater than 99.9% of all polyclonal organoid cultures (Supplementary Fig. S8A and S8B). We analyzed the transcriptional profile of each KP-MUT line compared with KP-WT organoids and identified a range of pathways activated in Kras-mutant cells (Supplementary Table S5), including MYC, p53, and AKT/MTOR signaling. In general, KP-G12R organoids showed less prominent KRAS transcriptional signatures (Fig. 5B), and a notable impairment of PI3K/AKT/mTORC1 signaling (Fig. 5B), consistent with recent reports in KRASG12R-mutant human PDAC cell lines (8). Similar to MEFs, RAS-GTP levels were elevated in all mutants, although moderately lower in KP-G13D organoids (Fig. 5C).
Pancreatic organoids from KPC tumors have been extensively characterized and reflect an accurate ex vivo surrogate for murine PDAC (35). Our KP-MUT organoids were generated ex vivo, from otherwise WT pancreatic ductal epithelium. To define the tumorigenic potential of the distinct KP-MUT models, we transplanted organoids from each genotype by direct orthotopic injection into the pancreas of recipient (Foxn1nu/nu) mice. For each genotype, we combined two independent organoid lines (1:1) to ensure that any lack of tumor growth was not simply due to an isolated effect in one biological replicate. As expected, KP-G12D and KP-G12C organoids formed tumors in 9 of 10 injected animals (5/5 and 4/5, respectively), each showing variability in tumor size at 8 weeks (Fig. 5D and E). In contrast to the defect in early-stage progression of premalignant lesions, but consistent with the observed frequency in human PDAC (2), KP-G12R organoids formed tumors in 5 of 5 injected animals, whereas no KP-G13D lines showed evidence of tumor growth (Fig. 5D and E). Together, these data suggest the pancreatic KP organoid lines represent a valuable preclinical platform to interrogate KRAS biology and possible therapeutic strategies.
G13D-Mutant Organoids Are Sensitive to EGFR Inhibition
Current clinical guidelines exclude all patients with KRAS-mutant tumors from treatment with small molecules or antibodies that target EGFR. However, retrospective clinical data (6, 7) and cell line analyses (36, 37) have suggested that colorectal cancers carrying KrasG13D mutations may be sensitive to the EGFR-targeted antibody cetuximab. Although KRASG13D mutations are infrequent in PDAC, we decided to ask whether a similar genotype–phenotype relationship would hold true in pancreatic organoids. Cetuximab does not recognize mouse EGFR, so we treated KP-MUT organoids with the cross-reactive small-molecule EGFR inhibitor gefitinib. Although KP-G12D organoids showed an acute molecular response to gefitinib treatment, including decreased phosphorylation of the downstream effectors ERK1/2 and AKT (Fig. 5F), and reduced EdU incorporation at high doses (Fig. 5G), they could be maintained over multiple passages in the presence of drug (Fig. 5H). In contrast, KP-G13D organoids showed a profound cell-cycle arrest within 48 hours of gefitinib treatment (Fig. 5G), even at intermediate doses, and could not survive long-term (>1 week) drug exposure (Fig. 5H). KP-G12C and KP-G12R organoids had an intermediate response to gefitinib but ultimately were able to maintain growth and proliferation with extended treatment (Supplementary Fig. S9A and S9B).
The response of KrasG13D-mutant cells to EGFR inhibition implies that they rely on additional pathway activation to achieve robust downstream mitogenic signaling. Although KRAS mutations are rarely observed with other MAPK-activating alterations, such as mutational activation of EGFR (38) or loss of negative regulators such as Neurofibromin 1 (NF1; ref. 39; Supplementary Fig. S10A and S10B), recent work has identified a subset of BRAF mutations (Class III) that require upstream RAS signaling to induce high levels of MAPK activation. Interestingly, class III BRAF mutations often co-occur with other RAS–MAPK pathway mutations (39). Similarly, our analysis of more than 55,000 cancers in the Project GENIE database (40) shows that tumors carrying KRAS codon 13 mutations are four times more likely to have additional driver mutations in MAPK pathway genes (Supplementary Fig. S10B). In particular, KRASG13D- or KRASG13C-mutant cancers showed 5-fold enrichment in truncating NF1 mutations (Supplementary Fig. S10B), suggesting that these mutant proteins may be subject to upstream RTK/RAS and/or NF1 regulation.
To test this hypothesis directly, we silenced NF1 in KP-G12D and KP-G13D pancreatic organoids using lentiviral-transduced miRNA-based shRNAs (Fig. 5I) and treated them with gefitinib for 3 days (Fig. 5J). KP-G13D organoids transduced with a control shRNA (shRen.713) remained sensitive to EGFR inhibition whereas NF1-silenced organoids showed elevated pERK and continued expansion in the presence of drug (Fig. 5J and K). As expected, KP-G12D organoids showed no change in growth following EGFR inhibition, regardless of NF1 expression (Fig. 5J).
