As summarized in , we obtained fresh HCC samples from 16 consecutive patients who were undergoing surgical resection of their tumors as primary therapy. None of the patients had received any form of neoadjuvant therapy. The diagnosis of HCC and degree of tumor differentiation in each resected specimen was determined through standard diagnostic assessments performed by clinical hepatopathologists at our institution independent of this study. Of 21 xenografts generated from the 16 different HCC specimens, we identified only 10 xenografts that resembled HCC on initial histopathological assessment (“HCC-like xenografts”). The remaining 11 xenografts resembled lymphoid neoplasms rather than HCC (“non-HCC-like xenografts”). In comparing the groups of patient samples that yielded HCC-like xenografts as compared with those that yielded non-HCC-like xenografts, there were no obvious differences between groups in patient age (68±11 years vs. 62±10 years, p
0.27), underlying liver disease, or degree of differentiation of primary tumors. HCC-like and non-HCC-like xenografts arose in both NOD/SCID and NSG mice, and the time intervals between HCC implantation and harvesting of the first xenografts were similar (180±81 days vs. 144±65 days, p
Patient demographics, parent HCC grade, and xenograft characteristics.
As shown in , HCC-like xenografts shared many histological features of HCC with parent tumors, including hepatocyte-like cells with nuclear atypia and high nuclear-to-cytoplasmic ratio, absence of portal tracts, and distorted trabeculae with increased thickness of hepatocellular plates. As shown in , non-HCC-like xenografts differed significantly from parent HCC tumors and HCC-like xenografts. Non-HCC-like xenografts had no architectural features of HCC, instead consisting of monomorphic populations of small lymphoid mononuclear cells with nuclear atypia and high mitotic index. As shown in , analysis of xenografts by RT-PCR revealed that HCC-like xenografts retained expression of many liver cell markers, while these were diminished or absent in non-HCC-like xenografts. We suspect that the persistent detection of liver epithelial markers such as cytokeratins, AAT and TDO in some non-HCC-like xenografts reflects the extreme sensitivity of RT-PCR to detect the presence of some persisting HCC cells in non-HCC-like xenografts despite near replacement of the originally implanted HCC tissue by rapidly proliferating lymphoid cells.
Xenografts arising from human HCC specimens.
Due to the resemblance of non-HCC-like xenografts to lymphoid neoplasms, we evaluated the expression of human and murine leukocyte markers by immunohistochemistry. In contrast to the typically sparse distribution of leukocytes observed in all of the parent human HCC tissues to a similar extent mainly along portal tracts when they were invaded by the tumor (), non-HCC-like xenografts were densely infiltrated or replaced by cells that could be characterized as human B lymphocytes by their expression of human CD45 and CD20 (). Flow cytometry confirmed that the human CD45+ population consisted predominantly of CD19+ cells, consistent with human B lymphocytes (). Staining for the human T-cell antigen CD3 was minimal, as was staining for the mouse lymphocyte antigen B220 and the mouse histocompatibility antigen H2k (data not shown). We did not identify any correlation between leukocyte infiltration in parent HCC tissues and the development of non-HCC-like xenografts. In addition, prior to their HCC resection, none of the source patients had a history of lymphoproliferative disease or an immunodeficient state known to predispose to lymphoproliferative disorders (eg. human immunodeficiency virus infection, pharmacological immunosuppression).
Expression of leukocyte markers and EBER in non-HCC-like xenografts.
Cognizant of the importance of EBV in the pathogenesis of lymphoproliferative disorders in immunocompromised humans 
, we evaluated our xenografts for evidence of EBV infection by in situ
hybridization (ISH) for EBV-encoded RNA (EBER), and found that this was strongly positive in the B lymphocytes populating the non-HCC-like xenografts (). We were unable to determine the presence or absence of latent EBV in parent HCC specimens using EBER ISH and PCR for EBV DNA, and this needs to be prospectively evaluated. Since EBV is so effectively suppressed in immunocompetent humans, we suspect that the tiny amount of EBER and EBV DNA that would be present in a few B-cells in the source tumor, of which we had extremely limited histological sections for retrospective analysis, was below the detection threshold of our assays. Preoperative EBV serology of source patients was not available since this is not routinely tested.
