Development and Histopathological Characterization of Tumorgraft Models of Pancreatic Ductal Adenocarcinoma

Ethics Statement: Human subjects

This study included human subjects and all procedures were approved by the University of Alabama at Birmingham Institutional Review Board (IRB approved protocol number: X10xxx8006) in accordance with the guiding ethical principles of the IRB-respect for persons, beneficence and justice, as embodied in the Belmont Report.  Written informed consent was obtained from all human participants after discussions of the procedures and potential risks and benefits prior to study participation. Written informed consent was obtained for use of these samples for this specific research purpose only.  No minors/children were included as participants. Tumor tissue received for implantation into mice was deemed in excess of that needed for standard of care. The consent procedure was approved by the Institutional Review Board committee of the University of Alabama at Birmingham.

Ethics Statement: Animal protocols

Animal studies were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee (IACUC) and were carried out in accordance with Animal Protocol Number (APN): 1x1009186. Four- to six-week old female CB17^-/- SCID mice were purchased from Taconic farms (Germantown, NY) and used as hosts for tumorgraft production. All animals were housed in the AAALAC accredited vivarium at UAB Research Support Building under barrier conditions with 12 hour light/dark cycles and ad libitum access to food and water. All mice were monitored for tumor growth daily, and tumors were measured twice a week. Mice were euthanized as soon as animals appeared to be in distress or discomfort. Bedding, chow, and cages were autoclaved and cages changed twice a week.

Tissue procurement

Pancreatic tumor tissue and normal pancreatic tissue were collected from patients who were undergoing surgical resection for pancreatic cancer at the University of Alabama at Birmingham Hospital (Birmingham, AL), and who had given prior consent per IRB-Approved Protocol #10xxx8006 as described in Ethics Statement. Immediately after resection (within an hour), tumor specimens not needed for diagnostic purposes were placed in M-199 medium (Lonza, Walkersville, MD) supplemented with penicillin G/ streptomycin (Pen/Strep, 50 units/mL; Gibco, Grand Island, NY) for transport to the Department of Pathology where tissue samples were grossly evaluated by a board certified pathologist (LNC) with a specialty in gastrointestinal malignancies, to confirm tissue viability and suitability for implantation.

Establishment of first generation (F1) tumorgrafts

Tumor tissue (primary tumor, F0 generation) was washed three times in M-199 media containing penicillin/streptomycin (Pen/Strep, 50 units/mL; Gibco, Grand Island, NY) and the entire specimen was dissected into fragments roughly 5 x 5 x 5 mm in size. Mice were anesthetized with 3% isoflurane (Animal Resources Program, Birmingham, AL) and incision sites were disinfected with cholorohexidine. Using surgical scissors, an incision was made 10 mm above the base of the tail, approximately 5 mm in length. A blunt dissection was then made with scissors over the right flank, creating a pocket in which to place the tumor specimen. A single tumor fragment was implanted into right flank of each of three mice. The incision site was treated with 2 drops of Pen/Strep and closure was completed with Vetbond™ (3M, St. Paul, MN). Mice were placed on a mat warmed to 37°C until they had recovered from anesthesia. Mice were monitored daily for discomfort or distress and for tumor growth. The entire procedure, from resection to pathological evaluation to implantation, was performed in less than 60 minutes. F1 generation tumors became palpable 8-12 weeks after implantation.

Tumor measurement and monitoring

Tumors were measured with vernier calipers (Fowler/Slyvac, Newton, MA) and tumor sizes were recorded twice weekly, after tumors had reached ~5 mm in diameter. Tumor volumes were calculated by assuming a perfect sphere and the equation v = (π/6)xd³, where d represents the mean diameter.

Tumor banking and tissue preservation

Tissue collected from normal pancreas or from primary tumor (F0) or subsequent F1 or F2 generation tumor specimens that were not needed for implantation was banked and preserved for future use by three methods. 1) Snap freezing: Tissue was dissected, placed in cryovials and immediately frozen in liquid nitrogen and stored at -80°C. 2) Viable samples: tumor fragments were placed 5 fragments/vial in 1.5 ml of M-199 + 10% DMSO in cold Mr. Frosty (Fisher Scientific, GA) and transferred to -80°C freezer overnight followed by storage in liquid nitrogen. 3) Formalin fixed paraffin embedded (FFPE) tissue: tissue was placed in 10% formalin, embedded in paraffin by standard methods, and thin-sectioned in the Comparative Pathology Lab (CPL) at UAB.

