Reversal of the T cell immune system reveals the molecular basis for T cell lineage fate determination in the thymus

T cell specificity and function are linked during development, as MHC-II-specific TCR signals generate CD4 helper T cells and MHC-I-specific TCR signals generate CD8 cytotoxic T cells, but the basis remains uncertain. We now report that switching coreceptor proteins encoded by Cd4 and Cd8 gene loci functionally reverses the T cell immune system, generating CD4 cytotoxic and CD8 helper T cells. Such functional reversal reveals that coreceptor proteins promote the helper-lineage fate when encoded by Cd4, but promote the cytotoxic-lineage fate when encoded in Cd8—regardless of the coreceptor proteins each locus encodes. Thus, T cell lineage fate is determined by cis-regulatory elements in coreceptor gene loci and is not determined by the coreceptor proteins they encode, invalidating coreceptor signal strength as the basis of lineage fate determination. Moreover, we consider that evolution selected the particular coreceptor proteins that Cd4 and Cd8 gene loci encode to avoid generating functionally reversed T cells because they fail to promote protective immunity against environmental pathogens.

D evelopment of mature T cells in the thymus is driven by signals transduced by components of the T cell antigen receptor (TCR). Fully assembled αβ TCR complexes are first expressed on immature CD4 + CD8 + (double positive, DP) thymocytes, and these cells are signaled by their TCR to undergo positive selection and to differentiate into mature CD4 or CD8 single positive (SP) T cells 1,2 . It is during positive selection that TCR-signaled DP thymocytes make lineage fate choices and differentiate into SP T cells with either helper or cytotoxic function 2 . Expression of the helper-lineage transcription factor ThPOK directs thymocytes to differentiate into helper T cells [3][4][5][6] , whereas expression of the cytotoxic-lineage transcription factor Runx3 directs thymocytes to differentiate into cytotoxic T cells [7][8][9][10][11] . It remains uncertain how TCR-signaled thymocytes are induced to express these lineage-specific factors, making lineage fate determination a critical feature of T cell development that still requires investigation.
T cell specificity and function are linked during thymic selection because MHC-II-specific TCR signals generate CD4 helper T cells and MHC-I-specific TCR signals generate CD8 cytotoxic T cells 12 . However, the molecular basis for this linkage continues to be a major issue of contention, with two main perspectives. The 'strength of signal' model attributes T cell lineage fates to differences in TCR/coreceptor signaling strengths, with strong CD4-dependent TCR signals inducing the helper-lineage fate and weak CD8-dependent TCR signals inducing the cytotoxic-lineage fate [13][14][15] . Although this perspective has been contradicted by a variety of experiments [16][17][18][19][20] , it has never been definitively invalidated. In fact, recent single-cell gene analyses of thymocytes undergoing positive selection are thought to support strength-of-signal as the basis for lineage fate determination 21,22 . The 'kinetic signaling model' attributes T cell lineage fates to the opposite effects of TCR signaling on Cd4 and Cd8 gene transcription in DP thymocytes 23,24 . TCR signaling upregulates Cd4 but terminates Cd8 transcription, causing persistent CD4-dependent TCR signaling and disrupted CD8-dependent TCR signaling during positive selection 24 . In the kinetic signaling perspective, persistent/long-duration TCR signaling induces ThPOK and the helper-lineage fate, while disrupted or short-duration TCR signaling results in Runx3 expression and the cytotoxic-lineage fate 23 . Because strong signals tend to have a long duration and weak signals tend to have a short duration, it has not been possible to unequivocally distinguish the effects of signal duration on lineage fate from those of signal intensity.
