To examine the consequences of activated CD4⁺ T cell deletion for immune homeostasis, we generated a genetic mouse model in which activated CD4⁺ T cells were killed in the absence of retroviral infection. Activated CD4⁺ T cells were targeted by conditional gene activation mediated by Tnfrsf4-driven Cre recombinase expression 26. Specificity for CD4⁺ T cells was confirmed by activation of a yellow fluorescent protein (YFP) reporter gene in the Gt(ROSA)26Sor (R26) locus 27, which revealed that about 95% of YFP⁺ cells were CD4⁺ T cells (Figure 1a, b). YFP expression in Tnfrsf4^Cre/+ R26^Yfp/+ mice marked 55% and 80% of memory and regulatory CD4⁺ T cells, respectively (Figure 1c). In contrast, the vast majority of naïve CD4⁺ T cells (92%) and naïve and memory CD8⁺ T cells (99% and 97%, respectively) were YFP^-(Figure 1c, d). Conditional deletion of CD134-expressing CD4⁺ T cells was achieved by Cre-mediated activation of a gene encoding diphtheria toxin fragment A (DTA) independently targeted into the R26 locus (Additional data file 1).

The efficiency of DTA-mediated T cell deletion was assessed in Tnfrsf4^Cre/+ R26^Yfp/Dta heterozygous mice. In comparison with Tnfrsf4^Cre/+ R26^Yfp/+ mice, the proportion of YFP⁺ memory and regulatory CD4⁺ T cells in Tnfrsf4^Cre/+ R26^Yfp/Dta mice was reduced by more than half (Figure 1e), suggesting that more than 50% of the cells that were tagged with YFP in the absence of DTA expression were killed on DTA activation. However, this analysis ignored the dynamic nature of T cell death and replacement. The relative presence of activated CD4⁺ T cells and proportion of YFP⁺ T cells in Tnfrsf4^Cre/+R26^Yfp/Dta mice reflected equilibrium between DTA-mediated killing, which would reduce, and homeostatic replacement, which would increase, the number of YFP⁺ activated CD4⁺ T cells, in addition to the relative kinetics of YFP and DTA induction following T cell activation. In vitro activated purified CD134^-YFP^- naïve CD4⁺ T cells from Tnfrsf4^Cre/+ R26^Yfp/+ mice began to express YFP by the first day of culture, with a delay of about 1 day relative to CD134 induction (Figure 1f, g). However, the effect of DTA activation on survival of in vitro activated naïve CD4⁺ T cells from Tnfrsf4^Cre/+ R26^Dta/+ mice was not evident until the second day of culture (Figure 1h).

To determine whether activated CD4⁺ T cell deletion had any impact on lymphocyte population dynamics, we analyzed lymphoid organ cellularity and composition. We observed significant systemic lymph node enlargement in Tnfrsf4^Cre/+ R26^Dta/+ mice compared with control Tnfrsf4^Cre/+ R26^+/+ mice (Figure 2a), which was also associated with elevated serum levels of several proinflammatory mediators (Figure 2b). Spleen size was not appreciably affected (Figure 2a). We thus calculated the total size of lymphocyte and myeloid populations as the sum of the cellular contents of the spleen and of inguinal, axillary, brachial, mesenteric and superficial cervical lymph nodes. B cells, but not T cells or myeloid cells, were significantly more numerous in Tnfrsf4^Cre/+ R26^Dta/+ mice than in control Tnfrsf4^Cre/+ R26^+/+ mice (Figure 2c).

Deletion of activated CD4⁺ T cells resulted in a substantial systemic drop in the CD4:CD8 ratio (Figure 2d). Remarkably, compared with control mice, total CD4⁺ T cell numbers in Tnfrsf4^Cre/+R26^Dta/+ mice were only marginally reduced and remained stable throughout a 6-month observation period (Figure 2e). To assess whether activated CD4⁺ T cell numbers were selectively reduced in Tnfrsf4^Cre/+ R26^Dta/+ mice, we determined the composition of the CD4⁺ T cell pool. Numbers of naïve CD4⁺ T cells and, notably, of memory CD4⁺ T cells were similar between Tnfrsf4^Cre/+ R26^Dta/+ and control Tnfrsf4^Cre/+ R26^+/+ mice (Figure 2e), whereas numbers of regulatory CD4⁺ T cells were reduced by about 40% in Tnfrsf4^Cre/+ R26^Dta/+ mice (Figure 2e). In contrast to CD4⁺ T cells, total numbers of CD8⁺ T cells were elevated in Tnfrsf4^Cre/+ R26^Dta/+ mice compared with Tnfrsf4^Cre/+ R26^+/+ mice, resulting from a systemic expansion exclusively of memory CD8⁺ T cells (Figure 2e), which was primarily responsible for the systemic reduction in the CD4:CD8 ratio.

