The present study is the first to investigate HIV-1 evolutionary patterns among tissues from lymphoma patients and to demonstrate the existence of a distinct viral sub-population at least in part associated with TAMs. Although further work will be required to test the actual lymphomagenic potential of HIV lymphoma-associated quasispecies, our findings have important implications towards a thorough understanding of the nature of tumorigenesis and suggest that HIV-1 DNA could provide a molecular marker capable of tracking the spread of TAMs and thus tumor in vivo. These data may have important implications for studying the role of TAMs in the pathogenesis of cancer metastasis and as targets for macrophage-specific therapeutic development.
The two ARL patients examined died within 2 to 4 weeks of their diagnosis and phylodynamic analysis of HIV sequences from post mortem tissues revealed several interesting aspects of viral evolution: 1) both patients displayed remarkably similar patterns suggesting the existence of tumor and non-tumor-specific viral sub-populations, characterized by different growth rates; 2) the results consistently showed that while HIV-1 strains were highly compartmentalized between tumor and non-tumor tissues, strains from lymph nodes were interspersed within the two groups; 3) the presence of two different subpopulations of viruses in the patients with lymphoma is not the result of drug selection since both patients were HAART-naïve, and 4) HIV gene flow among lymphoid and non-lymphoid tissues with different histopathology was non-random. In particular, subject AM showed viral sequences from left and right axillary lymph nodes compartmentalizing, respectively, with tumor and non-tumor tissues.
It is important to notice that although compartmentalization tests clearly showed that the HIV-1 subpopulation infecting tumor tissues significantly differed from the one infecting non-tumor tissues, some intermixing can still be observed. The result is not unexpected. First, occasional exchange of virus (gene-flow) between contiguous tumor and non-tumor cells is possible, as shown by gene-flow analysis (). Second, mixed tissue viral branches that occur between the non-tumor and tumor clades may represent a biologically ancestral form of the tumor virus. For example, patient IV showed a population of lymph node, spleen and gastric wall sequences branching out of the tumor clade that likely represented an older population of sequences. Due to the high viral turnover in lymph nodes, a mixture of viral populations from these tissues is expected. In only one case do we see complete tissue compartmentalization, and this is with Patient IV, stomach, non-tumor sequences. Van Marle et al. showed that HIV sequences derived from the gut-associated lymphoid tissues are frequently compartmentalized from PBMC, and even from regions within the gut (stomach, esophagus, duodenum, colon)
. In this sense, it is not surprising that normal stomach is compartmentalized from the other tissues in the study, yet it is interesting that nearly half of the tissues from a gastric wall tumor were found within the tumor clade. If gut compartmentalization is common, then the only explanation for sequences in tumor clade to be found in the gastric wall is due to the migration and establishment of a metastatic macrophage-harboring virus from a site of tumorigenesis.
The finding of strongly compartmentalized populations of sequences isolated from different tissues, especially tumor-associated viruses within HAART-naïve ARL patients, suggests two potential biological mechanisms: 1) the virus found in ARL is migrating within metastatic macrophages and/or 2) lymphoma-associated strains of HIV exist. This finding is significant because the identification and characterization of lymphoma-specific viral strains may shed further light into the study of lymphomagenic processes. In particular, such viruses may contain specific protein structures or amino acid substitutions, or may induce expression patterns in macrophages that are associated with lymphoma onset and progression.
One characteristic of end-stage AIDS is the presence of p24 expressing macrophages along with T cell depletion 
. A recent AIDS autopsy study showed that lymph nodes, spleen and brain contained p24 expressing macrophages
. Additionally, previous studies of ARL tissues demonstrated that p24 staining within tumor tissues was predominantly localized to macrophages interspersed in a background of p24-negative lymphocytes. Laser-capture microdissection studies of these tissues localized integrated forms of HIV to TAMs 
. The histology from patients AM and IV pointed out that a subset of TAMs identified with CD68 staining were also positive for HIV p24, implying that at least some of the HIV sequences generated for this study were derived from tumor associated tissue macrophages. Such conclusion was strengthened as well by the predicted co-receptor usage for the strains amplified from tumor tissues; however, the possibility of harvesting lymphocytes from the tissues collected exists. In one case, viruses from tumor tissues did not map within the tumor clade (5 spleen tumor sequences in patient AM) and contained highly charged V3 loop sequences. It is possible that these represented viruses infecting lymphocytes present in the sample.
Overall, these findings seem to indicate that during lymphoma metastasis, the lymph nodes acted as the conduit of metastatic cells migrating to distinct sites in the body and that the lymph nodes exchanged virus with both non-tumor and tumor associated TAMs. It has been recently proposed that metastasis is initiated by hematopoietic progenitor cells (HPCs) and macrophages that co-migrate to future metastatic sites establish a favorable microenvironment for the seeding of tumor cells 
. These HPCs and macrophages mobilize in response to growth factors produced by the primary tumor, such as vascular endothelial growth factor A (VEGFA), placental growth factor (PlGF) and transforming growth factor-beta (TGFb), where they establish a pre-metastatic niche. Our data suggest the possibility that a lymphoma specific sub-population specifically infecting TAMs may promote lymphomagenesis. Over time, infected macrophages migrate to future secondary tumor sites where they likely establish a pre-metastatic niche and set up a favorable environment for ARL metastatic cells. Strikingly, we found that the viral effective population size (Ne
) in the tumor clade for both patients expanded ~100 fold (from ~1 to 100) during the course of infection, while the viral population in the non-tumor clade remained constant. Molecular clock analysis revealed that the exponential increase of the Ne
was not due to an increased evolutionary or replication rate of HIV-1 in tissues with different histopathology. The exponential increase in viral population size, on the other hand, is consistent with an increase number of macrophages within the tumor environment that are susceptible to HIV-1 infection. Human tumors, including lymphomas, secrete chemokines such as monocyte chemotactic protein-1 (MCP-1), macrophage colony stimulating factor (M-CSF), and VEGF, which are believed to be important for the recruitment of macrophages into tumor sites 
. Therefore, tumor-derived chemokines could be attracting macrophages into the tumor where they would be promoting tumor growth and metastasis through TAM's known pro-tumor properties as well as becoming infected with HIV-1. Further support for this hypothesis is that the majority of the gene flow in both patients was between tumor and lymph nodes, while almost no gene flow was observed between either tumor or lymph nodes and normal tissues. The continuous recruitment-infection of new macrophages in tumor tissues can explain the exponential growth of the tumor-associated viral quasispecies that occurred at the onset of ARL and progressed till patients' death, and suggest that a similar HIV-1 population dynamic in TAMs may underlie ARL onset and progression. One limitation of our study is that only two cases could be included. We believe, however, that the highly significant finding of HIV-1 compartmentalization in tumor and non-tumor tissues in two patients with independent clinical histories and the different numbers and sets of tissues analyzed should strengthen the confidence in our results.
To date, the evolution of HIV infection in macrophages has been primarily addressed in the context of the T-cell-free central nervous system. These studies have identified macrophages as long-lived reservoirs of HIV infection and replication and as sites for increased viral recombination. TAMs are not targeted by HAART. Most importantly, studies have shown that HIV infected macrophages are not prone to cell lysis whereas T-cells are. These facts make TAMs interesting in the context of how virus populations might persist and migrate in the context of macrophage-associated disease processes, such as lymphoma and dementia. The evolution of HIV within TAMs likely represents an entirely different epigenetic event than that found in T-cells and deserves increased attention in future studies.