We undertook the present study to assess whether thymocyte lineage fate is determined by coreceptor gene loci that regulate TCR signal duration or by coreceptor proteins, which determine TCR signal strength. We constructed unique FlipFlop mice, whose Cd4 and Cd8 genes encode the opposite coreceptor proteins of wild-type (WT) mice. We discovered that switching the coreceptor proteins that the Cd4 and Cd8 genes encode generates a reversed T cell immune system, with cytotoxic CD4 T cells generated by MHC-II-specific TCR signals (CD4/MHC-II) and CD8 helper T cells generated by MHC-I-specific TCR signals (CD8/MHC-I). Such functional reversal revealed that weakly signaling CD8 coreceptors encoded by Cd4 gene loci promoted long-duration CD8/MHC-I TCR signaling and the helper-lineage fate, whereas strongly signaling CD4 coreceptors encoded in Cd8 gene loci promoted short-duration CD4/MHC-II assessment of gene expression revealed that FlipFlop CD4 T cells closely resembled B6 CD8 cytotoxic-lineage T cells, while FlipFlop CD8 T cells closely resembled B6 CD4 helper-lineage T cells (Fig. 2b). The few exceptions in FlipFlop T cells are due to strain 129 genes 32 remaining in FlipFlop mice from the 129R1 embryonic stem cell line that was originally used to generate Cd8 CD4 gene loci 25 . Principal component analysis confirmed primary similarities between FlipFlop CD4 and B6 CD8 T cells, and between FlipFlop CD8 and B6 CD4 T cells (Fig. 2c). Together, these data show that, in FlipFlop mice, CD4 T cells are cytotoxic-lineage cells and CD8 T cells are helper-lineage cells. We conclude that, regardless of which coreceptor proteins they encode, Cd4 gene loci regulate expression of helper-lineage genes and Cd8 gene loci regulate expression of cytotoxic-lineage genes. As a result, switching the coreceptor proteins that Cd4 and Cd8 gene loci encode functionally reverses the lineage fate of CD4-and CD8-expressing T cells so that the FlipFlop immune system consists of CD4 cytotoxic-lineage T cells and CD8 helper-lineage T cells.
Cd4 and Cd8 gene loci promote different lineage factors. Because CD4 and CD8 T cells express opposite lineage factors and acquire opposite lineage fates in FlipFlop compared with B6 mice, we examined their MHC recognition specificities. We found that CD8 T cell generation was impaired by MHC-I deficiency in FlipFlop.β2m KO and FlipFlop.MHC KO mice, and that CD4 T cell generation was impaired by MHC-II deficiency in FlipFlop.MHC-II KO and FlipFlop. MHC KO mice, which indicated that FlipFlop CD8 T cells are generated by MHC-I-specific selection and that FlipFlop CD4 T cells are generated by MHC-II-specific selection (Fig. 2d). Thus, the MHC recognition specificity of CD4 and CD8 T cells in FlipFlop mice is the same as that in B6 mice, indicating that MHC-I and MHC-II recognition by CD8 and CD4 coreceptor proteins is unaffected by the coreceptor gene loci in which they are encoded.
We then compared TCR-Vβ usage by CD4 and CD8 T cells in B6 and FlipFlop mice as a potential way to observe TCR repertoire shifts in FlipFlop and B6 T cells because of their opposite lineage fates. We identified TCR-Vβ proteins whose usage significantly differed between CD4/MHC-II and CD8/MHC-I T cells in B6 mice and found that their usage also significantly differed between CD4/ MHC-II and CD8/MHC-I FlipFlop T cells (Extended Data Fig. 2). Thus TCR-Vβ usage was similar in FlipFlop and B6 T cells, despite their opposite lineage fates.
Next, we examined positive selection and lineage fate determination by monoclonal OT-I and OT-II transgenic TCRs in Rag KO FlipFlop and Rag KO WT mice. MHC-I-specific OT-I TCR selected CD8 T cells and MHC-II-specific OT-II TCR selected CD4 T cells in both FlipFlop and WT mice (Fig. 2e), confirming that the MHC specificity of CD8 and CD4 T cell positive selection is identical in FlipFlop and WT mice. In contrast, lineage factor expression is opposite in FlipFlop and WT mice, as MHC-I-restricted OT-I CD8 T cells expressed ThPOK in FlipFlop mice but expressed Runx3 in WT mice (Fig. 2e left). Similarly, MHC-II-restricted OT-II CD4 T cells expressed Runx3 in FlipFlop mice but expressed ThPOK in WT mice (Fig. 2e right). These results document that monoclonal T cells signaled by identical TCRs and coreceptor proteins express opposite lineage factors in FlipFlop and WT mice. Consequently, both CD4 and CD8 T cells express the helper-lineage factor ThPOK when their coreceptor protein is encoded by Cd4 but express the cytotoxic-lineage factor Runx3 when their coreceptor protein is encoded in Cd8. We conclude that, regardless of which coreceptor protein Cd4 and Cd8 genes encode, Cd4 gene loci promote ThPOK expression and the helper-lineage fate, whereas Cd8 gene loci promote Runx3 expression and the cytotoxic-lineage fate.     