1 INTRODUCTION

Hepatocellular carcinoma (HCC) is one of the leading causes of cancer mortality worldwide, largely associated with chronic hepatitis B and C virus (HBV, HCV) infections.1 In the U.S. most HCC are caused by HCV infection and alcoholic liver disease (ALD) with a recently rising incidence in HCC secondary to non-alcoholic fatty liver disease (NAFLD).2 The current treatment paradigm consists of curative surgical or radiological interventions for early stage HCC. Tumors that have spread or are not amenable to local interventions are treated with systemic therapies. Sorafenib was the first agent to show a modest benefit for locally advanced and metastatic HCC, while other targeted therapies failed Phase 3 trials at the time of this effort.3 More recently, several agents including tyrosine kinase inhibitors4-6 and immune checkpoint blockade-based therapies7-9 have shown improvement in survival outcomes. Genomic diversity among HCC and intra-tumor heterogeneity challenge clinical trials targeting one or few genomic alterations.10 HCC may develop multicentrically from different clones or arise from a single original tumor via intrahepatic metastasis. It is estimated that 22%–79% of synchronous HCC vary clonally, and 12%–66% of single tumors contain intratumor heterogeneity.11 As a direct consequence of the tumor heterogeneity, biopsies are necessary for identifying valid biomarkers for HCC and better research models are needed that represent the heterogeneous nature of human HCC.

Human HCC cell lines have been utilized for research, but their translational value has thus far remained limited.12 For in vivo studies, many mouse models are available that develop murine HCC.13 However, existing mouse HCC models poorly recapitulate the human heterogeneity resulting from the aberrant function of multiple molecular pathways.14-16 Poor reproducibility of murine HCC models has further limited their use.17 Therefore, it is imperative to develop experimental models that hold more translational potential, can increase our understanding of human HCC biology and serve as platforms for pre-clinical testing of novel therapies.

A different strategy to study tumor biology in laboratory animals is to engraft human cancers into immunodeficient mice, commonly referred to as patient-derived xenografts (PDX).18, 19 For a number of human cancers, PDX models have been useful to study of cancer biology in vivo and predict patient responses to chemotherapeutic regimens.20-22 To date, PDX have advanced insights into both basic biology and anti-tumor strategies for breast, colorectal, and several other cancers.19 PDX from HCC case series are dominated by HBV-associated tumors.23-27 Establishing PDX from HCC in Western populations has proved more challenging with low success rates.28-30 With the exception of Zhu et al.30 these series compared patient and tumor characteristics rather than animal variables to determine which HCC could establish PDX. We here aimed to quantify to what extent mouse variables influenced PDX formation with HCC biopsy materials obtained from a U.S. patient population. We tested the effects of the murine immunodeficiency, the surgical implantation site, and whether mouse liver injury affected PDX formation.

2 MATERIALS AND METHODS 2.1 Human subjects

The protocol was reviewed and approved by Memorial Sloan Kettering Cancer Center Institutional Review Board, and by the Rockefeller University Institutional Review Board.

2.2 Biopsies

Adult patients of both genders and whose liver lesions were clinically highly suspicious for HCC, required a diagnostic biopsy were included in the study and provided written informed consent for use of tissue samples. Eligible HCC etiologies were HCV infection as determined by HCV serology and/or detectable HCV RT-PCR; HBV infection as determined by HBV core antibody and/or surface antigen serology, and/or detectable HBV PCR; alcoholic liver disease (ALD) as determined by the history of chronic alcohol use; NAFLD as determined clinically by history, past or current fatty liver on imaging, previous biopsy confirmation and associated metabolic conditions. Biopsies were performed at the interventional radiology department at MSKCC, tumor samples were obtained by image-guidance technique from primary HCC or metastatic site. Biopsy material was cut into small pieces (~0.5 mm3) for implantation.

2.3 Animals

Balb/c Rag2−/−, NOD Rag1−/− Il2rgnull (NRG) and Foxn1nu mice were obtained from Jackson Labs (Bar Harbor, Maine). Mice with a targeted disruption in fumaryl acetoacetate hydrolase (Fah−/−) were kindly provided by M. Grompe (Oregon Health & Science University) and crossed with Rag2−/− Il2rgnull or with NRG mice.31 Immunodeficient Fah−/− were bred and maintained on nitisinone (Yecuris, Tualatin, OR) to prevent liver damage. To induce chronic liver injury immunodeficient Fah−/− mice were cycled off nitisinone after HCC biopsy implantation or kept on continuous nitisinone as non-liver injury controls. At the time of surgery mice were 6–10 weeks of age with weight varying from 20 g up to 30 g. All procedures were reviewed and approved by Rockefeller University Institutional Review Board and Committee on Use and Care of Animals under protocol number 18063.

