Foxp3-deficient mice have an otherwise intact immune system, and they lack systemic inflammation in the early neonatal period. These features of Foxp3 deficiency, together with the aggressive nature of the autoimmune disease that subsequently develops in untreated mice, made this an ideal model for our studies designed to examine the role of in vivo-derived iTreg cells in tolerance induction. The studies herein identify iTreg cells as an essential, non-redundant regulatory subset that acts synergistically with nTreg cells to enforce peripheral tolerance.
Although nTreg cells were provided by adoptive transfer immunotherapy shortly after birth, we found that the frequency of in vivo derived iTreg cells was substantial, as it was in the lymphopenia-colitis model (Haribhai et al., 2009
). We also found that iTreg cells generated in vitro through a TGFβ-dependent pathway largely substituted for those iTreg cells derived in vivo, suggesting that the two sources of iTreg cells were functionally equivalent. We did not directly examine the TGFβ dependence of the in vivo production of iTreg cells in this model, so a significant contribution from the TGFβ-independent CD25+
Treg cell precursor population cannot be excluded (Schallenberg et al., 2010
). Indeed, the recovery of half as many iTreg cells after 50 days, when the iTreg cells were derived in vitro rather than in vivo, might argue that more than one iTreg production pathway is operational. Regardless of the mechanisms supporting iTreg cell induction, these data reinforce the conclusion that the iTreg cell induction pathways are important tolerogenic mechanisms, both in the setting of chronic inflammation and in normal individuals.
Our initial investigations into the mechanisms underpinning the functional synergy between iTreg and nTreg cells focused on comparing the gene expression profiles of the two cell types, which were found to be remarkably similar. One explanation for this similarity is that each transcriptional signature is a composite of several Treg cell subsets, which could tend to blur distinctions between the two Treg cell types (Feuerer et al.). Nevertheless, the current data argue strongly that iTreg and nTreg cells share overlapping effector mechanisms. Gene expression profiles do not exclude a skewed reliance on certain Treg cell mediators like IL-10, which showed 2.2 fold increase in the in vivo derived iTreg cells over nTreg cells isolated from the same mice. Other potential distinctions between the two Treg cell subsets include increased expression of PD-1 by iTreg cells and Gzm B by nTreg cells, and differences in localization. We also found that the transcriptional signature of iTreg cells produced in vivo was more similar to nTreg cells than to iTreg cells produced in vitro. The most likely explanation for the differences in gene expression between iTreg cells produced in vivo and in vitro is that the profile of iTreg cells generated in vitro is largely a function of TCR, TGFβ and IL-2-mediated signaling, at least shortly after induction (Hill et al., 2007
). Indeed, the gene expression profile of iTreg cells produced in vitro 72 hours after induction is independent of Foxp3 (Haribhai et al., 2009
). We have not yet examined the transcriptome of iTreg cells produced in vitro and subsequently allowed to achieve equilibrium in treated mice. Based on the current results, this gene expression profile is anticipated to closely match that of both iTreg cells derived in vivo and nTreg cells.
If iTreg and nTreg cells share similar effector mechanisms, as suggested by their overlapping gene expression signatures, then other factors must contribute to the capacity of iTreg cells to complement the function of nTreg cells. For example, a synergistic interaction could develop if iTreg cells provided different TCR specificities than those found in nTreg cells, as we observed. These results shed light on the origins of in vivo-derived iTreg cells. Published data demonstrates that some nTreg cells can lose Foxp3 expression and survive as ” ex-Treg” cells (Yang et al., 2008
; Zhou et al., 2009
). In our experiments, such cells would be sorted with the Tconv cells and might re-express Foxp3 after transfer (Lathrop et al., 2008
). A substantial overlap between nTreg and iTreg cells could therefore indicate that the two populations were clonally related, implying that iTreg cells were “ex-Treg” cells that reacquired Foxp3 expression and suppressive capability under the conditions of our experiment. Similarly, iTreg cells could be a special population of Tconv cells poised to express Foxp3 based on their shared TCR specificity with nTreg cells (Schallenberg et al., 2010
). Again, substantial TCR repertoire overlap would be predicted if this shared specificity was created at the clonal level during the cellular expansion that occurs following β-selection in the thymus. From another perspective, iTreg cells might come from those Tconv cells that bear no clonal relationship to nTreg cells, based on the well-documented differences in the TCR ligand affinity requirements for the selection of nTreg and Tconv cells in the thymus (Jordan et al., 2001
). Here, the predicted result would be essentially no TCRβ CDR3 overlap between the two populations, which is largely what we observed, particularly when considering that peripheral Treg cells probably contain a mixture of nTreg and iTreg cells. A recent estimate places the frequency of iTreg cells in the peripheral Treg compartment at 4–7% (Lathrop et al., 2008
). We estimate that the iTreg cell frequency in rescued Foxp3 mutant mice is closer to 10–15%. Such misidentified iTreg cell TCRs would create an apparent overlap in the iTreg versus nTreg cell TCR comparisons and may well be a contributor to the small overlap observed in our studies. Thus our work is most consistent with naturally arising nTreg and iTreg cell populations with distinct TCR repertoires, indicating the two populations are clonally unrelated.