EGFR Inhibition Reveals Sensitivity to Direct KRASG12C Inhibition
The recent development of covalent inhibitors of KRASG12C represents the first strategy to directly target oncogenic KRAS, and multiple small molecules have shown promise in early-stage clinical trials (41, 42). To determine whether our KrasLSL-G12C model is an effective preclinical tool to investigate response and resistance to clinical KRASG12C inhibitors, we treated KP-MUT organoids with ARS1620, a selective KRASG12C inhibitor that covalently binds to Cys12 when KRAS is in its GDP-bound inactive state (43). Surprisingly, KP-G12C organoids were completely insensitive to treatment with the KRASG12C inhibitor alone, having no effect on morphology (Fig. 6A), proliferation (Fig. 6B), or organoid size (Fig. 6C; Supplementary Fig. S11). Although ARS1620 reduced RAS-GTP levels selectively in KP-G12C and not KP-G12D organoids (Fig. 6D), it did not have a major impact on downstream signaling (Fig. 6E), perhaps due to endogenous receptor–mediated pathway activation. Consistent with a role for upstream mitogen signaling driving resistance to G12C inhibition, combined treatment with gefitinib induced a further reduction in RAS-GTP levels (Fig. 6D), a profound block in proliferation within 48 hours (Fig. 6B), and eliminated all organoids within 6 days of treatment (Fig. 6A–C). These data indicate that upstream signaling by RTKs can affect the outcome of downstream KRASG12C inhibition, in line with recent reports in cell lines and PDX models (42, 44, 45). Together, these findings highlight the fidelity of which the KrasLSL-G13D and KrasLSL-G12C alleles recapitulate key signaling and feedback regulation observed in human cancer cells.
Mutant KRAS is a clear driver of human cancer, and consequently there are numerous approved and investigational drugs in clinical use that target KRAS or upstream/downstream signaling mediators. Owing to expansive tumor-sequencing efforts, we now know the types and frequency of KRAS alterations in different cancer types in great detail. In contrast, exactly how each distinct KRAS mutation affects disease initiation and progression remains largely unknown. Defining the similarities and differences in cancer-associated KRAS mutations may reveal unique dependencies that can be exploited therapeutically. Here, we report the CRISPR-mediated generation of a series of new conditional KrasLSL-mut mouse models that recapitulate the tissue-restricted alterations seen in human colorectal, pancreas, and lung cancers. Using these new preclinical tools, we show that subtle changes in mutations at codon 12 and 13 have a dramatic impact on tumor initiation and premalignant progression in the colon and pancreas, and that pancreas-derived organoids carrying particular KRAS alterations are differentially sensitive to targeted therapies.
The biochemical properties of individual KRAS mutant proteins have been well characterized in in vitro systems, and show specific differences in intrinsic or GAP-mediated GTP hydrolysis and/or nucleotide exchange (4). The phenotypic consequences for each mutation have been much more difficult to define. Recently, Winters and colleagues used a multiplexed adeno-associated virus approach to engineer endogenous Kras mutations in the lung and pancreas of recipient mice (46). These experiments concluded that in multiple sensitized backgrounds, KrasG12D, KrasG12R, and KrasG13R mutations readily induce tumor growth, whereas KrasG12C is a comparatively weak transforming allele. Our data assessing premalignant transformation in the colon and pancreas, and in KP organoids and MEFs, suggest that KRASG12C is quite a potent KRAS mutant, whereas KRASG12R drives a less robust KRAS phenotype. There are notable differences between these studies, including the timing of KRAS activation, presence/absence of coaltered tumor suppressors, and tissue context. One important technical consideration that may have relevant biological consequences is that, using CRISPR-based HDR to engineer Kras mutations, Winters and colleagues invariably introduced disruptive mutations in the second Kras WT allele (46), as we observed during the first iteration of targeting ESCs, whereas our Cre-driven models retain the expression of WT KRAS. Indeed, in some contexts, WT KRAS acts as a tumor suppressor (47) and can influence the types of KRAS mutations that occur following chemical carcinogenesis (48). The role of WT KRAS protein is a poorly understood but important question in cancer biology, as up to 50% of all KRAS-mutant cancers show allelic imbalance involving amplification of the mutant gene or loss of the WT copy (49). In this regard, our new mouse alleles offer a controlled setting to study the impact of WT KRAS, by exploiting the silent mutations introduced to the mutant alleles (Fig. 1B) and selectively targeting the WT allele by CRISPR or RNAi-based strategies.