Preservation of CD4⁺ T cell numbers despite killing of activated CD4⁺ T cells in Tnfrsf4^Cre/+ R26^Dta/+ mice suggested increased replenishment, which would be associated with functional and phenotypic activation. Phenotypic differences between naïve or memory CD4⁺ T cells in Tnfrsf4^Cre/+ R26^Dta/+ and those in control Tnfrsf4^Cre/+ R26^+/+ mice were largely unremarkable, with modest increases in expression of cytokines and of the activation markers CD43 and CD49b in memory CD4⁺ T cells isolated from Tnfrsf4^Cre/+ R26^Dta/+ mice (Figure 3a, b). In contrast to regulatory CD4⁺ T cells from control mice, those from Tnfrsf4^Cre/+ R26^Dta/+ mice showed a highly activated phenotype, characterized by downregulation of CD62L and upregulation of CD44, CD43, CD49b and CD103 (Figure 3c). Thus, DTA-mediated destruction of activated CD4⁺ T cells had a significant effect on regulatory CD4⁺ T cell numbers and activation state, but little apparent effect on memory CD4⁺ T cells.

Memory CD4⁺ T cells, under physiological conditions, display higher turnover rates, self-renewal potential and activation profile than either naïve or regulatory CD4⁺ T cells 1. Indeed, naïve CD4⁺ T cells from either Tnfrsf4^Cre/+ R26^Dta/+ or control mice showed little evidence for cell division assessed either by incorporation of bromodeoxyuridine (BrdU) or staining with the Ki67 antibody (Figure 4a). In contrast, population turnover rates were very high in memory CD4⁺ T cells from both Tnfrsf4^Cre/+ R26^Dta/+ and control mice (Figure 4b). Ki67 staining, but not BrdU incorporation, in memory CD4⁺ T cells from Tnfrsf4^Cre/+R26^Dta/+ mice was elevated in comparison with that in memory CD4⁺ T cells from control mice (Figure 4b). Moreover, regulatory CD4⁺ T cells had a significantly higher turnover rate in Tnfrsf4^Cre/+ R26^Dta/+ mice than in control mice, which approached the high turnover rate of memory CD4⁺ T cells (Figure 4c).

Comparable representation of BrdU⁺ memory CD4⁺ T cells in both Tnfrsf4^Cre/+ R26^Dta/+ and control mice at the end of the 6-day BrdU pulsing period (Figure 4b) seemed discordant with the expected elevated turnover of memory CD4⁺ T cells in Tnfrsf4^Cre/+ R26^Dta/+ mice. However, in memory CD4⁺ T cells from Tnfrsf4^Cre/+ R26^Dta/+ mice, increased BrdU incorporation would be masked by increased DTA-mediated death of the proliferating cells during the pulsing period. We therefore examined the fate of BrdU⁺ memory CD4⁺ T cells 3 days after termination of BrdU administration (chase period). In contrast to the opposing action of cell proliferation and death, which would increase or decrease, respectively, the percentage of BrdU⁺ cells during the pulsing period, cell proliferation, by dilution of BrdU label, and cell death would both decrease the percentage of BrdU⁺ cells during the chase period. Indeed, almost twice as many BrdU⁺ memory CD4⁺ T cells were lost during the 3-day chase period in Tnfrsf4^Cre/+R26^Dta/+ mice as in control mice (Figure 4d). Consistent with the Ki67 staining, the difference in the loss of BrdU⁺ cells was even more pronounced in regulatory CD4⁺ T cells (Figure 4d). These data together suggested that memory CD4⁺ T cells, and to a higher degree regulatory CD4⁺ T cells, had significantly elevated turnover rates in Tnfrsf4^Cre/+ R26^Dta/+ mice than in control mice.