Since there are theoretically 1015
different TCRs that can be generated, the <1×106
different TCRβ chains found within an individual mouse virtually assures that these randomly selected repertoires are largely unique to each individual (i.e. “private”) (Quigley et al., 2010
). Thus we expected that there would be little overlap in TCRβ CDR3 sequences from mouse to mouse, as we observed. However, a much higher frequency of the overlapping sequences in our experiments are “public”, which could indicate that these were the most abundant clones in the donor mice, that iTreg and nTreg cells share clonal identity, and that the repertoires are largely overlapping. Several lines of evidence argue against this point of view. First, many private nTreg and iTreg sequences were equally abundant, based on the frequency of their recovery. Thus abundant clones are not restricted to the overlapping component of the repertoires. Second, if one source of overlap derives from misidentified iTreg cells within the transferred nTreg pool, these clones may be numerically advantaged at the time of transfer, relative to new iTreg cells that arise in transfer recipients. Furthermore, misidentified iTreg clones that are present in the nTreg cell inoculate may be particularly important, since they arise in healthy mice. It follows that these clones would be found frequently in the overlapping component of the Treg repertoires, and that they would also be more likely to be public if they fill a ubiquitous antigen-specific niche left vacant by nTreg cells. Indeed, public sequences are seen in well-characterized antigen-specific CD4+
Tconv cell responses (Kedzierska et al., 2004
; Menezes et al., 2007
A different explanation for the overlapping “public” CDR3 sequences deserves special attention, because it has little to do with antigen specificity. It has been proposed that an essential characteristic of “public” CDR3 sequences is that they are generated by “convergent recombination” (Venturi et al., 2008
). In this proposal, the number of different ways that a particular TCR CDR3 amino acid sequence can be generated by germline recombination is an important determinant of TCR production frequency and therefore of TCR sharing, quite apart from the antigen specificity of the TCR (Quigley et al., 2010
). In other words, certain TCRβ amino acid sequences are more likely to be “public” because they are more frequently generated in the thymus. In our studies, public TCRβ sequences generated by convergent recombination would also be more likely to be overlapping, since each new TCRβ sequence is paired with a different TCRα chain, generating a new TCR and an independent opportunity for clones to be selected into either the iTreg or nTreg pool. Indeed, we found that the most frequently identified public, overlapping TCRβ chain amino acid sequence was generated by 7 unique nucleotide sequences. Some of our data therefore supports this latter model.
It is important to note that in our experiments, the size and complexity of both Treg cell TCR repertoires is fixed, since there can be no new production of Foxp3+ cells by the host. This experimental design may have expanded the opportunity to observe the generation of iTreg cells in vivo as well as the expansion of a few high rank clonotypes. Most of our TCR repertoires fit a two-component model, with a power law-like component and a component consisting of a few high rank clonotypes. This indicates that peripheral selection events occur after adoptive transfer. Given the repertoire differences described herein, coupled with the profound differences in the thymic selection requirements of nTreg and Tconv cells, it seems very unlikely that iTreg and nTreg cells will recognize identical sets of self-antigens and foreign antigens and that they will do so at comparable frequencies. Nevertheless, we make no a priori claim about the antigen specificity of the responses analyzed, and the reader should draw a distinction between diversity (unique TCR sequences) and specificity (antigens recognized).
In rescued mice that received both nTreg and either Thy1.1+
Tconv or Thy1.1+
in vitro generated iTreg cells, a substantial number of Thy1.1+
cells were maintained long-term. There are a few possibilities for the origin of these cells. A bidirectional linear model where Tconv cells gain and lose Foxp3 expression based on the needs of the host has been proposed and is consistent with the data (Haribhai et al., 2009
). In a variation of this view, some Treg cells are not terminally differentiated and can adopt alternative TH
cell fates (Zhou et al., 2009
), or co-express other TH
lineage specification factors that alter their distribution and homeostatic properties (Koch et al., 2009
). It will be important to determine the cytokines produced by these Thy1.1+
cells, to compare their TCR repertoire with that of iTreg and nTreg cells, and to use genetic tools to map cell fates in future experiments. Rational design of adoptive transfer immunotherapy with iTreg cells hinges upon this type of information.