Fig. 3a,b). Compared with B6 DP thymocytes, FlipFlop DP thymocytes express more CD4 mRNA and protein and express less CD8 mRNA and protein (Fig. 3a,b and Extended Data Fig. 3a), which indicates that coreceptor transcription is greater from Cd8 gene loci than from Cd4 gene loci. Consistent with this finding, quantitative flow cytometric analysis further revealed that FlipFlop thymocytes express substantially more CD4 than CD8 surface coreceptors, while the reverse is true for B6 thymocytes ( Fig. 3c and Extended Data Fig. 3b). Thus, Cd4 and Cd8 gene loci regulate the amount of coreceptor protein that thymocytes express. Because FlipFlop thymocytes express lower amounts of CD8 surface coreceptors than do B6 thymocytes, they contained even less CD8-associated Lck tyrosine kinase 33,34 than did B6 thymocytes (0.5% versus 16%) (Fig. 3d). Consequently, CD8 coreceptor signaling would be weaker than CD4 coreceptor signaling in FlipFlop thymocytes than it is in B6 thymocytes. To examine this expectation, we signaled unstimulated DP thymocytes with anti-TCR/coreceptor monoclonal antibodies and assessed calcium mobilization. Because anti-TCR by itself fails to signal DP thymocytes 34 , calcium flux induced by anti-TCR/CD4 monoclonal antibodies and anti-TCR/CD8 monoclonal antibodies reflects the strength of CD4 and CD8 coreceptor signaling. As expected, TCR/CD8 coengagement generated weaker signals than TCR/CD4 coengagement in FlipFlop DP thymocytes (Extended Data Fig. 3c), revealing that CD8 coreceptor signaling is weaker than CD4 coreceptor signaling in FlipFlop thymocytes.
Because CD5 expression levels are thought to reflect strength and/or duration of TCR/coreceptor signaling 35 , we next examined CD5 expression on FlipFlop and B6 T cells in the thymus and periphery. CD5 expression was higher on CD4 than on CD8 T cells in B6 mice, as expected, but we were surprised to find that CD5 expression in FlipFlop mice was reversed and was higher on CD8 than on CD4 FlipFlop T cells (Fig. 3e). Because CD8 signaling is weaker in FlipFlop than in B6 mice because of their very low CD8 surface expression and their exceptionally low amounts of CD8-associated Lck (Fig. 3a-d and Extended Data Fig. 3c), high CD5 expression on FlipFlop CD8 T cells cannot reflect strong TCR/ CD8 signaling but might reflect long-duration TCR/CD8 signaling.
T lineage fate is determined by TCR signal duration. TCR signaling of positive selection transiently terminates Cd8 gene transcription but not Cd4 gene transcription, which causes surface expression of Cd8-encoded coreceptor proteins to acutely decline but surface expression of Cd4-encoded coreceptor proteins to persist. Consequently, Cd8-dependent signaling is disrupted while Cd4-dependent signaling persists 23,24 . To visualize changes in surface coreceptor protein expression on signaled thymocytes during positive selection, we identified thymocytes at sequential stages of positive selection by expression of CD69 and CCR7 (ref. 36 ). CD69 − CCR7 − cells are stage 1 pre-selection thymocytes; CD69 − CCR7 + cells are stage 2 thymocytes that have just been TCR-signaled to undergo positive selection; and CD69 + CCR7 + and CD69 − CCR7 + cells are TCR-signaled cells at subsequent stages of positive selection ( Fig. 3f and Extended Data Fig. 3d,e). Indeed, transient termination of Cd8 gene transcription caused an acute decline in CD4 expression on stage 3 FlipFlop thymocytes and in CD8 expression on stage 3 WT thymocytes (Fig. 3g). The acute decline in CD4 expression on FlipFlop thymocytes during positive selection would disrupt CD4-dependent MHC-II signaling and cause it to have a short duration, whereas the steady increase in CD8 surface expression on FlipFlop thymocytes would cause CD8-dependent MHC-I signaling to persist and have a long duration ( Fig. 3g left). Note that the opposite changes in coreceptor expression and positive selection signaling occurred in WT thymocytes (Fig. 3g right).