We show that subtle mutational differences in KRAS can dramatically alter the acute cellular response in an in vivo setting in multiple tissues. We noted striking differences among KRAS mutants in the hyperplastic response of the colonic epithelium as well as early transformation steps of the pancreatic acini. In particular, KrasG12R mutations had no measurable effect in the colon, and caused a partial acinar-to-ductal transition in the pancreas, but were unable to induce PanIN transformation. Even in the presence of inflammatory stimuli during acute pancreatitis, G12R mutants failed to induce obvious progression to PanINs. KrasG12R pancreata showed a significantly reduced frequency of DCLK1+ progenitor cells that have been linked to pancreatic cell transformation (34). This reduction in a known disease-initiating cell population may underlie the unusual difference observed in G12R mutants, but it remains unclear whether this is a cause or consequence of the stalled transition.
The observation that KrasG12R shows differences in the early steps of premalignant progression in the pancreas is surprising given the relatively high frequency of KRASG12R mutations in human PDAC. However, despite the atypical primary response to KrasG12R induction, KrasG12R mutations, but not KrasG13D mutations, were sufficient to drive tumor growth in orthotopically transplanted pancreatic organoids in the absence of p53. This suggests that KRASG12R is a potent oncogenic allele in PDAC, but the cellular response to this event is context-dependent. So, what could explain the altered progression in the premalignant state? First, the vast majority of our understanding of pancreatic cancer initiation has been driven by studies in the KrasLSL-G12D mouse and it is possible that not all KRAS mutants take the same path through tumor initiation. Second, it is also possible that the KrasG12R-mutant cells lack a key signal to drive PanIN development. Although KrasG12R lesions appear to have less ERK 1/2 phosphorylation, this is more likely linked to the lack of ductal cells where pERK is highest (Fig. 2). Hobbs and colleagues recently revealed that KrasG12R-mutant cells have reduced AKT/PI3K activation and macropinocytosis (8). Both PI3K signaling and macropinocytosis have been directly linked to tumor initiation and/or progression in KRAS-driven pancreas cancer (50–52), providing a potential mechanism for this atypical in vivo response. Indeed, we noted diminished AKT phosphorylation and lower PI3K/AKT/mTOR transcriptional signatures in KrasG12R-mutant MEFs and organoids (Figs. 1 and 5). This was less obvious in whole pancreas lysates, and perhaps indicates a need for more focused, dynamic, and epithelial cell–specific investigation of the in vivo response. As has been described for KrasG12D models, disruption of p53 enabled tumorigenic progression of KrasG12R-mutant cells, although it is unclear whether p53 is involved in the altered initial response of KrasG12R-mutant cells in situ. Whatever the underlying cause of restricted progression to the ductal state in KrasLSL-G12R mice, it is important to note that organoid culture conditions select for ductal-like cells, thus bypassing the stalled metaplastic progression observed in KC-G12R mice. As is clear from the data presented here, furthering our understanding of KRAS biology will require consideration of not only the specific oncogenic KRAS mutation, but also the cell context (cells vs. organoids vs. tissue) as well as co-occurring genetic alterations. These qualitative and quantitative differences in each setting will make it difficult to extrapolate general principles from any single experimental system.
To explore the utility of the KrasLSL-mut strains as preclinical tools, we generated KP pancreatic organoids and systematically tested two different targeted therapies. The response of KrasG13D mutants to EGFR inhibition that we observed in pancreatic organoids (Fig. 5) is consistent with retrospective clinical data in colorectal cancer (6, 7), and two recent publications using cancer cell lines (36, 37). Similarly, these studies, like ours, show that NF1 is an important regulator of MAPK signaling output in KRASG13D-mutant cells, although the exact mechanism remains controversial. These observations parallel similar findings in BRAF-mutant cells, where Rosen and colleagues identified distinct classes of oncogenic BRAF mutations based on their signaling dependencies (39). Class III BRAF mutations, like KRASG13D mutations, more frequently co-occur with mutations in additional MAPK pathway genes, and human cancers carrying these alterations are more sensitive to EGFR inhibition (53). Similarly, KRASA146T mutations commonly co-occur with MAPK pathway mutations in colorectal cancer, and Poulin and colleagues recently showed, using a similar Cre-conditional Kras approach, that KrasA146T mutations are poorly transforming in the colon and pancreas, like KrasG13D (25, 26). It is possible that the low oncogenicity of KrasG13D mutations in the pancreas may be driven by an increased reliance on EGFR-mediated signaling, which has a dose-dependent impact on PDAC development, even in the presence of p53 mutations (54, 55). Together, mutations such as KRASG13D and KRASA146T likely represent a distinct class of KRAS alterations that exert differential oncogenic effects and may confer sensitivity to existing clinical therapies.