Establishment of second generation (F2) tumorgrafts

When F1 tumors reached a volume of 1,000-1,500 mm³, they were excised, washed in medium + Pen/Strep, cut into multiple fragments (~5 x 5 x 5 mm each). Tumor fragments from one F1 generation mouse were transplanted into each of four naive mice to produce an F2 generation. Donor mice (bearing F1 tumors) were euthanized by CO₂ and cervical dislocation and submerged in 70% ethanol. Necropsy was performed on donor mice. Tumor tissue not needed for implantation was banked according to the procedures described above.

Histological analysis by hematoxylin/eosin (H&E) staining

FFPE sections were de-paraffinized in two changes of xylene, followed by rehydration in two changes of absolute ethanol, and two changes of 95% and 70% ethanol. Tissue was washed briefly in deionized water and stained with Harris hematoxylin (Fischer-Scientific, Suwannee, GA). Slides were then processed in 0.25% acid alcohol, blued in lithium carbonate, and counterstained with eosin solution (Acros Organics-Thermo Fisher Scientific, Fair Lawn, NJ). Tissue was dehydrated in two changes of 95% and absolute ethanol and cleared in xylene. Photomicrographs were taken using an Olympus BH-2 microscope with DP70 camera operating with DPS-BSW v3.1 software (Center Valley, PA).

Mutational analysis of the KRAS codons 12 and 13

Genomic DNA was extracted from primary tumor (F0) and corresponding tumorgrafts, using a DNA/RNA extraction kit (EpiCentre, Madison, WI). The DNA sequence of codons 12 and 13 in exon 2 of the KRAS gene was determined using standard PCR/direct sequencing methods [30]. The sequenced fragment was a 214-bp PCR product generated with the primers KRAS F: 5’gtgtgacatgttctaatatagtca3’ and KRAS R:5’gaatggtcctgcaccagtaa3’ and 400 ng of genomic DNA [30]. Primers were designed to anneal to human intronic sequences that flank exon 2, to avoid the amplification of murine KRAS sequences. PCR conditions were: initial denaturation for 1 min at 95°C, denaturation for 30 s at 94°C, annealing for 1 min at 58°C, extension for 1 min at 72°C for 30 cycles, and final extension for 7 min at 72°C. After separation in 2% agarose gels, PCR products were extracted using a gel purification kit (Fermentas-Fisher Scientific, Savana, GA). Purified DNA (reaction products) concentration and quality was determined by ND-1000 spectrophotometer using NanoDrop 3.0.1 software (Coleman Technologies, Inc., Wilmington, DE); 280/260 ratios for all DNA examined ranged from 1.80 to 2.00. Reaction products were sequenced by UAB Center for Aids Research (CFAR) DNA Sequencing Core using ABI 3730 sequencer (Life Technologies, Carlsbad, CA). All PCR products were generated and sequenced twice in the forward and reverse directions, in independent experiments. An electropherogram of each sample was provided by CFAR at UAB and visualized using FinchTV (version1.4.0; www.geospiza.com). Of note, some of the electropherogram tracings depicting results for F0 specimens had attenuated, but readily discernible, peaks corresponding to nucleotide substitutions in the sequence of KRAS codon 12. This type of result is expected for DNA obtained from heterogeneous tissue, as demonstrated by Ogino, et al [31]. This concept is discussed in more details in the Discussion.

Statistics

Fisher’s exact tests or paired t test were performed using Prism 5.0 (San Diego, CA). A P value less than 0.05 was considered statistically significant.