To further reveal the full extent of memory CD4⁺ T cell killing in Tnfrsf4^Cre/+ R26^Dta/+ mice, we generated mixed bone marrow chimeras. Compared with immunodeficient mice reconstituted with wild-type bone marrow alone, those reconstituted with Tnfrsf4^Cre/+ R26^Dta/+ bone marrow alone showed a paradoxical increase in memory CD44⁺CD4⁺ T cells and a small reduction in regulatory CD25⁺CD4⁺ T cells (Figure 4e). In contrast, mice reconstituted with a mixture of wild-type and Tnfrsf4^Cre/+ R26^Dta/+ bone marrow had a severe reduction in memory and regulatory CD4⁺ T cell numbers of Tnfrsf4^Cre/+ R26^Dta/+ origin (Figure 4e). Comparison of the Tnfrsf4^Cre/+R26^Dta/+:wild-type ratio in thymocyte and peripheral lymphocyte subsets confirmed a significant selective loss of Tnfrsf4^Cre/+ R26^Dta/+-origin memory and regulatory CD4⁺ T cells (Additional data file 2). Thus, although both subsets were being killed with equal efficiency by DTA activation in Tnfrsf4^Cre/+ R26^Dta/+ mice, regulatory CD4⁺ T cells were incompletely replenished, whereas the loss of memory CD4⁺ T cells was overcompensated.

Continuous replenishment of memory CD4⁺ T cells in Tnfrsf4^Cre/+ R26^Dta/+ mice could be due to constant recruitment of new cells from either memory CD4⁺ T cells that did not express CD134 or naïve CD4⁺ T cells, thymic production and peripheral numbers of which were minimally affected in these mice. To examine the contribution of thymic T cell production and of naïve CD4⁺ T cells to the preservation of memory CD4⁺ T cell numbers in Tnfrsf4^Cre/+ R26^Dta/+ mice, we infused purified CD4⁺ T cells into lymphopenic recipients (adoptive transfer). Rag1^-/- mice, which are lymphopenic due to genetic deficiency in the V(D)J recombination activation gene Rag1 that precludes generation of lymphocytes, were chosen as recipients. Adoptive transfer of Tnfrsf4^Cre/+ R26^Yfp/+ CD4⁺ T cells revealed that T cell proliferation in lymphopenic Rag1^-/- recipients was sufficient to drive full activation, as nearly all of the transferred CD4⁺ T cells expressed the YFP reporter within 3 weeks of transfer (Figure 4f). In contrast, the proportion of YFP⁺ cells in transferred Tnfrsf4^Cre/+ R26^Yfp/Dta CD4⁺ T cells that could activate both YFP and DTA remained low throughout the 7-week observation period (Figure 4f), similar to the low percentage of YFP⁺ memory CD4⁺ T cells found in donor Tnfrsf4^Cre/+R26^Yfp/Dta mice (Figure 1e). More importantly, in contrast to Tnfrsf4^Cre/+ R26^Yfp/+ CD4⁺ T cells, which progressively expanded over time in Rag1^-/- recipients (Figure 4g), numbers of transferred Tnfrsf4^Cre/+R26^Yfp/Dta CD4⁺ T cells remained low in the blood throughout the observation period (Figure 4g) and in the lymphoid organs of Rag1^-/- recipients 7 weeks after transfer (Figure 4h). Failure of Tnfrsf4^Cre/+ R26^Yfp/Dta memory CD4⁺ T cells to accumulate in the setting of naïve CD4⁺ T cell deficiency demonstrated the requirement for recruitment of naïve CD4⁺ T cells in order to maintain the pool of memory CD4⁺ T cells.

To directly visualize the recruitment of naïve CD4⁺ T cells into the memory pool of Tnfrsf4^Cre/+ R26^Dta/+ mice, we adoptively transferred purified carboxyfluorescein succinimidyl ester (CFSE)-labeled naïve (CD44^loCD25^-) CD4⁺ T cells from wild-type donor mice expressing the allotypic marker CD45.1. As expected, naïve CD4⁺ T cells failed to proliferate or activate within 6 days of transfer into control Tnfrsf4^Cre/+ R26^+/+ mice (Figure 4i). In contrast, a substantial proportion of the progeny of transferred naïve CD4⁺ T cells had divided extensively and acquired CD44 expression during the same time in Tnfrsf4^Cre/+R26^Dta/+ mice (Figure 4i). Together, these observations support a model in which both memory and regulatory CD4⁺ T cells are being killed with equal efficiency by expression of DTA in Tnfrsf4^Cre/+ R26^Dta/+ mice. However, as long as production and maintenance of naïve CD4⁺ T cells is unaffected, the continual loss of memory but not regulatory CD4⁺ T cells can be efficiently compensated for by continual recruitment of naïve CD4⁺ T cells.