Finally, we show that KP organoids carrying an endogenous KrasG12C mutation are sensitive to a recently described covalent G12C inhibitor, but that response to this drug it is effective only in the presence of an EGFR inhibitor. Similar synergistic effects of G12C and EGFR inhibitors have been noted in human cancer cell lines (42, 44, 45), potentially due to reduced SOS-dependent GDP–GTP exchange, increasing the GDP-bound pool of KRASG12C and rendering it more vulnerable to covalent modification (43, 44, 56). It may also be possible that the presence of WT KRAS in these cells allows escape from targeted G12C inhibition, and that this activity is RTK-dependent. Further work in genetically defined systems such as these will provide a complete picture of the signaling and phenotypic consequences of clinical KRASG12C inhibitors, to further guide clinical application of these exciting small molecules.
Together, these data describe the development of three new broadly useful precision oncology models and highlight unique downstream consequences of subtle and cancer-relevant changes in KRAS mutations. The models faithfully represent the signaling dynamics and therapeutic response of human cancer cells, and we expect they will serve as valuable immunocompetent preclinical tools to understand KRAS biology and develop more effective treatment strategies for KRAS-driven cancers.
VP12 vector expressing spCas9-HF1 (15) was codon optimized (16) and renamed HF1*. Sequences encoding Kras sgRNAs (Supplementary Table S6) were cloned into BbsI site of pX458, VP12-U6 and pXHF1* vectors. P53c sgRNA (Supplementary Table S6) was cloned into BsmBI site of Lenti-Cas9-Cre vector. For shRNA cloning, NF1 and Renilla 713 shRNAs (Supplementary Table S6) were cloned into XhoI/EcoRI site of SGEN vector (57).
The p48 ESC line, derived previously as described elsewhere (14), was used to generate the new conditional LSL-Kras strains. ESCs were cultured in KOSR + 2i media (58) on irradiated MEF feeder layers. 2 × 105 cells were cotransfected with 2 μg of the Cas9/sgRNA vector PxHFc and 4 μL of single-stranded oligonucleotide HDR template (20 μmol/L) using a Lonza X Unit Nucleofector with P3 buffer kit (Lonza #V4XP-3032). Four days after transfection, cells were plated at low density (500 cells) to enable clonal growth, and the remaining culture was used to assess targeting efficiency from bulk population (see below). Clones were picked when they were visible without microscope. PCR amplification following digest to confirm template integration were carried out. Digested products were analyzed by QIAxcel (Qiagen). Positive clones were expanded and further validated by allele-specific PCR and Sanger sequencing before sending them to perform blastocyst injection.
Estimation of Expected Mutation Frequency from Published Data.
Cervantes and colleagues calculated a mutation rate of approximately 8 × 10−8 mutations per cell, per generation (17). As genome size is 3 × 109 bp, we estimate 3.75 mutations per genome/cell division. It took approximately 5 weeks to target, clone, and expand cells for cryopreservation. Mouse ESCs doubled every 4 to 5 hours (59), approximately 5.3 generations/day. So, 35 days of culture × 5.3 generations/day = 185.5 generations. 185.5 generations × 3.75 mutations/cell/generation = 696 mutations/cell. Because we performed whole- exome sequencing that covers approximately 1.5% of whole genome: 696 mutations/cell × 0.015 genome coverage = 10.4 mutations/clone.
Clones were picked and trypsinized. Half of the volume was mixed with 2× DNA lysis buffer (20 mmol/L Tris, pH 8.8, 40 mmol/L (NH4)2SO4, 20 mmol/L MgCl2, 10% Triton X 100, proteinase K 800 ug/mL, β-ME) and incubated for 2 hours at 55°C following 20 minutes at 95°C to inactivate the Proteinase K. The remaining half was resuspended in media and transferred to a 48-well plate already containing 500 μL of ESC media.
The region of interest was amplified using 1 μL of the crude gDNA lysis in 16 μL volume using Promega 2X PCR master mix. HDR targeting was confirmed in each clone by digesting for 2 hours half of the PCR product (8 μL) with the specific restriction enzyme for each integrated template (Fig. 1B).
