Humanized mice are a promising model for large-scale studies of HIV transmission, and their characterization is critical to future studies. Our study describes human engraftment in GALT of RAG2−/−γc−/− mice into which have been transplanted cord blood-derived CD34+ cells. We show that, in this particular model, human cell numbers in the GALT are lower than in other lymphatic organs and increase only moderately after proinflammatory treatment. Furthermore, we examined the susceptibility of these humanized mice to rectal transmission of HIV and found a very small rate of rectal transmission of HIV. Only 2 of 56 challenged mice became HIV positive. Thus, this model may have limited value for the study of rectal transmission of HIV.
Two other studies showed successful rectal infections of humanized mice (4
), with transmission rates of 6/7 and 11/14, respectively. Sun et al. (37
) used NOD/SCID mice transplanted with fetal liver and thymus and later reconstituted with liver-derived CD34+
cells (BLT mice). Berges et al. (4
) transplanted RAG2−/−
mice with CD34+
cells from fetal liver. Thus, a major difference is that we used cord blood-derived cells. So far, it is not clear which source of hematopoietic stem cells is optimal for the generation of humanized mice. In vitro, the colony-forming efficiency of CD34+
cells derived from fetal liver is higher than that of cord blood CD34+
). However, in vivo cord blood-derived hematopoietic stem cells produced four times more mature human leukocytes than liver-derived cells, when the same number of stem cells was transplanted into adult NOD/SCID mice (18
). When telomere length is used as a measure of replicative capacity of hematopoietic stem cells from fetal liver and cord blood, the differences are minimal (40
), indicating that both are suitable for generating humanized mice. However, it is not known to what extent the origin of lymphoid cells may be critical for repopulation of the intestines with human cells. Different expression of cell adhesion molecules (30
) could influence homing behavior and subsequent repopulation of lymphoid organs. That, in turn, may explain the lower infection rates we observed in our model using cord blood-derived cells for transplantation. But when we reconstituted mice in parallel either with cord blood-derived or with fetal liver-derived cells, rectal transmission of HIV was not more efficient in the mice which had received liver-derived cells.
Apart from the different origins of CD34+ cells, Berges et al. and we used similar models based on mice with the same genetic background. However, Berges et al. reported higher engraftment levels than the ones we detected. They found blood engraftment levels of almost 90%, whereas the mice used in our study had a mean engraftment of about 11%. But these values cannot be directly compared. Berges et al. measured the percentage of human cells in blood lymphocytes; we measured the percentage of human cells in all blood leukocytes, which include murine cells as well. Since RAG2−/−γc−/− mice have no B, T, and NK cells of their own, obviously all lymphocytes detected in humanized mice should be of human origin. In our cohort of mice transplanted with liver-derived cells, we saw no differences in engraftment levels compared to engraftment levels in littermates transplanted with cord blood-derived cells. This indicates that the origin of the stem cells probably has no major impact, at least in our hands, and that other factors are influencing HIV transmission rates in humanized mice.
The intestinal microflora plays important roles in GALT development: intestinal bacteria influence the migration pattern of lymphocytes into mucosal sites; in germ-free animals, intestinal lymphocyte numbers are drastically decreased (38
). Certainly, the microflora in the gut of mice from different genetic backgrounds and from different animal facilities varies and could influence human engraftment in GALT and subsequently HIV transmission susceptibility. Besides potential differences in gut microflora, especially in BLT mice, gastrointestinal engraftment appears to be significantly better than the GALT engraftment we obtained in RAG2−/−
mice. BLT mice have the advantage of a human thymic tissue transplant, which could facilitate T-cell development.
In our humanized mouse model, T-lymphocyte development occurs in the anlage of the murine thymus, which is almost completely repopulated with human cells. Nonetheless, the stroma of this thymus is of murine origin and may have unknown disadvantageous effects on overall lymphoid development. Furthermore, to generate BLT mice, the NOD/SCID mouse strain was used. In contrast to RAG2−/−
mice, NOD/SCID mice do not have mutations or loss of γc
, the common gamma chain in receptors of multiple cytokines, including IL-7. Signaling through the IL-7 receptor is essential for formation of the Peyer's patches' anlagen (28
), and humanized mice lacking this important feature may have more difficulties in repopulating the GALT than mice that have Peyer's patches' anlagen.
