Previous work shows that the conversion of conventional CD4+
T cells into Tregs is the predominant method of replenishment in tumor-bearing mice.27
The ability of conventional T cells to convert into Tregs is likely due to the expression of TGF-β. It has been demonstrated that high levels of TGF-β result in the conversion of CD4+
T cells into Tregs, which is abrogated with a neutralizing TGF-β antibody.18
Accordingly, Treg conversion takes place in a mouse model of pancreatic cancer by wild-type conventional CD4+
T cells converting into Tregs, but not by naive conventional CD4+
T cells expressing a dominant negative TGF-β receptor II.28
In contrast to these reports, recent evidence examining the T-cell receptor repertoire in a carcinogen-induced tumor mouse model has suggested that conventional CD4+
T cells and Tregs do not interconvert,29
which supports our observations demonstrating that brain tumor-resident Tregs are largely thymus derived and not a result of Treg conversion. Although further study is required to understand the nature of the conflicting reports regarding the origin of tumor-resident Tregs, it is possible that the stroma of the tumor is a critical factor in contributing to whether Tregs are thymus derived or converted from conventional CD4+
T cells. Given the anatomical specialization of the brain, including the presence of the blood-brain barrier (BBB), a difference in the composition of the stroma, and the lack of a conventional lymphatic drainage system,30
the mechanisms of Treg recruitment in brain tumors may differ from those found in many peripheral environments.
Although the brain tumor is encapsulated in the CNS by the BBB, it causes a systemic response throughout the mouse, as shown by the H&E and immunofluorescence staining in the spleen, thymus, and cervical dLNs. Our analysis at 3 WPO coincides, although it is a fluid process, with signs of weight loss, hunched posture, and a lack of normal physical activity. Furthermore, the smaller size of the follicles in the spleen of brain
tumor-bearing mice may be related to the generalized atrophy that occurs throughout the whole animal. Likewise, the cortex-medulla border of the thymus is no longer distinguishable in the thymus at 3 WPO. In contrast, both the high and low cervical dLNs enlarge and become easy to spot during dissection. Part of the reason for the systemic effects of the brain tumor on some of the lymphoid organs in the periphery may be related to the release of endogenous glucocorticoids, since it has been shown that adrenalectomy can partially prevent the apoptosis that leads to the rapid loss of thymocytes.31
Another possibility is that the direct innervation to lymphoid organs is compromised as a result of brain tumor compression on CNS-resident neurons. It has been well established that the thymus is innervated by neurons in the medulla of the brain as well as in the spinal cord.32
If projections to the thymus or neurons giving rise to those projections are compromised, it seems reasonable that this would be another mechanism by which thymic atrophy could occur.
Our work tested 2 mouse brain tumor models for the presence of Tregs. The first model was an orthotopic model by which GL261 cells were i.c. implanted. However, this model has intuitively inconvenient drawbacks. Most notably, by virtue of i.c. implantation, the BBB would be at least temporarily damaged and/or destroyed by the surgical drilling of the burr hole through the skull followed by a needle puncturing the dura of the brain. This may be the reason that we observe a low level of detectable CD4+
Tregs in the brain of mice i.c. implanted with saline at 1 WPO. Alternatively, it may be the result of a lack of perfusion prior to harvesting the brain tumor. However, given that Tregs are not detectable in the mouse brain i.c. injected with saline at 3 WPO, we think that the lack of perfusion is unlikely to be the cause of the low level of Treg infiltration. To confirm whether nTregs predominate in a non-orthotopic model, we utilized the spontaneously developing brain tumor, RasB8, transgenic mouse model.33
The RasB8 mouse model develops low- to high-grade astrocytoma without the need for i.c. injection of glioma cells or cellular transformation by carcinogens. In accordance with what we observed in the orthotopic mouse brain tumor model, Tregs were found to be present in the symptomatic RasB8 mouse brain. Most importantly, the Tregs in the brains of symptomatic RasB8 mice were almost universally thymus derived, as indicated by the coexpression of Helios by CD4+
T cells. More translationally relevant, we confirmed the predominance of thymus-derived Tregs in human GBM via immunofluorescent colocalization of FoxP3 and Helios. Collectively, the data suggest that nTregs predominate in both mouse and human brain tumors.