Genomic DNA Isolation, and T7 Assay
ESCs were lysed in genomic lysis buffer (10 mmol/L Tris, pH 7.5, 10 mmol/L EDTA, 0.5% SDS, and 400 μg/mL proteinase K) for at least 2 hours at 55°C. After proteinase K heat inactivation at 95°C for 15 minutes, 0.5 volume of 5 mol/L NaCl was added, and samples were centrifuged for 10 minutes at 15,000 rpm. Supernatants were mixed with one volume of isopropanol, and DNA precipitates were washed in 70% EtOH before resuspension in 10 mmol/L Tris, pH 8.0. Cas9-induced mutations were detected using the T7 endonuclease I. Briefly, the target region surrounding the expected mutation site was PCR-amplified using Herculase II (600675, Agilent Technologies). PCR products were column-purified (Qiagen) and subjected to a series of melt–anneal temperature cycles with annealing temperatures gradually lowered in each successive cycle. T7 endonuclease I was then added to selectively digest heteroduplex DNA. Digested products were visualized on a 2.5% agarose gel.
DNA-Library Preparation and MiSeq
Deep sequencing was performed on Clones Cas9-only transfected or cotransfected alongside an HDR template and having successfully integrated it. Briefly, DNA-library preparation and sequencing reactions were conducted at GENEWIZ. A NEB NextUltra DNA Library Preparation Kit was used according to the manufacturer's recommendations (Illumina). Adaptor-ligated DNA was indexed and enriched through limited-cycle PCR. The DNA library was validated with a TapeStation (Agilent) and was quantified with a Qubit 2.0 fluorometer. The DNA library was quantified through real-time PCR (Applied Biosystems). The DNA library was loaded on an Illumina MiSeq instrument according to the manufacturer's instructions (Illumina). Sequencing was performed with a 2 × 150 paired-end configuration. Image analysis and base calling were conducted in MiSeq Control Software on a MiSeq instrument and verified independently with a custom workflow in Geneious R11.
For virus production, HEK293T cells (ATCC CRL-3216) were plated in a 10-cm plate and transfected 12 hours later (at 95% confluence) with a prepared mix in DMEM (with no supplements) containing 15 μg of lentiviral backbone LCMCp53c (pLenti-U6-p53c-sgRNA-Cas9-p2A-Cre), 7.5 μg of PAX2, 3.75 μg of VSV-G, and 78 μL of polyethylenimine (1 mg/mL). Thirty-six hours after transfection, the medium was replaced with target cell collection medium and supernatants were harvested every 8 to 12 hours up to 72 hours after transfection.
Production of mice and all treatments described were approved by the Institutional Animal Care and Use Committee (IACUC) at Weill Cornell Medicine (New York, NY), under protocol number 2014-0038. ESC-derived mice were produced by blastocyst injection, and animals were either maintained on a mixed C57B6/129 background for experimental breeding or back-crossed to C57BL/6N mice. All LSL-Kras strains are available from Jackson Labs [G12C: B6N.129S4-Krasem1Ldow/J (#033068); G12R: B6N.129S4-Krasem2Ldow/J (#033316); G13D: B6N.129S4-Krasem3Ldow/J (#033317)]. Progeny of both sexes were used for experiments and were genotyped for specific alleles (KrasLSL-G12D, KrasLSL-G12C, KrasLSL-G12R, KrasLSL-G13D, Ptf1/p48-Cre, Rosa26-LSL-tdTomato, CAGS-LSL-RIK) using primers described in Supplementary Table S6 and protocols available at www.dowlab.org/Protocols. Production of mice and all treatments described were approved by the IACUC at Weill Cornell Medicine under protocol number 2014-0038. For experiments analyzing the colon, all mice were collected at 8 weeks of age and injected with BrdU (1 mg/mouse) 2 hours prior to harvest by intraperitoneal injection. Fabp1-Cre and KrasLSL-mut age-matched littermates were used as controls. To induce experimental pancreatitis, 9-week-old mice were subjected to 8 intraperitoneal injections of cerulein (50 μg/kg) once every 1 hour (60). Mice were monitored daily and euthanized 20 days after the acute treatment.
KrasLSL-mut/LSL-tdTomato males were bred with C57BL/6N females. MEFs were derived from day 12.5–14.5 postcoitum embryos following previously described protocol (61). Cells were cultured and expanded for one passage in DMEM (Corning) supplemented with 10% (v/v) FBS, and frozen at passage 2.
MEFs were immortalized by using a vector encoding Cas9 and a p53c sgRNA, as well as Cre recombinase to be able to activate Kras. Cells were thawed and immediately transduced with viral supernatants (1:2) in the presence of polybrene (8 μg/μL). Two days after transduction, cells were selected in Nutlin-3 (10 μmol/L). Established MEF cell lines expressing different Kras mutants (G12D, G12C, G12R, and G13D) or KrasWT and Trp53 loss were consequently used to perform RNA-seq and Western blot analysis. For protein experiments, MEFs were starved overnight (2% FBS DMEM) and then stimulated with EGF 20 ng/mL for 10 minutes (62). All MEF data were obtained using at least three independent MEFs/genotype.