To assess signaling during positive selection, we analyzed CD5 expression on TCR-signaled FlipFlop and WT thymocytes at sequential stages of positive selection. We compared CD5 expression on TCR-signaled thymocytes (stages 2-5) with TCR-unsignaled thymocytes (stage 1) whose values were set equal to 1.0, and found that CD5 on TCR-signaled stage 2 thymocytes increased by six-to eightfold in both FlipFlop and WT mice (Fig. 3h). Importantly, CD5 on FlipFlop thymocytes at subsequent stages 3-5 further increased during CD8/MHC-I signaling but declined during CD4/MHC-II signaling, with the opposite occurring in WT thymocytes (Fig. 3h). These findings reveal that, in FlipFlop mice, CD8/MHC-I-signaling is long duration and CD4/MHC-II-signaling is short duration, which is opposite of that in WT mice. Thus, in FlipFlop mice, long-duration CD8/MHC-I signaling induced the helper-lineage fate and high CD5 expression, despite weak CD8 signaling; short-duration CD4/ MHC-II signaling induced the cytotoxic-lineage fate and low CD5 expression, despite strong CD4 signaling. Consequently, thymocyte lineage fate and CD5 expression levels reflect TCR signaling duration, not TCR signaling strength.
We conclude that coreceptor proteins encoded in the Cd4 gene locus promote persistent and long-duration positive-selection signaling that induces high CD5 expression, ThPOK expression, and the helper-lineage fate, and that coreceptor proteins encoded in the Cd8 gene locus promote disrupted and short-duration positive selection signaling that stimulates low CD5 expression and allows induction of Runx3 and cytotoxic-lineage fate (schematized in Extended Data Fig. 4). Thus, T cell lineage fate is determined by Cd4 and Cd8 gene loci that transcriptionally regulate the kinetics of coreceptor protein expression and determine TCR signaling duration during positive selection.
A reversed T cell immune system. Switching the coreceptor proteins that Cd4 and Cd8 gene loci encode resulted in generation of a reversed T cell immune system, with CD4/MHC-II cytotoxic T cells and CD8/MHC-I helper T cells. To determine the immunocompetence of reversed-function T cells, we examined in vivo immune responses.
We first considered that species survival requires a functional and self-tolerant T cell immune system, and that self-tolerance requires regulatory T (T reg ) cells expressing the Foxp3 transcription factor 37 . Because T reg cells in B6 mice are generated from CD4/MHC-II helper-lineage T cells, we were surprised to find that FlipFlop mice generate Foxp3 + T cells from CD8/MHC-I helper-lineage T cells and that these CD8 T reg cells display the same expression profiles of CD25, CTLA-4, Helios, and Neuropilin-1 (Nrp-1) proteins as those of B6 CD4 T reg cells 37 (Fig. 4a,b and Extended Data Fig. 5a,b). To determine whether CD8 T reg cells were important for self-tolerance in FlipFlop mice, we introduced the X-chromosome-linked scurfy (sfy) Foxp3 gene mutation into FlipFlop mice 38,39 . In fact, self-tolerance was abrogated in FlipFlop sfy male mice, which lacked T reg cells and developed markedly enlarged lymph nodes and lymphocytic infiltrations into liver and lung ( Fig. 4c and Extended Data Fig. 5c). We then further assessed T reg function in vitro and found that FlipFlop CD8 T reg cells were as effective as B6 CD4 T reg cells in suppressing responses of stimulated B6 CD4 responder T cells (Fig. 4d). Thus, CD8 Foxp3 + T reg cells maintain self-tolerance in vivo and inhibit T cell activation in vitro.
Species survival also depends on a functional T cell immune system to provide protection against environmental pathogens. To document that peripheral FlipFlop T cells respond to TCR stimulation, we stimulated FlipFlop T cells in vitro with immobilized anti-TCR/CD28 monoclonal antibodies and observed that FlipFlop CD4 and CD8 T cells both upregulated surface CD69 expression (Extended Data Fig. 5d right panels). However, only stimulated helper-lineage T cells upregulated surface CD40L expression, and these were CD8 + in FlipFlop mice but CD4 + in B6 mice (Extended Data Fig. 5d left panels). As a first test of in vivo T cell function, we assessed rejection of skin allografts 40    B6 mice (P = 0.15) ( Fig. 4e and Extended Data Fig. 5e). Thus, functionally reversed T cells in FlipFlop mice are immunocompetent to reject skin allografts.
