Microarray expression profiling has been quite successful in differentiating subclasses within tumors with a similar histological/morphological diagnosis [18, 30]. Tumors with common genetic lesions tend to cluster together, reflecting the fact that deregulation of one or more genes leads to downstream gene activation/inactivation cascades that are unique to that lesion. It has also made it possible to predict the prognosis of cancer patients and the likelihood of their responding to chemotherapy, based on gene expression patterns [18, 31]. Major progress in experimental cancer research will require animal model systems and in vitro cell culture systems. Experimentally induced PCTs and related BCLs have long been, and continue to be, useful for the study of plasma cell neoplasms.
In this study, we generated gene expression profiles of mouse B-lymphocytic neoplasms, including four groups of PCTs that arose with different latent periods in chronic inflammatory tissue induced by ip pristane. We also studied two subgroups of PCTs that arose spontaneously in hyperplastic lymph nodes of IL-6-transgenic mice and B-cell lymphomas from iMyc gene-insertion mice. Statistical methods were employed to compare these expression profiles to search for specific gene expression signatures that might aid in understanding the molecular mechanisms at work in each subtype of B cell malignancy and the differences in tumor latent periods resulting from variation in the PCT induction schemes.
Classical unsupervised clustering analysis and statistical analysis of microarray data revealed a tight clustering of PCTs regardless of subtype. The expression patterns in this group showed major differences from those of the B-cell lymphoma group. As seen in earlier analyses , the stage of B cell differentiation strongly influenced the major gene expression differences. It is important to note that virtually all of the PCTs clustered together, including both intraperitoneal oil-induced PCTs as well as those that arose spontaneously in IL-6-transgenic lymph nodes, independent of ip pristane. These results indicated that all PCTs share a common genetic expression signature. It can also be seen that many genes present in the B-cell signature have been downregulated in the PCT signature, a well-known feature of the specialized plasma cell, which seems to reflect this cell's single-minded dedication to maximize synthesis of antibody Ig.
Many of the genes that emerged as part of the PCT signature after unsupervised clustering analysis had been recognized in previous studies of gene expression patterns as characteristic of this stage of B cell differentiation, before and after neoplastic transformation [17, 19, 32]. Irf4 and Syndecan1 are characteristically expressed in plasma cells, and cyclin D2 has been reported to have high expression in PCT, presumably secondary to deregulated c-Myc.
Subclusters within the PCT cluster were seen, chiefly reflecting the slow vs. rapid appearance of the PCTs and the nature of the accelerating oncogenes. Unlike human myelomas, mouse plasma cell tumors have a common initiating event, a Myc-activating chromosome translocation. We found that unsupervised clustering of these tumors did not reveal major differences in gene expression that could be associated with position of breakpoint and translocation on chromosome 15. That is, the nature of c-Myc-activating chromosomal translocations [i.e., typical T(12;15) vs. variant T(6;15)] did not affect their inclusion in the PCT group, nor did it lead to subclusters within the broad PCT grouping or the ABPC or TEPC subgroups, in this particular analysis. This supports our hypothesis that even T(6;15) translocations that occur within the Pvt-1 region initiate PCTs via induction of constitutive c-Myc expression, and that fine structural differences in the site of T(12;15) within or near the c-Myc locus itself do not materially affect the way Myc influences PCT induction. Of course, we know that additional genetic events are required for complete transformation of plasma cells, for Myc-activating chromosome translocations have been found in normal cells and organs that did not develop plasma cell tumors . Future studies will focus on identifying these additional events that occur during tumor progression.
Our cluster analysis did, however, reflect differences in the cooperating genes used to accelerate tumorigenesis. The most striking gene expression changes associated with tumors accelerated by Abelson virus infection are the high levels of Socs1 expression. The Socs proteins are inhibitors of signaling pathways and are generally expressed at low levels in unstimulated cells. They become rapidly induced by cytokines, thereby inhibiting Jak-STAT signaling, forming a classic negative feedback loop . It was previously shown that the v-Abl oncogene activates Jak-STAT signaling during transformation of pre-B cells in mice . Although Socs-1 is highly expressed in v-Abl-transformed cells, it is unable to inhibit v-Abl-mediated Jak- STAT signaling. This is thought to be the result of one of the non-tyrosine kinase effects of v-Abl, such as phosphorylation of Socs-1 on non-tyrosine residues, leading to disruption of the interaction between Socs-1 and the elongin B/C complex and inhibition of Socs1-mediated proteasome targeting of activated Jaks . STI-571/Imatinib (Gleevec, Novartis Pharma), at low concentrations, is a specific inhibitor of Abl kinase and shows effectiveness in the treatment of chronic myelogenous leukemia [27, 35]. ABPC20, ABPC22 and pre B v-Abl are much more sensitive to killing by STI-571 than TEPC1165 or TEPC2027 (Figure 3A), indicating that the Abl kinase activity is still required for viability of PCTs induced with the assistance of Abelson Virus. Since nearly all ABPCs also show c-Myc-activating chromosome translocations, some cooperation between the signaling pathways of v-Abl and c-Myc must be responsible for the rapid transformation of ABPCs. Unexpectedly, however, this cooperation appears to be required even after full transformation is achieved and even after adaptation to growth in culture. It will be interesting to see if resistance to STI-571 develops in these cultures, as is often seen in human CML .
Stat4 is more highly expressed in the BCL group than in PCTs. However, the level of stat3 expression does not differ significantly between BCLs and PCTs, both groups showing relatively high expression levels. Jak1 also showed higher expression in the BCL group compared to PCTs, but Jak1 is relatively highly expressed, even in PCTs. The accelerating mechanisms engaged after v-Abl infection seems to utilize these pathways (Figure 3B), despite the concomitant induction of the counteracting Socs family of genes. These pathways are currently being studied in greater depth at the translational and post-translational levels within the PCT system, following up the leads afforded by our gene expression studies and the initial phosphorylation studies shown here, with the goals of understanding the mechanisms at work.
It has been illuminating to analyze our mouse expression data in conjunction with already published Affymetrix data from human multiple myeloma. Cluster analysis showed that human MM1 clustered most closely with PCT4 and PCT5, IL6PC and KiPC, the two groups of PCTs from IL-6-transgenic mice, while the more aggressive myeloma groups, MM3-MM4, clustered more tightly with PCT1 and PCT2, ABLMYCPC and ABPC, those with appearance accelerated by v-Abl activity. This similarity includes differences in expression of genes associated with proliferation. This was unexpected but significant, because plasma cell neoplasms are not generally associated with rapid proliferation. Instead, increased survival or escape from apoptosis is thought to be the chief mechanism responsible for the expansion of lymphocytes or plasma cells in lymph nodes or bone marrow, respectively. This similarity brings to mind the possibility that Imatinib, the activated Abl inhibitor, might be effective in treating aggressive myeloma patients.
This co-clustering suggests that different pathways can be utilized to achieve a similar outcome, namely transformation of plasma cells. Thus, the mouse PCT model, despite its biological differences from MM, offers an experimental model for studying the details of the etiology of plasma cell neoplasms with different degrees of aggressiveness, much as seen in human myelomas. This aspect of our study will be broadened to include new data on additional myeloma patients  in which expression data are used to define seven subgroups that differ in their molecular characteristics. This study will be the subject of a separate manuscript.