2.4 Implantation and engraftment monitoring

For intrahepatic implantation, the upper abdomen of mice was shaved and sterilized with iodine. After mice had been anesthetized using isoflurane, a 10–15 mm skin incision was made over the left subcostal margin after which the peritoneum was mobilized. Using a cautery device, 10–15 mm of peritoneum was opened, and the large lobe of the liver was exposed. An ~1 mm opening was created in the center of the lobe and two to three ~0.5 mm3 pieces of biopsy material were loaded into the liver. The outflow tract was cauterized to prevent bleeding and then the peritoneum closed with a Vicryl suture and the skin with hemostat wound clips.

For subrenal capsule (SRC) implantation the left flank of mice was shaved and sterilized with iodine. After mice were anesthetized using isoflurane, a 5–7 mm skin incision was made in the left midaxillary plane, after which the peritoneum was mobilized. Using a cautery device 7–8 mm of peritoneum was opened, the left kidney exposed and at the caudal pole. A ~1 mm opening in the capsule was created. Two to three ~0.5 mm3 pieces of biopsy material were placed under the kidney capsule and moved cranially. The kidney was mobilized back into the peritoneum, closed with a Vicryl suture and the skin approximated with hemostat wound clips.

Engraftment growth was monitored by quantification of human HCC markers in mouse serum every 3–4 weeks by using commercially available human-specific ELISA antibodies and protocols as described previously.31 If animal health were adversely affected by tumor growth as shown by displaying lethargy, hunched appearance, ruffled fur, and/or cachexia, the experiment was terminated in accordance with the Rockefeller University Institutional Animal Care and Use Committee (IACUC) protocol.

2.5 Tissue harvest, passaging, and cryopreservation

Once tumors plateaued as assessed by serum markers or imaging, PDX tissues were harvested for histology and cryopreservation. After CO2 asphyxiation the liver and the left kidney were harvested. Tumors were weighed after separation from the liver or kidney. For tumors in which the kidney could not be separated or identified the median left kidney weight for that strain was subtracted from total weight. Tumor tissue was collected in 10% buffered formalin for histology. The remaining tissue was cut into small, 1–2 mm3 pieces for immediate passaging or put in hepatocyte freezing medium containing 10% DMSO for long-term storage at −150ºC.

2.6 Histological comparison

Samples were formalin-fixed, paraffin-embedded, cut into 5.0 micron sections, and stained with hematoxylin and eosin (H&E). Histopathology was examined under light microscopy.

2.7 Statistical analysis

Pearson correlation coefficient was used to assess the linear correlation between tumor weight of mice with serum hAAT among PDX1. A linear mixed model was used to further evaluate the association between tumor weight and serum hAAT from different PDX. At the biopsy level, Fisher's exact test was used to examine the association between immunodeficiency status and implantation site on the success of PDX formation. To account the clustering of repeated binary observations from mice within a PDX, the generalized estimating equations (GEE) logistic regression model was used to examine the association between the factors mentioned above and PDX formation in the mouse level. To look at the association between liver injured and no liver injury on the 18 biopsies implanted under both conditions, the exact McNemar's test was used. Biopsies that were implanted into Rag2-/- mice were excluded.

All analyses were carried out in GraphPad Prism Software and SAS 9.4 (SAS Institution, Cary, NC). All p-values were two-sided and p-value of 0.05 indicated statistical significance.

3 RESULTS 3.1 Clinical HCC characteristics and overall PDX formation rate

Sixty-two biopsies were obtained from 60 patients with presumed HCC enrolled in the study. Seventeen samples were excluded from the analysis as they were determined either as cirrhotic liver without HCC (n = 4), a non-HCC cancer (n = 5) or because there was no or not enough tissue for implantation (n = 8). Of the 45 HCC biopsy specimens that were implanted, 37 were obtained from male patients (82%). Patient demographics and tumor characteristics are summarized in Table 1. Our biopsy material selection highlights the HCC etiological distribution most commonly noted in the United States. The underlying liver disease etiologies of the 45 HCC biopsies were HCV (n = 15), alcoholic liver disease (ALD) (n = 6), HBV (n = 8), NAFLD (n = 5), mixed etiology (alcohol and HCV n = 2; alcohol and NAFLD n = 1; HCV and NAFLD n = 1), and no identifiable liver disease etiology or cryptogenic cirrhosis (n = 7).