Establishment of patient tumor-derived tumorgraft models

Between January and October, 2012 a total of 41 pancreatic adenocarcinoma specimens from 41 individual patients were implanted into immunocompromised mice (Table 1 ). Specimens were obtained from the primary pancreatic site (N = 39) or from a metastatic site (N = 2; 1 lymph node; 1 ligament of Treitz). Table 1 summarizes clinical data from the 41 patients who participated in this study. Engraftment was deemed successful if the F0 implant reached approximately 1,000-1,500 mm³ in size (F1 generation), a size sufficient to perform serial transplantation. Of the 41 implants, 35 were successfully propagated in mice, for an overall take rate of 85% (Table 2 ). Of the 35 implants, all remain viable. To date, 34 of the 35 have produced an F2 generation and 20 of the 34 have produced an F3 generation. Table 2 summarizes the current status of these models. The P values in Table 1 refer to the likelihood that a given characteristic was/was not related to successful propagation in mice. These data were generated using Fisher’s exact test to allow analysis of groups of unequal sizes to be compared with nominal variables (in this case, clinical parameters compared with take/no take). These analyses indicated that the size of the primary tumor from which specimens obtained significantly impacted take rate. Primary tumors measuring ≥ 2.5 cm in at least one dimension (greatest dimension) produced viable tumorgrafts more frequently than smaller tumors (96.6% take rate vs 37.5% take rate, respectively [P<0.0001]. (The percent viable tumor cells present in implanted specimens may have also affected take rate, as discussed below.) Our analyses did not identify a correlation between take rate with patient age or gender, tumor differentiation status (histology) or stage, tumor origin (head, body or tail of the pancreas), presence of lymph node metastases, lymphovascular invasion, or perineural invasion (Fisher’s exact test, Prism 5.0 software, San Diego, CA). The apparent lack of correlation between any of these parameters and take rate may reflect a real biological phenomenon or may be due to insufficient numbers of tumors in some categories to accurately predict the factors that significantly affect take rate.

Conservation of KRAS mutations in codon 12 of primary tumors and corresponding tumorgrafts

KRAS mutation is one of the most common oncogenic mutations in human cancers, including PC [32,33]. Mutations of the KRAS oncogene constitutively activate the KRAS signaling pathway. KRAS mutations are present in approximately 90% of PDAC tumors, with mutations occur most commonly in codon 12 [34]. To evaluate whether key genetic characteristics of F0 specimens were also present in tumografts derived from these specimens, we analyzed twenty tumorgraft models that have progressed to the F3 generation. Based on the availability of sufficient material from F0 tumors and corresponding F1 and F2 tumorgraft specimens, we used polymerase chain reaction (PCR) with human-specific primers (Figures 1a and 1b ) and sequencing techniques to identify the mutational status of codons 12 and 13 in the KRAS gene (Figure 1 and data not shown). No mutations were detected in codon 13 in any of the F0, F1, or F2 specimens analyzed. In contrast, as shown in Table 3 , mutations were present in primary tumor tissue (F0) in codon 12 of the KRAS gene of all specimens analyzed and these mutations were conserved in 100% of F1 and F2 generation tumorgrafts. More specifically, in pancreatic cancer, the most commonly reported KRAS mutation is the substitution of valine (V) or aspartic acid (D) for glycine (G) at position 2 in codon 12 (GGT [encoding G] → GTT [encoding V] or GAT [encoding D]) [35]. Consistent with the literature, 16 of the 20 tumors analyzed contained V (6 cases) or D (10 cases) substitutions at this position (Table 3 , Figures 1c and S1 ). The other four models analyzed harbored more rare glycine (G) → cysteine (C) or glycine (G) → arginine (R) substitutions at codon 12. Specific KRAS mutations in each of the 20 models evaluated were conserved among F0, F1 and F2 specimens. To then evaluate conservation of KRAS status among tumors of a given generation, we analyzed the sequence of KRAS codons 12 and 13 in tumors from multiple mice bearing F1 and F2 tumors that originated from each of three independent F0 specimens.

Sister tumorgrafts originating from the same F0 or F1 tumor retained KRAS codon 12 and 13 mutational status

We randomly selected three independent primary (F0) tumors, and analyzed the sequence of KRAS codons 12 and 13 of these tumors. We also analyzed three F1 progeny derived from each F0 tumor (three “sets” of three F1 sister tumorgrafts) and four F2 progeny of one tumor from each set of F1 tumors (three “sets” of four F4 sister tumorgrafts) (Figure 2 ). The data demonstrate that sister tumorgrafts (from mouse 1 [m1], mouse 2 [m2], mouse 3 [m3], and mouse 4 [m4] of each set) from all three models (UAB-PA2; UAB-PA4; UAB-PA10) conserved KRAS codon 12 status. As in previous analyses, no mutations in codon 13 were identified (data not shown). These results indicate that tumorgrafts in this study having the same origin displayed fidelity of KRAS status within tumors of a given generation.