Since nTregs but not iTregs were the primary resident in brain tumors, we hypothesized that the depletion of nTregs would lead to a significant increase in lifespan. This possibility was tested by thymectomizing mice, with or without the Treg-depleting, CD25 mAb, prior to i.c. implantation of GL261 cells. We were surprised to find that combining thymectomy with CD25 mAb was not as effective in increasing the lifespan of brain tumor-bearing mice as was the administration of CD25 mAb alone. This could be due to the removal of the thymus, which not only participates in nTreg development but also contributes to the development of NK+
, conventional CD4+
, and CD8+
T cells, all of which have the potential to contribute to tumor eradication.34,35
Accordingly, recent thymic emigrants of CD8+
T cells are enriched in progressively growing T9 gliomas, further suggesting the importance of the thymus to contributing to other cells than nTregs during tumor progression.31
An alternative explanation is that thymectomy primes Tregs for a faster recovery of CD25 expression after CD25 mAb-mediated Treg depletion. It has previously been shown that the systemic administration of CD25 mAb decreases overall CD25 expression on lymphoid cells, which contributes to Treg inactivation.36
Here, we have confirmed the downregulation and/or loss of CD25 on CD4+
Tregs after systemic CD25 mAb treatment in vivo (Supplementary Fig. S4
). Interestingly, when treated with the CD25 mAb, fewer Tregs in the dLN and ndLN from nontumor bearing unthymectomized mice expressed high levels of CD25 compared with thymectomized mice at 3 WPO. This finding suggests that thymectomy primes Tregs for a faster recovery of high CD25 expression. While we did not study the detailed kinetics of this regulatory mechanism, it is possible that the faster renewal of CD25 on the surface of Tregs in thymectomized mice is associated with a higher level of suppressor function, since CD25 is critical for interleukin-2 signaling and consequent Treg function.37
If this were true, then it would likely be a contributing factor to the faster time of death of thymectomized mice treated with CD25 mAb compared with unthymectomized mice with brain tumors.
One of the unexpected effects of i.c. injection with saline and/or GL261 cells was the low levels of Tregs in the thymus at 1 WPO followed by an increase in Treg levels at 3 WPO in mice not treated with CD25 mAb. We were unable to find other studies that had analyzed thymic Treg levels at comparable time points in mouse models after injury. While we can only speculate at this time what causes the increase in Treg accumulation in the thymus at the latter time point post-injection, it may be related to injury-associated glucocorticoid signaling or an increase in thymic stromal lymphopoetin and/or keratinocyte growth factor (fibroblast growth factor 7), both of which have been shown to affect Treg development in the thymus.38–40
It would be interesting to test the effects of adrenalectomy on the injury-associated Treg increase in the thymus. However, to rule out other factors that may be playing a role in Treg development after injury, a global expression analysis comparing factors that govern Treg development between the 1- and 3-WPO time points is planned. Certainly, this could have important implications for future therapies that are not specific to brain tumor burden, but it may also have ramifications for autoimmune-, allergic-, and/or transplant-related pathogenesis.
Although thymus-derived Tregs compose the predominant population in brain tumors, it was interesting to find that they also proliferate at higher levels than iTregs or conventional CD4+
T cells, which was also recapitulated in the dLNs. However, based on our experiments in thymectomy, CD25 mAb, and brain tumor, we propose that nTregs emigrate from the thymus to the dLNs before infiltrating the brain tumor. Thus, it is possible that the higher level of Ki67+
nTregs that was observed both in the dLN and in the brain tumor may simply be a result of expansion in the dLN and remnant expression of Ki67 after transit to the site of the tumor. Furthermore, we believe that the gradual decrease in nTreg levels in the dLN between 1 and 3 WPO occurs as a result of contraction to normal nTreg levels, which has been established to be ~60%.21
Moreover, recent work has demonstrated the remarkable stability and continuous self-renewal that contributes to the Treg lineage.41
Our data support this observation and suggest that the population of Tregs that undergo the highest level of self-renewal is thymus derived, since we showed increased levels of CD4+
cells that coexpressed Ki67 compared with CD4+
cells. Furthermore, our data reciprocally support recent work demonstrating an increased level of Tregs that express Ki67 in the tumor infiltrates of primary human breast tumors,42
as well as work that demonstrates increased intratumoral Treg proliferation in a peripheral mouse tumor model.43
Thus, the presence of proliferative tumor-resident Tregs appears to extend beyond brain tumors and is likely a feature of both central and peripheral compartmental malignancies, although our data indicate that thymus-derived Tregs are the primary proliferative CD4+
T cell population in brain tumors and dLNs.