TABLE 1. Patients demographics and tumors characteristics of implanted HCC biopsies Biopsy Race Ethnicity Sex Site of biopsy Etiology Differentiation grade Growth AFP, ng/ml HCC 1 White Non-Hispanic Male Liver ALD Moderate Yes 4.2 HCC 2 White Non-Hispanic Male Liver ALD Well to moderate Yes 38.4 HCC 3 Asian Non-Hispanic Female Liver HBV Moderate Yes 1389 HCC 4 White Non-Hispanic Female Liver NAFLD Poor Yes 1062 HCC 5 N/A Non-Hispanic Male Liver ALD moderate Yes 10.1 HCC 6 Asian Non-Hispanic Male Peritoneum ALD Moderate Yes 27.6 HCC 7 White Unknown Male Liver NAFLD N/A No 263 HCC 8 White Non-Hispanic Male Perirenal None N/A No 5 HCC 10 Other Hispanic Male Liver ALD, HCV Mostly necrotic Yes 14,121 HCC 13 White Non-Hispanic Male Liver HCV N/A Yes 949,118 HCC 14 White Non-Hispanic Male Liver ALD, NAFLD Poor Yes 2719 HCC 15 Asian Non-Hispanic Male Liver ALD Moderate Yes 208 HCC 16 White Non-Hispanic Male Liver HCV N/A Yes 1035 HCC 17 White Non-Hispanic Male Liver HCV Moderate Yes 15.4 HCC 18 White Non-Hispanic Male Liver HCV Well No 39.6 HCC 19 White Non-Hispanic Male Liver HBV, HIV Moderate No 6732 HCC 20 White Non-Hispanic Male Liver HCV Moderate Yes 43 HCC 21 N/A Non-Hispanic Male Liver HCV Well to moderate Yes 2.9 HCC 22 White Non-Hispanic Female Liver None Moderate No 42,323 HCC 24 White Non-Hispanic Male Liver HCV Well No 180 HCC 25 White Non-Hispanic Male Liver ALD N/A Yes 55,089 HCC 26 White Non-Hispanic Male Pancreas HCV Poor Yes 1470 HCC 27 White Non-Hispanic Male Liver ALD Moderate No 381,770 HCC 28 Black Non-Hispanic Male Liver HCV Moderate Yes 51,687 HCC 29 White Non-Hispanic Female Paracolic NAFLD Poor No 2293 HCC 31 White Non-Hispanic Female Abdominal HBV N/A No 16,469 HCC 31 (2) Peritoneal Node Poor Yes HCC 32 Black Non-Hispanic Female Adrenal None Well Yes 355 HCC 32 (2) Adrenal N/A No HCC 33 White Non-Hispanic Male Liver HBV N/A No 6 HCC 35 White Non-Hispanic Male Liver HCV N/A No 6 HCC 36 Other Hispanic Male Liver HCV Moderate Yes 1,249,574 HCC 40 White Non-Hispanic Female Liver Cirrhosis Moderate No 228 HCC 42 White Non-Hispanic Male Liver Crohns Poor No 52 HCC 46 Pacific Islander Non-Hispanic Female Liver HCV Moderate No 991 HCC 48 Black Non-Hispanic Male Liver HBV N/A No 143,236 HCC 49 White Non-Hispanic Male Liver HCV Well No 542 HCC 50 White Non-Hispanic Male Liver HCV, HIV Poor No 30 HCC 52 White Non-Hispanic Male Liver HBV Moderate No 4.4 HCC 53 Asian Non-Hispanic Male Liver HBV Moderate No 125 HCC 56 White Non-Hispanic Male Liver None Moderate No 2.2 HCC 57 White Non-Hispanic Male Liver HCV Poor No 74 HCC 59 White Non-Hispanic Male Liver HCV Well to moderate No 110 HCC 60 White Non-Hispanic Male Liver Unknown Well to moderate No 1227 HCC 62 White Non-Hispanic Male Lung Cirrhosis N/A No 263 Abbreviations: ALD, alcoholic liver disease; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; NAFLD, non-alcoholic fatty liver disease; N/A, not available.

We found that 20 out of 45 HCC biopsy specimens could establish PDX lines for an overall success rate of 44%. Expectantly, the majority came from male patients (n = 16; 80%). These were equally distributed among the four major HCC etiologies in the United States, with 8 (40%) from HCV, 6 (30%) from ALD, 1 (5%) from NAFLD, 2(10%) from HBV, 2 (10%) from alcohol with a secondary liver disease, and 1 (5%) with cryptogenic cirrhosis. These results on a small sample size suggest that HCC biopsy materials from a patient population in the United States can form PDX without a clear preference toward certain liver disease etiologies.