Fluorescence Competitive Proliferation Assays.
MEFs were thawed and after one passage infected with adenovirus-Cre purchased from the University of Iowa (Iowa City, IA; 5 × 107 plaque-forming units/1 × 106 cells). One day after infection, the percentage of tdTomato-positive cells was measured by flow cytometry (Attune NxT Flow Cytometer, Thermo Fisher Scientific) and cells were mixed at defined proportions with their respective parental cells. tdTomato fluorescence was then tracked every 5 days by flow cytometry.
Ras-GTP levels were assessed by Active Ras Pull-Down and Detection Kit (Thermo Fisher Scientific, catalog no. 16117Y) using RAF-RBD fused to GST to bind active (GTP-bound) RAS. Protein lysates (300—500 μg) were incubated with 30 μL glutathione resin and GST protein-binding domains for 1 hour at 4°C to capture active small GTPases according to the manufacturer's protocol. After washing, the bound GTPase was recovered by eluting the GST-fusion protein from the glutathione resin. The purified GTPase was detected by Western blot analysis using mouse monoclonal anti-KRAS provided by the kit.
Murine Pancreatic Ductal Organoid Culture
Isolation of normal pancreatic ducts was done modifying previously described protocol (63). Briefly, pancreata were minced and washed in Hank's Balanced Salt Solution (Corning), and then incubated for 30 minutes at 37°C with Collagenase V to release the ducts. After washing twice with DMEM/10% FBS media, ducts were resuspended in basal media [advanced DMEM/F12 (Corning) containing 1% penicillin/streptomycin, 1% glutamine, 1.25 mmol/L N-acetylcysteine (Sigma Aldrich A9165-SG), and B27 Supplement (Gibco)], and mixed 1:10 with growth factor–reduced (GFR) Matrigel (BD Biosciences). Forty microliters of the resuspension was plated per well in a 48-well plate and placed in a 37°C incubator to polymerize for 10 minutes. To culture ductal pancreatic organoids, the basal media described above was supplemented with 10 nmol/L Gastrin (Sigma), 50 ng/mL EGF (PeproTech), 10% RSPO1-conditioned media, 100 ng/mL Noggin (PeproTech), 100 ng/mL FGF10 (PeproTech), and 10 mmol/L nicotinamide [Sigma; note: culture freshly isolated organoids in pancreatic organoid media (POM) containing 10 μmol/L Rock inhibitor (Y2732) during 48–72 hours]. For subculture and maintenance, media were changed on organoids every two days and they were passaged 1:3 every 5 days. To passage, the growth media was removed and the Matrigel was resuspended in cold basal media and transferred to a 15-mL Falcon tube. Organoids were mechanically disassociated using a P1000 and pipetting 40 times. Five milliliters of cold PBS were added to the tube and cells were then centrifuged at 1,200 rpm for 5 minutes and the supernatant was aspirated. Cells were then resuspended in GFR Matrigel and replated as above. For freezing, after spinning, the cells were resuspended in complete media containing 10% FBS and 10% DMSO and stored in liquid nitrogen indefinitely.
To generate KP organoids (Krasmut/Trp53 loss), normal pancreatic organoids were cultured in transduction media [POM containing CHIR99021 (5 μmol/L) and Y-27632 (10 μmol/L)] for 2 days prior to transduction. Single-cell suspensions were produced by dissociating organoids with TrypLE Express (Invitrogen, #12604) for 5 minutes at 37°C. After trypsinization, cell clusters were resuspended in 400 μL of transduction media containing concentrated lentiviral particles in the presence of polybrene (8 μg/μL) and transferred into a 48-well culture plate. The plate was centrifuged at 600 × g at 32°C for 60 minutes, followed by another 4-hour incubation at 37°C. Cell clusters were spun down and plated in Matrigel.
Orthotopic Pancreatic Organoid Transplantation
Two different KP lines carrying the same Kras mutation (G12D or G12C or G12R or G13D) or Kras WT lines were cultured over 3 days in basal media, then mixed at 1:1 ratio and injected (total volume 25 μL: 12.5 μL organoid in basal media/12.5 μL Matrigel) in the pancreata of nude (Foxn1nu) mice.
Organoid Drug Treatment
Organoids were plated in 120 μL Matrigel (3 × 40 μL droplets) in one 12-well plate and cultured in basal media with either DMSO or gefitinib (1 μmol/L) or ARS1620 (1 μmol/L) or gefitinib/ARS1620 combined. Organoids were passaged 1:3 every 72 hours and then cultured again in DMSO, gefitinib, ARS1620, or gefitinib/ARS1620 combination.