F1 and F2 generation tumorgraft tissues are morphologically similar to F0 primary tumor tissue of origin

Having demonstrated fidelity in F0, F1, and F2 tissue specimens with respect to the mutation status of codon 12 of KRAS, we next addressed fidelity of tissue morphology of the same twenty models. As shown in Table 1 , a majority of these models were derived from stage IIA or IIB tumors, stages of PDAC regarded as resectable disease.

Histology was evaluated using criteria for clinical staging and diagnosis of PDAC, a subtype that accounts for 90% of all cases of PC. PDAC displays characteristic lesions that have at least one of three specific histological traits: pancreatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasm (IPMN), or mucinous cystic neoplasia (MCN) [36]. Among these, the most common and well-characterized precursor lesion is PanIN. PanIN lesions, in turn, are subdivided into four grades (PanIN-1A, PanIN-1B, PanIN-2, and PanIN-3) based on degree of dysplasia, as reflected by cytologic atypia and architectural changes [37]. Briefly, PanIN-1A and 1B are characterized by tall columnar cells and mucin production. PanIN-1B lesions have a papillary (or micropapillary) architecture. PanIN-2 lesions produce mucin, and have predominant nuclear abnormalities. Histological abnormalities of PanIN-2 type lesions include cytonuclear atypia, nuclear crowding, and nuclear abnormalities (most commonly nuclear enlargement). PanIN-3 lesions are characterized by many of the cytologic hallmarks of solid tumors in general, such as cribriforming, loss of nuclear polarity, nuclear atypia, luminal necrosis, abnormal mitoses, and budding off of groups in small clusters into the ductal lumen. Eventually, progressive development of dysplasia by all four subclasses of PanIN precursor lesions produces invasive pancreatic adenocarcinoma.

Summaries of the histological and morphological characteristics (by LNC) of each of the twenty tumorgraft models that have progressed to the F3 passage are based on: degree of differentiation, atypical gland formation, cytonuclear atypia, and nuclear abnormalities (e.g., nuclear:cytoplasmic ratio). These characteristics comprise the predominant indicators of PDAC malignancy in primary patient tumors (F0) and are used to describe the degree to which tumorgraft (F1, F2) progeny reflected characteristics of F0 tissue. We defined the degree of differentiation based on the presence or absence of glandular formation. Tumors containing gland tissue were designated as well-differentiated or moderately differentiated. Tissue closely resembling normal pancreatic (NP) tissue is routinely designated as well-differentiated. However, none of our tumor specimens (presented in Figures 3 and S2 ) fulfilled this criterion; all specimens contained moderately to poorly differentiated tumor cells. Tumors with complex histology, for example with cystic, cribriformed, solid nest, or malignant cytology were designated as moderately differentiated. Tumors that had not formed glands, as evidenced by sheets of cells, sarcomatoids differentiation, or anaplastic cells, were designated as poorly differentiated. In general, all tumorgraft models evaluated retained the essential characteristics of primary pancreatic adenocarcinomas from the F0 through the F2 generation. H&E images and detailed histological characterization of eight tumors are shown below (Figure 3 , panels a-h); similar analyses of the remaining twelve tumors are depicted and described in Supporting Information (Figure S2 , panels j-u).

The F3 generation of UAB-PA2 tumorgrafts retains the histology of the F0 generation

To date, we have sufficient material for histologic analysis from the F3 generation of only model UAB-PA2. Notably, as shown in Figure 4 , the F3 tissues have characteristics very similar to those of F0 UAB-PA2 tissue. All generations, F0, F1, F2 and F3 tissues show equivalent N:C ratios, display consistent cellular differentiation, and have similar histological features. This in vivo F3 generation preserved the histological features of its tumor of origin. We next compared the growth patterns of F1 and F2 generations of each of these twenty models.