3.2 Human alpha1-antitrypsin is a serum marker of HCC PDX formation in mice

Most HCC tumor specimens have been implanted under the skin of mice, which allows for the detection of large PDX by palpation. When HCC are implanted into other anatomical sites palpation becomes insensitive for small tumors, imaging can be used but is operator-dependent, labor-intensive, and costly. We thus set out to test if human markers in mouse serum could identify animals that grew PDX. This would allow for easier PDX monitoring in non-subcutaneous implantation sites such as intrahepatic (IH) or subrenal capsule (SRC). Serum from mice that were implanted with HCC biopsies were screened for several human markers including albumin (hAlb), α1-antitrypsin (hAAT), transferrin, and α-fetoprotein (hAFP). As illustrated by PDX1 (Figure 1A), hAAT became detectable in 3 out of 5 mice shortly after HCC implantation and rose over time along with hAlb, while hAFP (not shown) remained undetectable. PDX that released multiple human markers in mouse serum typically did so with similar ratios as illustrated by PDX3 (Figure 1B). Of the first 3 HCC biopsies that were transplanted into mice and that resulted in the rise of any human serum marker, all secreted hAAT. Other markers were inconsistently detectable. From then on serum hAAT was used as a marker for successful PDX formation. Indeed, all mice that contained macroscopic tumors showed rising hAAT levels. And conversely, of mice that were transplanted with HCC biopsy materials and that had unquantifiable (<50 ng/ml) hAAT levels for up to 6 months after surgery, none were found to have a macroscopic PDX. This suggests that serum hAAT, which historically has been investigated as a clinical HCC marker,32 is an effective tool to monitor PDX formation.

Human AAT is a serum marker for PDX formation. (A) After implantation of HCC1 biopsy materials mice were serially bled and human alpha1-antitrypsin (hAAT, grey circles) and human albumin (hAlb, white squares) were quantified in mouse serum. Three of five mice showed rising levels of both hAAT and hAlb over time. (B) For PDX3 that secreted hAAT, hAlb, and human alpha-fetoprotein (hAFP), hAAT levels correlated to the other proteins even though the ratio of hAlb and hAFP varied between mice. (C) A mouse that was transplanted with HCC1 under the SRC and had serum hAAT levels of 1.5 mg/ml was examined by ultrasound. The PDX could readily be visualized as a hypoechoic mass adjacent to the kidney. (D) Tumors from mice that received passaged PDX1 under the SRC were harvested, dissected, and weighed. Serum hAAT levels on the day of harvest correlated with the tumor weight. Pearson correlation coefficient. (E) Tumors from mice transplanted with various PDX lines were weighed and serum hAAT levels quantified. There was no association between tumor weights and hAAT levels across PDX lines

Next, sonographic detection of PDX was explored in mice with rising hAAT serum levels. In mice with high hAAT levels (e.g., >1 mg/ml for PDX1), PDX in the SRC could be visualized by ultrasound as a hypoechoic mass adjacent to the left kidney (Figure 1C). However, PDX in mice with low serum hAAT levels could not convincingly be visualized by ultrasound either for SRC or IH implanted biopsy materials.

Finally, serum hAAT levels were compared to tumor size. For passaged PDX1 there was a correlation (r = 0.8, p = 0.003 by Pearson correlation coefficient) between hAAT serum levels and tumor weight (Figure 1D). After accounting for the clustering serum hAAT and tumor weight data from different PDX lines did not show such correlation, which was in line with the highly heterogeneous nature of these tumors (beta = −434.72, p = 0.258) (Figure 1E).

These data show that hAAT quantification in mouse serum can serve as an effective tool to screen for PDX formation, and that for an individual PDX line hAAT serum levels correlate with tumor weight.

3.3 Only severely immunodeficient mice allow for PDX formation from HCC biopsy material

Innate, humoral and T cell-mediated rejection all form major immune barriers to xenotransplantation,33 which has led to the widespread use of T and B cell-deficient mice such as the severe combined immunodeficient (SCID) and recombination activating gene knockout (Rag−/−) strains to establish PDX.34 We here aimed to test to what extent the immunodeficient background of recipient mice influenced PDX formation. Three immunodeficient strains were used for PDX formation1: T and B cell deficient Rag2−/− animals2; Rag2−/− mice additionally lacking the interleukin-2 receptor gamma chain (Rag2−/− Il2rgnull), which results in the absence of natural killer and innate lymphoid cells and some impaired myeloid functions3; animals that in addition to Rag1−/− Il2rgnull were on the non-obese diabetic background (NOD Rag1−/− Il2rgnull or NRG), which impairs macrophage xenorejection.