To investigate whether the accelerated death and replenishment of memory CD4⁺ T cells compromised immune competence, we evaluated CD4⁺ T cell function in Tnfrsf4^Cre/+ R26^Dta/+ mice. In comparison with Tnfrsf4^Cre/+ R26^+/+ littermates and C57BL/6 (B6) control mice, which showed a strong neutralizing antibody (nAb) response to and effectively contained infection with Friend virus (FV), a retrovirus that causes persistent infection in mice, the FV-specific nAb response was undetectable in four out of seven Tnfrsf4^Cre/+R26^Dta/+ mice and significantly delayed in the rest (Figure 5a). Defective FV-specific nAb response also correlated with inability of Tnfrsf4^Cre/+ R26^Dta/+ mice to control FV replication (Figure 5b). Furthermore, following acute influenza A virus (IAV) infection, titers of CD4⁺ T cell-dependent virus-neutralizing antibodies (nAbs) in the serum of Tnfrsf4^Cre/+R26^Dta/+ mice were reduced to 48% and to 36% of those in Tnfrsf4^Cre/+ R26^+/+ littermates and B6 mice, respectively (Figure 5c). Titration of CD4⁺ T cells into T cell-deficient mice revealed that this degree of reduction in IAV-specific nAb titers corresponded to about 80% loss of CD4⁺ T cells (Additional data file 3). Lastly, in contrast to Tnfrsf4^Cre/+R26^+/+ and B6 mice, which effectively cleared experimental infection with Pneumocystis murina, Tnfrsf4^Cre/+R26^Dta/+ mice remained persistently infected and failed to thrive, with susceptibility intermediate between T cell-replete and MHC II-deficient mice, which lacked CD4⁺ T cells (Figure 5d). These results showed that, despite the presence of normal numbers of CD4⁺ T cells, Tnfrsf4^Cre/+ R26^Dta/+ mice were CD4⁺ T cell immune deficient.

CD8⁺ T cells in HIV infection are increased in numbers and also have a higher turnover rate and activation state 245. This could result from either a composition change from quiescent naïve to activated memory CD8⁺ T cells and/or further activation within the memory pool. Following in vitro stimulation of total spleen and lymph node cells, the expanded population of memory CD8⁺ T cells in Tnfrsf4^Cre/+R26^Dta/+ mice showed a pattern of cytokine production similar to that in control mice (Figure 6a), indicating that they were functionally intact. Compared with the relatively homogeneous memory CD8⁺ T cell pool in control mice, memory CD8⁺ T cells in Tnfrsf4^Cre/+ R26^Dta/+ mice contained higher numbers of recently activated/effector CD8⁺ T cells, characterized by downregulation of CD62L and upregulation of CD43 expression (Figure 6a). To examine whether or not the expansion of the memory CD8⁺ T cell pool in Tnfrsf4^Cre/+ R26^Dta/+ mice was due to recruitment of a greater fraction of naïve CD8⁺ T cells, we adoptively purified CD45.1⁺ naïve (CD44^lo) CD8⁺ T cells labeled with CFSE from wild-type donor mice. Naïve CD8⁺ T cells remained undivided, as evident from the retention of CFSE label 6 days after transfer into either Tnfrsf4^Cre/+ R26^Dta/+ or control mice (Figure 6b). In contrast, naïve CD8⁺ T cells divided extensively in lymphopenic Rag1^-/- recipients (Figure 6b). This finding suggested that naïve CD8⁺ T cells were not constantly recruited into the memory pool of Tnfrsf4^Cre/+ R26^Dta/+ mice. The proportion of proliferating cells in memory CD8⁺ T cells, assessed by BrdU incorporation during a 6-day administration period, was similarly high in Tnfrsf4^Cre/+ R26^Dta/+ and control mice (Figure 6c). However, because of the difference in memory CD8⁺ T cell pool size, absolute numbers of BrdU⁺ memory CD8⁺ T cells were significantly higher in Tnfrsf4^Cre/+ R26^Dta/+ mice than in control mice (Figure 6d). Thus, the expanded memory CD8⁺ T cell population in Tnfrsf4^Cre/+ R26^Dta/+ mice seemed to be maintained by elevated turnover within the memory pool.