Comparison of the growth patterns of F1 with F2 generation tumorgrafts

The growth curves of F1 tumors produced by individual F0 tumor specimens following implantation into mice are shown in Figures 5a , S3a and S3c. The F1 tumors demonstrated a range of growth rates that differed by ~12-fold, based on an approximated slope during exponential growth (range of slope values: 28 [UAB-PA8] to 344 [UAB-PA2]). F1 tumors reached ~1,500 mm³ 11-38 weeks after implantation.

F2 tumor growth curves (Figures 5b , S3b and S3d), generated from data obtained from 2-4 mice per tumor model, show that while F2 tumors grew at rates similar to each other, differences were observed in the interval between implantation and measurable tumor progression (lag phase). Compare, for example, the slopes of the lines representing UAB-PA2-F2 tumor growth compared to UAB-PA16-F2 after a ~7-week lag period. F2 tumors reached ~1,500 mm³ 6-25 weeks after transplantation.

Cryopreserved tumorgraft transplant shows the growth similar to that following direct transplantation of non-frozen specimen

Because it will be important to ensure the availability of tumorgraft models as novel chemotherapeutic agents become available for testing, we also examined whether cryopreserved F1 specimens formed tumors as readily as F1 specimens that had not been frozen. We compared in vivo growth characteristics of F2 tumors derived from frozen F1 specimens (UAB-PA2-F2-FV tumors) compared to fresh F1 specimens with the UAB-PA2 model. As outlined in Methods, after dissection, F1 specimens were either transplanted directly into mice (to produce UAB-PA2-F2) or stored in liquid nitrogen for a minimum of one month and then transplanted into mice to produce UAB-PA2-F2-FV tumors. H&E-stained sections of F2 tumors produced by both fresh and frozen F1 specimens showed the same degree of moderate differentiation and similar morphology: complex gland formation and increased N:C ratio with relatively intact nuclear polarity (Figure 6a ). The single difference noted between the two F2 generation progeny was that tumorgrafts from cryopreserved specimens had an apparent lag time of ~2 weeks (P=0.0809; Figure 6b ) compared to no apparent lag time for tumors derived from fresh specimens.

Primary tumor specimens that failed to produce tumors in mice showed distinct morphologies

Our primary tumor take rate in mice for pancreatic adenocarcinoma is 85%, higher than the previously reported 22-62% [38-40]. However, six of the original 41- F0 tumor specimens failed to produce tumors in mice. To determine whether these specimens represented a unique subset of PC tumors, we evaluated four of these six tumors for which we had sufficient FFPE tissue for analysis (Figure 7 ). H&E stained sections of the first specimen (UAB-PA11) (Figure 7a ) shows F0 tissue containing benign pancreatic tissue with chronic pancreatitis, but no identifiable tumor cells (0% tumor content). The second specimen (UAB-PA17) (Figure 7b ), shows F0 tissue containing benign acinar pancreatic tissue, but no identifiable tumor cells (0% tumor content). The third specimen (UAB-PA21) (Figure 7c ) appears to contain some tumor cells (~10% of total cellular content), but is comprised primarily of stromal tissue. The fourth specimen (UAB-PA24) (Figure 7d ) shows focal suspicious glands in a degenerating pancreatic lobule against a background characteristic of chronic pancreatitis and remnant acinar tissue (red arrow), but no identifiable tumor cells (<1 % tumor content). These observations are consistent with the hypothesis that a critical ratio of tumor cells: stroma or a critical number of tumor cells is required to support growth in immunocompromised mice. In addition, the morphological features of specimens that differ from those successfully propagated in mice include fewer atypical glands and the presence of acini/stroma consistent with chronic inflammation. We concluded that the F0 specimens that did not produce tumors in mice had common histological features and a minimal percentage of malignant cells (0-10%). Of note, all primary specimens obtained from tumors of patients diagnosed with PDAC. Specimens that did not produce tumors when implanted into mice simply did not contain a sufficient number of tumor cells to produce tumors, emphasizing the heterogeneity of these specimens and the importance of verification of tumor histology prior to implant.