Humoral response to soluble antigens. To assess the in vivo function of CD8/MHC-I helper T cells, we examined the humoral response of FlipFlop mice. T follicular helper (T FH ) cells are a specialized population of antigen-specific CD4 helper T cells resident in lymphoid organs that interact with antigen-specific B cells and activate them to form germinal centers (GCs) and to secrete IgG antibodies 41,42 . Because CD8 TFH cells have not previously been described, we immunized FlipFlop mice with NP-KLH and then looked for T FH cells, which are CXCR5 + PD1 + T cells. In fact, CD8 T FH cells did arise in immunized FlipFlop mice, and these were generated in comparable numbers to CD4 T FH cells in B6 mice ( Fig. 5a and Extended Data Fig. 6a). Moreover, FlipFlop CD8 T FH cells expressed Bcl6, ICOS, and CD40L, as did B6 CD4 T FH cells (Fig. 5b). We then assessed the ability of CD8 T FH cells in FlipFlop mice to promote anti-NP humoral immune responses stimulated by NP-KLH. Notably, although FlipFlop mice and B6 mice had similar numbers of B cells (Fig. 5c), FlipFlop B cells produced only IgM anti-NP antibodies after immunization like T cell-deficient TCRα KO immunized mice (Fig. 5d). FlipFlop mice failed to produce any IgG1 anti-NP antibodies, even at 9 weeks after immunization when B6 mice were producing high amounts of high-affinity IgG1 antibodies (Fig. 5d,e). Thus, CD8 T FH cells failed to activate FlipFlop B cells to undergo T-dependent immunoglobulin (Ig) class switching and Ig affinity maturation.
Histologic examination of FlipFlop spleens revealed that they were virtually devoid of GCs ( Fig. 5f green color), even though their T cell follicles (Fig. 5f red color) were located adjacent to surrounding B cell follicles (Fig. 5f white color). The few GC B cells (identified as Fas + GL7 + B220 + cells) that were present in FlipFlop spleens were not NP-binding and so were not specific for the immunizing NP antigen (Fig. 5g and Extended Data Fig. 6b). Thus, FlipFlop B cells could not be stimulated to become NP-specific GC B cells by CD8 T FH cells. However, they could be stimulated to do so by CD4 T FH cells of B6 origin in mixed bone marrow (BM) chimeras ( Fig. 5h and Extended Data Fig. 6c-e). These results indicate that CD8 T FH cells are specifically unable to productively interact with antigen-specific B cells.
Our finding that CD8/MHC-I TFH cells are generated in immunized FlipFlop mice indicates that dendritic cells (DCs) that the T FH cells interact with can cross-present the immunizing antigen onto their surface MHC-I complexes 43 . However, these CD8/MHC-I T FH cells fail to activate antigen-specific B cells to form GCs or to undergo class switching or affinity maturation, which all indicate that CD8 T FH cells cannot mediate cognate interactions with B cells, likely because B cells cannot cross-present exogenous antigens onto MHC-I surface complexes 44 .

Immune response to virus infection.
Finally, to assess the in vivo functionality of CD4/MHC-II cytotoxic T cells, we examined responses of FlipFlop mice to acute infection with lymphocytic choriomeningitis virus Armstrong strain (LCMV-Arm) 45 . B6 mice successfully cleared the virus within 8 days, but FlipFlop mice failed to clear the virus, as revealed by the persistence of viral proteins in the serum and in the spleen (Fig. 6a,b). Nevertheless, both FlipFlop and B6 mice generated virus-specific cytotoxic T cells during virus infection (Fig. 6c). However, virus-specific T cells in FlipFlop mice were CD4 cytotoxic T cells that bound MHC-II tetramers composed of I-A b + gp66 virus peptide, whereas virus-specific T cells in B6 mice were CD8 cytotoxic T cells that bound MHC-I tetramers composed of H-2D b + gp33 virus peptide. FlipFlop virus-specific CD4 T cells expressed KLRG1 as did B6 CD8 T cells, confirming they were terminally differentiated cytotoxic T cells 46 despite failing to provide protection against virus infection (Fig. 6c and Extended Data Fig. 7a).
Because virus-specific cytotoxic T cells generated in FlipFlop mice were CD4/MHC-II-specific, we wondered whether they might eliminate MHC-II + virus-presenting cells and actively interfere with propagation of antiviral immune responses. To assess this possibility, we constructed mixed BM chimeras with FlipFlop and B6 BM cells which would contain B6-origin cytotoxic T cells that are capable of clearing infectious virus (Extended Data Fig. 7b). Remarkably, FlipFlop-origin cells did in fact prevent viral clearance in infected chimeras (Fig. 6d compare groups 1 and 2). Moreover, interference with viral clearance was specifically mediated by FlipFlop-origin cytotoxic T cells because the interference did not occur (that is, the virus was cleared) when FlipFlop-origin T cells were derived from FlipFlop.Prf KO BM cells lacking the cytolytic protein Perforin (Prf) 30,47 (Fig. 6d compare groups 2 and 3).