We next confirmed that expansion and activation of memory CD8⁺ T cells in Tnfrsf4^Cre/+ R26^Dta/+ mice was a consequence of accelerated turnover of activated CD4⁺ T cells. In mice reconstituted with a mixture of wild-type and Tnfrsf4^Cre/+R26^Dta/+ bone marrow, in which CD4⁺ T cell homeostasis is restored by wild-type CD4⁺ T cells (Figure 4e), memory CD8⁺ T cells of either wild-type or Tnfrsf4^Cre/+ R26^Dta/+ origin were comparable, with no signs of activation (Figure 6e) or competitive disadvantage (Additional data file 2), and CD4:CD8 ratios were restored (Figure 6f). Furthermore, memory CD8⁺ T cells in MHC II-deficient mice contained higher percentages of CD62L^- and CD43⁺ activated/effector CD8⁺ T cells than wild-type mice, despite proportional increases in numbers of both naïve and memory CD8⁺ T cells (Additional data file 4). Thus, memory CD8⁺ T cell activation and expansion in Tnfrsf4^Cre/+ R26^Dta/+ mice resulted from CD4⁺ T cell insufficiency, rather than a cell-autonomous effect.

Accelerated turnover of activated CD4⁺ T cells could trigger generalized immune activation by several distinct mechanisms, including self-reactivity to apoptotic CD4⁺ T cells 28, translocation of microbial products into the intestinal mucosa as a result of local effector CD4⁺ T cell depletion, and loss of regulatory CD4⁺ T cell activity 2930. Reactivity to apoptosis-related self peptides could be excluded as the cause of CD8⁺ T cell activation in Tnfrsf4^Cre/+R26^Dta/+ mice because activation of these cells was not observed in mixed bone marrow chimeras in the presence of wild-type CD4⁺ T cells (Figure 6e), despite continuous apoptosis of Tnfrsf4^Cre/+ R26^Dta/+ CD4⁺ T cells.

Although we found no evidence for intestinal pathology in Tnfrsf4^Cre/+ R26^Dta/+ mice, it could be that a small degree of histologically undetectable bacterial translocation was occurring. To answer this question, we examined the effect of microbial translocation on the immune system of Ikbkg^fl/YVil-Cre mice, in which intestinal epithelial cell-specific deletion of NFκB essential modulator (NEMO, encoded by Ikbkg) leads to epithelial cell apoptosis, translocation of bacteria into the mucosa and myeloid differentiation primary response gene 88 (MyD88)-dependent intestinal inflammation and colitis 31. Consistent with disruption of the intestinal epithelial barrier, Ikbkg^fl/YVil-Cre mice showed significantly elevated serum levels of lipopolysaccharide-binding protein (LBP), a surrogate marker for the systemic presence of bacterial lipopolysaccharide, in comparison with control mice (Figure 7a). Numbers of B cells, CD4⁺ and CD8⁺ T cell subsets and the CD4:CD8 ratio were comparable between Ikbkg^fl/YVil-Cre mice and control Ikbkg^fl/Y mice (Figure 7b-e) and sporadic lymphocyte expansions at inflamed mesenteric lymph nodes 31 were at the expense of splenic counterparts (Figure 7e; and data not shown). Moreover, T cell turnover was not systemically increased in Ikbkg^fl/Y Vil-Cre mice compared with control mice, and an increase in BrdU incorporation in memory CD8⁺ T cells in mesenteric lymph nodes was balanced by a proportional decrease in the spleen (Additional data file 5). In contrast to lymphocytes, numbers of cells expressing the myeloid markers CD11b⁺ and F4/80⁺ were dramatically elevated in all secondary lymphoid organs of Ikbkg^fl/YVil-Cre mice and outnumbered B cells or T cells (Figure 7b-e). Thus, microbial translocation in this mouse model was associated with systemic expansion of myeloid cells, but was not sufficient for T cell activation.

We next examined the contribution of microbial exposure to the immune activation status of Tnfrsf4^Cre/+ R26^Dta/+ mice. In agreement with the absence of obvious intestinal pathology, serum levels of LBP in Tnfrsf4^Cre/+ R26^Dta/+ mice were found to be similar to those in control mice (Figure 7f), arguing against a requirement for microbial translocation for immune activation in Tnfrsf4^Cre/+ R26^Dta/+ mice. To formally exclude any potential contribution of microbial exposure, we rendered Tnfrsf4^Cre/+ R26^Dta/+ mice deficient in MyD88. In contrast to the essential role of MyD88 in microbial recognition driving disease in Ikbkg^fl/Y Vil-Cre mice 31, MyD88 deficiency had no effect on immune activation in Tnfrsf4^Cre/+R26^Dta/+ mice, which still experienced lymph node enlargement with accumulation of B cells, expansion of memory CD8⁺ T cells, particularly of the activated CD62L^-CD43⁺ phenotype, and a drop in the CD4:CD8 ratio (Figure 7g-j). Thus, MyD88-dependent microbial exposure was not necessary for immune activation in Tnfrsf4^Cre/+R26^Dta/+ mice.