Although the Treg developmental program is regulated by FoxP3, genes regulated by FoxP3 demonstrate plasticity depending on the tissue context.43
Previous work has demonstrated that GITR and CD62L are expressed at higher and lower levels, respectively, on Tregs, while the level of CD103+
Tregs is increased in dLN of tumor-bearing mice, relative to tumor-free mice.27
Here, we have extended those observations to both Tregs and conventional CD4+
T cells and compared the spleen and brain in tumor-bearing mice. We showed that all of the brain-resident Tregs expressed the co-inhibitory ligand, GITR, while only ~50% of spleen-resident Tregs express it. Interestingly, a significantly increased fraction of brain-resident conventional CD4+
T cells expressed GITR compared with those cells in the spleen. A similar trend was observed for CD103. Likewise, there were fewer CD45RB-expressing Tregs in the brain compared with spleen. Also, fewer conventional CD4+
T cells expressed CD45RB compared with those in the spleen. Collectively, these data suggest that the brain tumor microenvironment amplifies the lineage programming of Tregs. Moreover, conventional CD4+
T cells are also affected by the brain tumor environment, which may be the cause for their recently noted plasticity.
While this is the first report that we know of demonstrating the predominance of nTregs in brain tumors, we plan to revisit other experimental tumor models that have been suggested to consist predominantly of iTregs, to confirm their origin, the comparative numbers of Tregs expressing GITR, CD62L, CD103, and CD45RB, as well as the proliferative status. It would be interesting to find that the tumor microenvironment within the CNS programmed Tregs differently than in subcutaneous tumor models in the peripheral nervous system. Already we have shown that a distinct subset of brain-resident Tregs, in vivo, coexpress the pro-inflammatory cytokine interleukin-17A.44
This is a surprising finding given that almost all of the brain tumor-resident Tregs are nTregs and previous data demonstrating the cytokine-producing stability of nTregs when compared with iTregs.21
However, it is important to point out that the conditions by which cytokine production occurs in nTregs and iTregs in vitro are unlikely to be recapitulated by the tumor microenvironment. To confirm this suggestion, we are currently engaged in tracking the stability of nTreg conversion in vivo utilizing FoxP3–interleukin-17 coreporter mice.
In summary, this work identifies that the thymus-derived nTreg is the predominant type of Treg in human GBM and experimental mouse brain tumors. A working model describing our data in the context of the current literature is presented in Figure . Functionally, we tested the relevance of nTregs in a mouse brain tumor model, using thymectomy and CD25 mAb and found that the thymus is critical for the beneficial effects of CD25 mAb. Furthermore, more brain-resident Tregs express GITR and CD103 compared with spleen-resident Tregs in tumor-bearing mice. In contrast, fewer brain-resident Tregs express CD62L and CD45RB compared with spleen-resident Tregs in tumor-bearing mice. Finally, we showed that more nTregs express Ki67 in the brain and cervical dLNs compared with iTregs and conventional CD4+ T cells. While most of the data that we have presented here are representative of the Treg origin in a mouse brain tumor model, we believe that these data have implications for the clinical treatment of GBM, given that we confirmed that nTregs predominate in human GBM. First, Tregs will need to be targeted in the periphery, where nTreg development and expansion predominates. Likewise, Tregs may need to be targeted in the brain, since brain tumor patients will present with Tregs already infiltrating the malignancy, in addition to the proliferative status. This will require chemotherapeutics and/or immunotherapeutic modalities that are capable of freely passing through the BBB, as well as targeting both proliferating tumor cells and nTregs. Coincidentally, these studies are currently ongoing in a clinical trial, NCT00626483, which is testing the combined efficacy of temozolomide and daclizumab in patients with GBM. The therapeutic effect of temozolomide depends on the ability to alkylate/methylate DNA, triggering tumor cell death, while daclizumab is a humanized anti–interleukin-2 receptor antibody. Our data support the rationale for this combinatorial strategy as well as the development of future strategies whereby both reagents readily cross the BBB.
Fig. 8. Model of Treg recruitment in brain tumors. Common lymphoid progenitor (CLP) cells develop in the bone marrow and migrate to the thymus. In the thymus, CLP cells can terminally differentiate into naive conventional CD4+ T cells (light green cell : white (more ...)