Viral clearance in B6 mice is mediated by CD8 cytotoxic T cells, as shown by the fact that B6 and MHC-II KO mice clear virus whereas Prf KO mice fail to clear virus (Extended Data Fig. 7c) 45,47 . Consequently, to determine whether FlipFlop CD4 cytotoxic T cells impaired the expansion of B6 virus-specific CD8 T cells in chimeras, we quantified the number of B6-origin virus-specific CD8 T cells (CD45.1) in their spleens (Fig. 6e). We found in chimeras with FlipFlop-origin cells that the number of B6-origin virus-specific CD8 T cells was markedly reduced, but their number was significantly higher (P < 0.05) when the FlipFlop-origin cells lacked Prf (Fig. 6e compare groups 1 and 2, and groups 2 and 3, and Extended Data Fig. 7d,e). These results reveal that the reduction in B6-origin virus-specific CD8 T cells was caused by FlipFlop-origin cytotoxic T cells (Fig. 6e compare groups 2 and 3, and Extended Data Fig. 7d,e).  Fig. 6d,e (n = 6 per strain, 2 independent experiments). *P < 0.05, **P < 0.01, ***P < 0.001 (two-tailed unpaired t-test); mean ± s.e.m. (c,g,h).
We confirmed these findings by measuring IFNγ production as the response of B6-origin (CD45.1) CD8 T cells to in vitro stimulation with MHC-I-specific viral peptide ( Fig. 6f and Extended Data Fig. 7f). We found that IFNγ responses were significantly reduced when B6-origin virus-specific CD8 T cells were from chimeras containing FlipFlop-origin cells, and the responses were fully restored when the FlipFlop-origin cells lacked Prf (Fig. 6f and Extended Data Fig. 7f)   found that FlipFlop cells reduced the number of B6-origin GC B cells generated in virus-infected chimeras, consistent with the loss of MHC-II + virus-presenting cells (Fig. 6g). Thus FlipFlop-origin CD4/MHC-II cytotoxic T cells do function in vivo but they do not protect against virus infection, partly because they cannot eliminate MHC-IIparenchymal cells that are infected with virus and partly because they actively inhibit propagation of protective antiviral immune responses that are dependent on MHC-II + virus-presenting cells.

Discussion
The present study documents that T cell lineage fate is determined by Cd4 and Cd8 coreceptor gene loci that regulate the kinetics and duration of TCR signaling during positive selection, invalidating coreceptor signal-strength as the basis of T lineage determination. Switching coreceptor proteins encoded in Cd4 and Cd8 gene loci reverses the T cell immune system to contain CD4/MHC-II cytotoxic T cells and CD8/MHC-I helper T cells. Thus, both CD4 and CD8 coreceptor proteins promote helper T cell generation when encoded in Cd4 but promote cytotoxic T cell generation when encoded in Cd8-explaining why helper-lineage T cells invariably express Cd4-encoded coreceptor proteins and cytotoxic-lineage T cells invariably express Cd8-encoded coreceptor proteins. We also documented that Cd4 and Cd8 gene loci dictate the duration of coreceptor-dependent TCR signaling during positive selection, with Cd4-encoded coreceptors promoting long-duration TCR signaling to induce helper-lineage fate and Cd8-encoded coreceptors promoting short-duration TCR signaling and inducing cytotoxic-lineage fate. Finally, reversed-function T cells fail to promote protective immunity, which explains why evolution fixed the particular coreceptor proteins that Cd4 and Cd8 gene loci encode in all surviving species. The molecular basis for lineage determination during positive selection has remained an issue of contention, with lineage choice attributed to either TCR and coreceptor signaling strength or signaling duration 13,21,23 . FlipFlop mice with switched coreceptor proteins clearly distinguished TCR signal strength from TCR signal duration because weakly signaling CD8 coreceptors were controlled by Cd4 gene loci, which promote long-duration TCR signaling and strongly signaling CD4 coreceptors were controlled by Cd8 gene loci, which promote short-duration TCR signaling. Thus, CD8/MHC-I signaling was weak but of long duration, and CD4/MHC-II signaling was strong but of short duration. That CD8/MHC-I signaling generated CD8 helper T cells and CD4/MHC-II signaling generated CD4 cytotoxic T cells reveals that T cell lineage fate is not determined by TCR signaling strength but is determined by TCR signaling duration during positive selection.