A deficit in Treg cells has been linked by some studies to immune activation and disease progression in HIV infection 2930. We therefore examined whether immune activation in Tnfrsf4^Cre/+ R26^Dta/+ mice resulted from Treg cell insufficiency. Wild-type regulatory CD4⁺ T cells adoptively transferred into Tnfrsf4^Cre/+R26^Dta/+ mice, but not into CD4⁺ T cell-replete control mice, expanded efficiently, reaching numbers comparable to the number of regulatory CD4⁺ T cells in wild-type mice, and maintained expression of FoxP3, the master regulator of Treg cell differentiation (Figure 8a, b), revealing a homeostatic deficit in the Treg cell pool of Tnfrsf4^Cre/+ R26^Dta/+ mice. The adoptively transferred regulatory CD4⁺ T cells had a significant effect on the proportion of memory CD8⁺ T cells, particularly those with an activated CD44^hiCD62L^- phenotype, in Tnfrsf4^Cre/+R26^Dta/+ mice; the proportion returned to wild-type levels over a period of 3 weeks (Figure 8c, d). Regulatory T cell reconstitution was also associated with reduction of pro-inflammatory cytokine serum levels in Tnfrsf4^Cre/+ R26^Dta/+ mice to levels comparable to those of control mice (Figure 8e). In addition to Treg cell insufficiency, Tnfrsf4^Cre/+ R26^Dta/+ mice were also characterized by impaired effector CD4⁺ T cell function. To examine whether reduction of immune activation in Tnfrsf4^Cre/+R26^Dta/+ mice upon transfer of wild-type CD4⁺ T cells was specific to Treg cells, we performed similar adoptive transfer experiments using total wild-type CD4⁺ T cells, consisting predominantly of non-regulatory conventional CD4⁺ T cells, which gave rise to effector CD4⁺ T cells. Although wild-type effector CD4⁺ T cells efficiently reconstituted Tnfrsf4^Cre/+R26^Dta/+ mice, but not control mice, they had no effect on the proportion of memory CD8⁺ T cells or activated CD44^hiCD62L^-CD8⁺ T cells (Additional data file 6). Reconstitution of Tnfrsf4^Cre/+ R26^Dta/+ mice with wild-type effector CD4⁺ T cells also failed to rescue the reduced host Treg cell numbers in Tnfrsf4^Cre/+ R26^Dta/+ mice (Additional data file 7). Thus, immune activation in Tnfrsf4^Cre/+ R26^Dta/+ mice was due to a deficit in regulatory, but not effector, CD4⁺ T cell function.

Inbred C57BL/6 (B6) and CD45.1-congenic B6 mice (B6.SJL-Ptprc^aPep3^b/BoyJ) were originally obtained from the Jackson Laboratory (Bar Harbor, USA) and were subsequently maintained at NIMR. B6-backcrossed Rag1-deficient mice 52 (B6.129S7-Rag1^tm1Mom/J, called Rag1^-/-here), T cell receptor α (TCRα)-deficient mice 53 (B6.129-Tcra^tm1Phi, called Tcra^-/-), MHC II-deficient mice 54 (B6.129S2-H2^dlAb1-Ea/J, called MHC II^-/-) and MyD88-deficient mice 55 (B6.129-Myd88^tm1Aki, called Myd88^-/-) have been previously described and have also been maintained at NIMR. Mice with an activatable gene encoding YFP targeted into the ubiquitously expressed Gt(ROSA)26Sor (R26) locus have been described 27 and were backcrossed onto the B6 genetic background for at least ten generations.

Mice with an activatable gene encoding DTA targeted into the R26 locus were generated by gene targeting in embryonic stem cells. Briefly, the pUC-DTA vector containing a truncated gene encoding DTA (amino acids 3-193), in which the initial Met-Asp-Pro sequence is donated by the human metallothionein IIA, and which is followed by an SV40 carboxy-terminal Ser-Leu and small t intron, was kindly provided by Ian Maxwell, University of Colorado, Denver, USA. This fragment was subsequently inserted into the pBigT cassette, which also contained a loxP-flanked (floxed) neomycin-resistance gene (neo) and a triple poly-adenylation signal (tpA). The floxed neo-tpA-DTA fragment was subcloned into the RODA26PA vector, which was then linearized and electroporated into R1 embryonic stem cells. Targeted embryonic stem cell clones were injected into B6 blastocysts and germline transmitting R26^Dta/+ mice were backcrossed onto the B6 genetic background for at least six generations.