EdU Flow Cytometry and Imaging in Organoids
Organoid EdU flow cytometry was performed using the Click-iT Plus EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Thermo Fisher Scientific, # C10634). Pancreatic organoids were first incubated with 10 μmol/L EdU for 4 hours at 37°C. One well of a 12-well plate was broken up by pipetting vigorously 50 times in 1 mL PBS, then diluted in 5 mL of PBS. Cells were pelleted at 1,100 rpm for 4 minutes at 4°C, then resuspended in 50 μL TrypLE and incubated at 37°C for 5 minutes. Five milliliters of PBS were then added to inactivate the TrypLE, and cells were pelleted. Cells were resuspended in 250 μL of 1% BSA in PBS, transferred to a 1.7-mL tube, and then pelleted at 3,000 rpm for 4 minutes. Cells were then resuspended in 100 μL Click-iT fixative, and processed as instructed in the Click-iT Plus EdU protocol (starting with Step 4.3). Wash and reaction volumes were 250 μL.
RNA Isolation, cDNA Synthesis, and qPCR
To isolate RNA from MEFs and pancreatic organoids, we used TRIzol (Thermo Fisher Scientific, #15596018) according to the manufacturer's instructions, and contaminating DNA was removed by DNase treatment for 10 minutes and column purification (Qiagen RNeasy #74106). Pancreas tissue portion for RNA purification was consistently collected from the tail of the organ and immediately cut into smaller pieces and immersed in RNAlater stabilization solution (Thermo Fisher Scientific) and incubated at 4°C overnight before storing the sample at -80°C until RNA extraction was performed. Samples were homogenized using a Tissue Master 125 (Omni) and RNA purified using the RNeasy Kit (Qiagen).
Total RNA was isolated using TRIzol, DNAse treated, and purified using the RNeasy Mini Kit (Qiagen). Following RNA isolation, total RNA integrity was checked using a 2100 Bioanalyzer (Agilent Technologies). RNA concentrations were measured using the NanoDrop system (Thermo Fisher Scientific, Inc.). Preparation of RNA sample library and RNA-seq were performed by the Genomics Core Laboratory at Weill Cornell Medicine. Messenger RNA was prepared using TruSeq Stranded mRNA Sample Library Preparation kit (Illumina), according to the manufacturer's instructions. The normalized cDNA libraries were pooled and sequenced on Illumina NextSeq500 sequencer with single-end 75 cycles.
Raw FASTQ files were mapped to mouse reference GRCm38 using STAR two-pass alignment (v2.4.1d; default parameters; ref. 64), and transcript abundance estimates were performed using Kallisto (65), aligned to the same (GRCm38) reference genome. Kallisto transcript count data for each sample were concatenated, and transcript per million (TPM) data were reported for each gene after mapping gene symbols to ensemble IDs using R packages “tximport,” tximportData,” “ensembldb,” and “EnsDb.Mmusculus.v79.” Differential gene expression was estimated using DESeq2 (66). For data visualization and gene ranking, log fold changes were adjusted using lfcShrink in DESeq2, to minimize the effect size of poorly expressed genes. Gene set enrichment analysis (v3.0) was performed on preranked gene sets from differential expression between control and treated groups. We used R (v3.6.1) and R Studio (v1.2.1335) to create all visualizations and perform hierarchical clustering and principal component analysis. Volcano plots, heat maps, and other visualizations were produced using the following software packages: Enhanced Volcano (https://bioconductor.org/packages/devel/bioc/html/EnhancedVolcano.html), pheatmap (https://cran.r-project.org/web/packages/pheatmap/index.html), and ggplot2 (https://cran.r-project.org/web/packages/ggplot2/index.html).
Each gDNA sample based on Qubit quantification was mechanically fragmented on a Covaris E220 focused ultrasonicator (Covaris). Two hundred nanograms of sheared gDNA were used to perform end repair, A-tailing, and adapter ligation with Agilent SureSelect XT (Agilent Technologies) Library Preparation Kit following the manufacturer's instructions. Then, the libraries were captured using Agilent SureSelectXT Mouse All Exon probes and amplified. The quality and quantities of the final libraries were checked by Agilent 2100 Bioanalyzer and Invitrogen Qubit 4.0 Fluorometer (Thermo Fisher Scientific). Libraries were pooled at 8 samples per lane and sequenced on an Illumina HiSeq 4000 sequencer (Illumina Inc) at PE 2 × 100 cycles. Copy-number alterations were identified and plotted using cnvkit (v0.9.6) and single-nucleotide variants were called using MuTect2.