To enrich for human cells of lymphoid origin in the rectal mucosa in our humanized mouse model, we induced local inflammation. Mucosal inflammation as seen in sexually transmitted diseases is a risk factor for HIV acquisition by increasing numbers of HIV target cells and disturbing epithelial integrity (14
). We treated humanized mice with Il-1β rectally 24 h before HIV challenge. In untreated mice, no infections could be seen; in treated mice, 1/17 showed a detectable plasma viral load. Immunohistochemical analysis of Il-1β-treated uninfected mice suggests that, despite an inflammatory response and cell infiltration, human cell numbers were probably still too low to establish infection. It is believed that HIV first infects a local founder population in the lamina propria of the mucosal tissue and amplifies there before systemic seeding and dissemination of the infection (16
). In our model, local expansion of infection is difficult with only a few human cells present in the lamina propria. We only observed disseminated infection when mice were injected i.p. with HIV, and the mucosal amplification step was not needed.
To attract human cells more efficiently to the GALT, we treated humanized mice with DSS, a compound widely used in murine models of inflammatory bowel disease. Repeated treatment with DSS leads to a chronic colitis with infiltration of lymphocytes and macrophages (21
). After three cycles of DSS treatment of humanized mice, we also saw infiltration of human cells into the rectal mucosa. Moreover, we confirmed colitis induction by monitoring weight changes and finally by histological scoring of rectal tissue. Mice tolerated treatment well up to 1 week after the last DSS administration, when a rapid decline in health occurred and half of the mice had to be euthanized. The HIV challenge could be a reason for the observed health decline. However, all mice were similarly affected, even surviving mice, which remained HIV negative. Notably, in non-DSS-treated mice, the same HIV challenge never elicited any symptoms. We therefore conclude that the DSS chronic colitis model is not suitable for rendering humanized mice permissive to rectal challenges with HIV.
It is still unknown whether cell-free or cell-associated HIV is preferably transmitted. To develop microbicides or vaccines, it is essential to know whether protection is needed against free virions, infected cells, or both. In simian and feline models, both cell-free and cell-associated virus transmission can be observed with different efficiencies, depending on the experimental design (9
). Results from studies in humans are conflicting: both free virions and infected cells are detected in cervicovaginal fluid (34
) and semen (41
). In cervicovaginal explants, which frequently are used for preclinical microbicide testing, cell-free HIV and cell-associated HIV are infectious (23
). Here, we detected only minimal transmigration of rectally applied mononuclear cells into the mucosa both in untreated and Il-1β-treated humanized mice. Further infection experiments with cell-associated HIV confirmed this observation. None of the mice exposed to HIV-infected PBMCs showed systemic viral replication, not even mice that had DSS colitis. Thus, cell-associated HIV transmission is, at least in our model, not more efficient than cell-free HIV transmission.
HIV strains with selective coreceptor use or even more subtle viral variants may differ in their abilities to establish infection by the mucosal route. In humans, CCR5-tropic HIV is transmitted preferably over CXCR4-tropic HIV (24
), and even in the group of CCR5 viruses, potential for transmission is diverse. Patients during acute infection show a more homogenous viral population, whereas patients in the chronic phase harbor many distinct variants (13
), indicating that only some of the viral variants in the transmitter are passed on. However, the characteristics of HIV variants preferentially transmitted are unknown so far. The CCR5-tropic HIV variant YU-2, which we used in our study for rectal challenge, was first isolated from neural tissue of a child suffering from AIDS (22
). There is some uncertainty whether it is easily transmitted by the mucosal route or not. In any case, YU-2 replicates well in humanized mice and establishes disseminated infection after i.p. injection (3
), and the three other HIV strains we tested in this study were not more efficient in rectal transmission of HIV. The other two studies (4
) demonstrating mucosal transmission of HIV in humanized mice showed that the mice were permissive to CXCR4-tropic HIV infection by the rectal route. However, Sun et al. challenged mice with HIV after mechanical disruption of the epithelial layer. It remains unknown whether the CXCR4-tropic viral strain used would have been transmitted otherwise. So far, we do not know whether the same bottleneck seen in humans for mucosal transmission of HIV exists in humanized mice.
In conclusion, our data indicate that GALT reconstitution in RAG2−/−γc−/− mice transplanted with CD34+ cells from cord blood is low and these mice seem to be quite resistant to rectal transmission of HIV, even in an inflammatory setting. Their value to study measures preventing mucosal transmission of HIV probably is limited. Further efforts are needed to clarify which mouse strain and transplantation protocol are best suited to generate the optimal humanized mouse.