Mice with a targeted insertion of Cre recombinase into the Tnfrsf4 locus were generated by gene targeting in embryonic stem cells 26. Tnfrsf4^Cre mice were backcrossed onto the B6 genetic background for at least six generations. To obtain Tnfrsf4^Cre/+R26^Dta/+ progeny, homozygous Tnfrsf4^Cre/Cre mice were mated with heterozygous R26^Dta/+ mice. Tnfrsf4^Cre/+ R26^Dta/+ mice were born at expected Mendelian ratios and were viable with no clinical or histological signs of disease or pathology. A small proportion (<8%) of Tnfrsf4^Cre/+R26^Dta/+ mice showed retarded development, which was evident as early as weaning. Histopathology analysis revealed exocrine pancreatic atrophy, consistent with the animals' small size, with absence of inflammatory infiltrates. All remaining organs, including endocrine pancreas, were normal. These mice were excluded from further analysis. In all experiments Tnfrsf4^Cre/+ R26^+/+ littermates were included as controls for Tnfrsf4^Cre/+R26^Dta/+ mice to control for any potential effects of CD134-hemizygosity.

Mice with intestinal epithelial cell-specific deletion of Ikbkg (encoding IKKγ, also called NEMO) were obtained by crossing mice with a Cre-deletable Ikbkg allele (Ikbkg^fl) with mice expressing Cre under the intestinal epithelial-specific villin promoter (Vil-Cre) and have been previously described 31. Ikbkg^fl/Y Vil-Cre and control Ikbkg^fl/Y mice were bred and maintained at the Institute for Genetics, Cologne, Germany. Spleens and lymph nodes from Ikbkg^fl/YVil-Cre and control Ikbkg^fl/Y male mice were harvested at Cologne and shipped in Iscove's modified Dulbecco's medium (IMDM) to NIMR, where they were analyzed the following day. All animal experiments were conducted according to local government regulations and institutional guidelines.

The A/PR/8/34 (PR8) strain of influenza A virus (IAV; kindly provided by Rose Gonsalves, Division of Virology, NIMR) was an allantoic fluid preparation from PR8-infected embryonated eggs. Non-anesthetized mice were infected with 250 hemagglutinin units of PR8 by instillation onto their nasal cavities. Serum titers of IAV-neutralizing antibodies were measured as previously described 56. The Friend virus (FV) used in this study is a retroviral complex of a replication-competent B-tropic helper murine leukemia virus (F-MuLV-B) and a replication-defective polycythemia-inducing spleen focus-forming virus, referred to as FV. The FV stock (kindly provided by Kim Hasenkrug, Laboratory of Persistent Viral Diseases, Rocky Mountain Laboratories, NIAID, NIH, Hamilton, USA) was free of lactate dehydro-genase-elevating virus and was obtained as previously described 57.

FV was propagated in vivo and prepared as 10% w/v homogenate from the spleen of 12-day infected BALB/c mice. Mice received an inoculum of FV complex containing 1,000-2,000 spleen focus-forming units injected via the tail vein in 0.1 ml of phosphate-buffered saline. Cell-associated virus in infected mice was estimated by flow cytometric detection of infected cells using surface staining for the glycosylated product of the viral gag gene (glyco-Gag), using the matrix-specific monoclonal antibody 34 (mouse IgG2b), followed by an anti-mouse IgG2b-fluorescein isothiocyanate (FITC) secondary reagent (BD Biosciences, San Jose, USA). Serum titers of FV-neutralizing antibodies were measured as previously described 57.

Pneumocystis murina was obtained from ATCC/LGC Promochem (stock PRA-111) and administered to non-anesthetized mice by instillation onto their nasal cavities. P. murina was detected in formalin-fixed lung tissue by Gomori's silver stain of lung sections by IZVG Pathology (Leeds, UK) and in fresh lung tissue by PCR specific for the P. murina mitochondrial large-subunit rRNA gene 58.