Whole pancreas samples were minced and lysed in 600 μL of RIPA buffer by homogenization on the TissueLyser II (Qiagen). Pancreatic organoids were grown in 120 μL of Matrigel in one well of a 12-well dish. Organoids were then recovered from the Matrigel using Cell Recovery Solution. Organoid pellets were lysed in 30 μL RIPA buffer. MEFs were cultured in 6-well plates and lysed in 150 μL of RIPA buffer. Antibodies used for Western blot analysis were: anti-actin-HRP (Abcam #ab49900), anti-α-Tubulin (Millipore Sigma #CP06), anti-pERK 44/42 (Cell Signaling Technology #4370), anti-ERK 44/42 (Cell Signaling Technology #9107), anti-pAKT (ser 473; Cell Signaling Technology #4060), anti-AKT (Cell Signaling Technology #4691), pMEK (Cell Signaling Technology #9154), MEK (Cell Signaling Technology #8727), pS6 (Cell Signaling Technology #4858), S6 (Cell Signaling Technology #2317), and anti-NF1 (Cell Signaling Technology #14623).
Immunofluorescence and IHC
Tissue, fixed in freshly prepared 4% paraformaldehyde for 24 hours, was embedded in paraffin and sectioned by IDEXX RADIL. Sections were rehydrated and unmasked (antigen retrieval) by heat treatment for 10 minutes in a pressure cooker in 10 mmol/L Tris/1 mmol/L EDTA buffer (pH 9) containing 0.05% Tween 20. For IHC, sections were treated with 3% H2O2 for 10 minutes and blocked in TBS/0.1% Triton X-100 containing 1% BSA. For immunofluorescence, sections were not treated with peroxidase. Primary antibodies, incubated at 4°C overnight in blocking buffer, were rabbit anti-CK19 (1:400, Abcam #ab133496), rabbit anti-DCLK1 (1:400, Abcam #109029), goat anti-CPA1 (1:400, R&D Systems AF2765), rabbit anti-αSMA (1:400, Abcam #ab5694), rabbit anti-pERK 44/42 (1:1,000, Cell Signaling Technology #9101), rabbit anti-pS6 (1:1,000, Cell Signaling Technology #2211), rabbit anti-SOX9 (1:1,000, Millipore #AB5535), rabbit anti-tRFP (1:2,000, Evogren #AB233), rat anti-BRDU (1:200 #ab6326), rabbit anti-KRT20 (1:200, Cell Signaling Technology #13063). For IHC, sections were incubated with anti-rabbit ImmPRESS HRP-conjugated secondary antibodies (Vector Laboratories, #MP7401) and chromagen development was performed using ImmPact DAB (Vector Laboratories, #SK4105). Stained slides were counterstained with Harris hematoxylin and mounted with Mowiol mounting media. For immunofluorescence stains, secondary antibodies were applied in TBS for 1 hour at room temperature in the dark, washed twice with TBS, counterstained for 5 minutes with DAPI and mounted in ProLong Gold (Life Technologies, #P36930). Secondary antibodies used were donkey anti-rabbit 594 (1:500, Invitrogen #A21207) and donkey anti-goat 488 (1:500, Invitrogen #A11055). Masson Trichrome staining was performed by IDEXX Radil. Images of fluorescent and IHC-stained sections were acquired on a Zeiss Axioscope Imager (chromogenic stains), Nikon Eclipse T1 microscope (IF stains). Raw .tif files were processed using FIJI (Image J) and/or Photoshop (Adobe Systems) to create stacks, adjust levels, and/or apply false coloring. Crypt length and proliferative zone were quantified in FIJI (Image J) using the “Measure” feature.
Raw exomeSeq and RNA-seq data have been deposited in the sequence read archive (SRA) under accession PRJNA578549.
Disclosure of Potential Conflicts of Interest
L.E. Dow reports grants from NCI during the conduct of the study and personal fees from Mirimus Inc. (scientific advisory board) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.
M.P. Zafra: Conceptualization, software, investigation, methodology, writing-original draft, writing-review and editing. M. Parsons: Methodology, writing-review and editing. J. Kim: Methodology. D. Alonso-Curbelo: Methodology. S. Goswami: Resources. E. Schatoff: Resources, software. T. Han: Methodology. A. Katti: Methodology. M. Calvo Fernandez: Methodology. J. Wilkinson: Formal analysis. E. Piskounova: Investigation. L.E. Dow: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing-original draft, writing-review and editing.
We thank Miguel Foronda for technical and experimental advice. This work was supported by a project grant from the NIH/NCI under award R01CA195787. We thank the Weill Cornell Genomics Resource Core Facility who performed library preparation and sequencing for WES and RNA-seq. M.P. Zafra is supported in part by NCI Grant NIH T32 CA203702. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.