To further characterize the non-HCC-like xenografts, we evaluated the clonality of the human B lymphocyte proliferations by assaying immunoglobulin heavy chain (IgH) gene rearrangements. As shown in , B-cell proliferations arose from a single clone in 10 of 11 non-HCC-like xenografts, and from two clones in the remaining case. In two instances, independent xenografts obtained from separate fragments of a single parent HCC developed into lymphoid tumors that demonstrated different IgH rearrangements. This suggests that the malignant transformation and clonal expansion responsible for the EBV-associated lymphomas observed in this study occurred subsequent to the xenotransplantation of human tissue into immunodeficient mice, and that lymphomas were not already present in source tissues.
Immunoglobulin heavy chain gene rearrangements in non-HCC-like xenografts.
To exclude contamination by a lymphocyte cell line or cross-contamination between xenografts as a source of the lymphomas, we performed short tandem repeat (STR) analysis on genomic DNA from each xenograft and confirmed that xenografts from different patients were genetically distinct ().
Short tandem repeat (STR) analysis demonstrating unique genetic identity of each xenograft.
Collectively, these findings demonstrate that non-HCC-like xenografts were comprised of EBV-associated human B-cell lymphomas that had developed spontaneously from xenografted human HCC tissues. Analogous to EBV-associated lymphoproliferative disorders in immunocompromised humans, this likely occurred through reactivation of latent EBV and malignant transformation of intratumoral passenger B lymphocytes made possible by the immunodeficiency of the recipient mice. In future work, we plan to apply cell sorting techniques to exclude CD45+ cells from human HCC samples prior to xenotransplantation in order to prevent this process.
The phenomenon of EBV-associated lymphomagenesis has not been reported in the context of human solid tumor xenografts. Considering the high prevalence of latent EBV in humans 
and the presence of B lymphocytes in solid tumors from all human tissues 
, it is likely that most solid tumors would be vulnerable to this process. It is unlikely that this phenomenon is unique to HCC, as EBV has not been implicated in the pathogenesis of human HCC 
, primary hepatic lymphoid malignancies are very rare 
, and there is no clearly established causative link between chronic liver disease and the incidence of lymphoproliferative disease. We thus believe that our observations are of importance to all investigators utilizing human solid tumor xenotransplantation assays. As we have shown, lymphomas can arise within the first xenografts obtained, and do not require serial tumor passaging. Unrecognized, this process may confound experimental data such as measures of tumor growth and markers of tumor cell populations. Lymphomas may also competitively eliminate the tumor tissue of interest, resulting in loss of valuable samples.
This process can best be recognized through phenotyping of xenografted tissues using leukocyte markers. Histopathological assessment is not sufficient, as lymphomas may be misinterpreted as poorly differentiated epithelial tumors. Initial growth kinetics of lymphomas in our model was similar to HCC-like xenografts, indicating that this process cannot be detected by gross observations alone. Because lymphomas may arise from one or two B-cell clones, immunomagnetic depletion of CD45+
cells may not be sufficient to prevent this process since all CD45+
cells are not removed 
. Flow cytometry-based single-cell sorting is the most stringent technique to identify and eliminate human leukocytes from solid tumor xenotransplantation assays. Most rigorous analyses of human solid cancer xenografts incorporate routine quantification, exclusion or depletion of contaminating CD45+
cells from assays targeting biological and functional properties of specific subpopulations of tumor cells 
; our observations provide a compelling biological rationale to support this practice and extend it to include all observations involving solid tumor xenografts.
It is important to note that the generalizability of our observations should be interpreted with caution in the context of a relatively small sample size of patients, HCC specimens, and xenografts. The clinical characteristics of the source patients are observational only and were retrospectively analyzed, making this study vulnerable to potentially confounding patient factors or differences that it is underpowered to detect. Similarly, although our data does not demonstrate any obvious relationships between the xenografting methods used (eg. tumor fragment vs. cell suspension, NSG vs. NOD/SCID mouse strain) and the development of EBV-associated lymphomas, the study is underpowered to detect the relative influence of these variables on the results.
In summary, we have shown that human solid tumor xenografts in immunodeficient mice are vulnerable to lymphomagenesis associated with EBV. This potentially confounding process can be recognized through immunophenotyping of xenografts using leukocyte markers and should be preventable by excluding leukocytes from source tissues. This phenomenon should be recognized as an important pitfall of human solid tumor xenotransplantation assays.