Cells were stained with directly conjugated antibodies to surface markers, obtained from eBiosciences (San Diego, USA), CALTAG/Invitrogen (Carlsbad, USA) or BD Biosciences. B cells and macrophages were identified as B220⁺ and CD11b⁺F4/80⁺ cells, respectively. Naïve, memory and regulatory T cells were identified as CD44^loCD25^-, CD44^hiCD25^- and CD25⁺ cells, respectively. Four- and eight-color cytometry were performed on FACSCalibur (BD Biosciences) and CyAn (Dako, Fort Collins, USA) flow cytometers, respectively, and analyzed with FlowJo v8.7 (Tree Star Inc., Ashland, USA) or Summit v4.3 (Dako) analysis software, respectively. For detection of cytokine synthesis, cells were stained for surface markers and stimulated for 4 h with phorbol 12,13-dibutyrate and ionomycin (both at 500 ng/ml), in the presence of monensin (1 μg/ml). Cells were then fixed and permeabilized using buffers from eBiosciences, before intracellular staining with tumor necrosis factor (TNF)-α- and interferon (IFN)-γ-specific antibodies (eBiosciences). FoxP3 was detected by intranuclear staining using a FoxP3-staining kit (eBiosciences) according to the manufacturer's instructions. Apoptotic cells were stained using an Annexin V staining kit (BD Biosciences) according to the manufacturer's instructions. Cellular turnover was assessed by BrdU incorporation during a 6-day administration into the drinking water of mice, and in addition 3 days after cessation of BrdU administration. BrdU incorporation was measured using a BrdU staining kit (BD Biosciences) according to the manufacturer's instructions. Ki67-expressing cells were identified using an anti-human Ki67 or matched isotype control (clones B56 and MOPC-2, respectively, BD Biosciences).

Single cell suspensions were prepared from thymus, spleen or lymph nodes of donor mice by mechanical disruption. Spleen suspensions were treated with ammonium chloride for erythrocyte lysis. Lymph node cellularity was calculated as the sum of the cellular contents of inguinal, axillary, brachial, mesenteric and superficial cervical lymph nodes. Total cell numbers of the various lymphoid and myeloid subsets were the sum of the splenic and lymph node content.

Bone marrow cell suspensions were prepared by flushing the bone cavities of femurs and tibiae from donor mice with IMDM. Target cells were enriched in lymph node and spleen suspensions using immunomagnetic positive selection (EasySep beads, StemCell Technologies, Vancouver, Canada) according to the manufacturer's instructions. Enriched cell suspensions were stained with antibodies to surface markers and then further purified by cell sorting, performed on MoFlo cell sorters (Dako). Typical cell purity following cell sorting was higher than 98%. In some experiments, purified cells were further labeled with CFSE (Molecular Probes/Invitrogen, Carlsbad, USA). Purified cells (1 × 10⁶ per recipient for unlabeled cells or 5 × 10⁶ per recipient for CFSE-labeled cells) were injected into recipient mice through the tail vein in 0.1 ml of air-buffered IMDM.

CD45.2⁺ Tnfrsf4^Cre/+ R26^Dta/+ (DTA) and CD45.1⁺ C57BL/6 (B6) wild-type bone marrow cells were injected separately or mixed together (mixed bone marrow chimeras) into non-irradiated Rag1^-/- recipients expressing the allotypic marker CD45.2. Each recipient received one mouse-equivalent of bone marrow cells. Mice were bled periodically for assessment of reconstitution and lymphoid organs were analyzed 12 weeks after bone marrow transfer. In separate experiments, CD45.1⁺ B6 bone marrow cells were injected into non-irradiated CD45.2⁺ Rag1^-/- recipients and reconstitution of lymphoid, myeloid and erythroid lineages by donor-type cells was assessed. This analysis revealed that in non-irradiated Rag1^-/- recipients only the lymphoid lineage (T and B cells) is reconstituted by donor-type cells, whereas all other lineages were still host-derived.

Single cell suspensions were prepared from the spleen or lymph nodes of donor mice and 0.5 × 10⁶ purified T cells per well were stimulated in 96-well plates with CD3/CD28-coated beads (Mouse CD3/CD28 T Cell Expander, Dynal/Invitrogen, Carlsbad, USA) at 1:1 ratio for the indicated length of time.

Levels of serum cytokines were assessed by multiplex cytokine bead arrays (20-plex Panel, BioSource/Invitrogen, Carlsbad, USA; or 23-plex Panel, Bio-Rad, Hercules, USA) using the Luminex 100 System (Luminex, Austin, USA). Levels of serum LBP were determined by ELISA (HyCult Biotech, Uden, The Netherlands) according to manufacturer's instructions.

Parametric comparisons were made by Student's t-test performed using SigmaPlot v10 software (Systat Software Inc., San Jose USA). Fisher's exact test was used specifically for the non-parametric comparison of P. murina-infected and non-infected mice.