It has been a classical paradigm of T cell immunology that TCR specificity dictates T cell function in the thymus, so that MHC-II-specific TCRs select helper T cells and MHC-I-specific TCRs select cytotoxic T cells 12 . The present study overturns this classical paradigm and significantly alters the understanding of why T cell specificity and function are linked during positive selection in the thymus. As shown in this study, MHC-II-specific TCR signaling generates helper T cells only when MHC-II-specific CD4 coreceptor proteins are encoded in Cd4 gene loci; MHC-I-specific TCR signaling generates cytotoxic T cells only when MHC-I-specific CD8 coreceptors are encoded in Cd8 gene loci. Thus, Cd4 and Cd8 gene loci determine helper and cytotoxic T lineage fates, respectively, and determine which coreceptor protein each T cell type expresses. Because functional T cells must express TCR and coreceptors with matching MHC specificities to transduce intracellular signals 33,48,49 , the MHC specificity of helper and cytotoxic T cells depends on which coreceptor protein Cd4 and Cd8 gene loci respectively encode.
Because FlipFlop mice contain a reversed T cell immune system, they revealed unique T cell subsets that have not previously been known. For example, FlipFlop mice generated CD8 T reg and CD8 T FH cells. T reg cells in WT mice are almost exclusively CD4/MHC-II T cells, which has been interpreted as indicating a unique role in thymic T reg generation for strongly signaling CD4 coreceptors and MHC-II-dependent ligands 37 . However, FlipFlop mice contained T reg cells that were CD8/MHC-I helper-lineage T cells and which were as effective as WT T reg cells in mediating self-tolerance and suppressing autoimmune responses. Thus, neither the generation nor function of T reg cells uniquely depends on either CD4 coreceptor proteins or MHC-II-dependent self-ligands. Similarly, T FH cells in WT mice are almost exclusively CD4/MHC-II T cells 41

. However, in vivo immunization of FlipFlop mice with soluble antigen revealed that CD8/MHC-I helper T cells become T FH cells by interacting with
DCs capable of cross-presenting peptides of soluble antigens onto MHC-I complexes 43,44 . Interestingly, CD8/MHC-I TFH cells are incapable of mediating cognate interactions with antigen-specific B cells, likely because B cells cannot cross-present peptides of soluble antigens onto MHC-I complexes. As a result, CD8/MHC-I T FH cells fail to promote Ig class switching or the production of protective IgG antibodies. To further assess immune protection against viral pathogens, we examined the ability of FlipFlop mice to clear viral infection. Somewhat surprisingly, FlipFlop mice were unable to clear virus despite generating virus-specific cytotoxic T cells. Failure to clear virus was shown to be partly due to the fact that their virus-specific cytotoxic T cells were CD4/MHC-II-specific and actively interfered with antiviral immune responses by their elimination of MHC-II + virus-presenting cells. Thus, reversed function CD8/MHC-I helper and CD4/MHC-II cytotoxic T cells fail to provide protective immunity against viral infections. Because protective immunity is necessary for species survival, our findings provide an evolutionary explanation for why Cd4 gene loci encode  test for a and d, two-tailed unpaired t-test for e-g); mean ± s.e.m. (e-g).
In conclusion, the present study has established that the functional polarity of the T cell immune system is determined by the MHC specificity of coreceptor proteins that Cd4 and Cd8 gene loci encode, and that in vivo protective immunity strictly requires the Cd4 gene to encode an MHC-II-specific coreceptor protein and the Cd8 gene to encode an MHC-I-specific coreceptor protein. In addition, this study has established that thymocyte lineage fate is determined by cis-regulatory elements in Cd4 and Cd8 gene loci that regulate the kinetics of coreceptor protein expression and the duration of TCR signaling during positive selection so that Cd4-encoded coreceptor proteins promote helper-lineage fate and Cd8-encoded coreceptor proteins promote the cytotoxic-lineage fate-regardless of the coreceptor protein each gene locus encodes.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41590-022-01187-1.

Methods
Mice. C57BL/6 (CD45.1 and CD45.2) (B6) mice were obtained from Charles River Laboratory. BALB/c, CB6, β2m KO , Scurfy and Perforin KO mice were purchased from The Jackson Laboratory and maintained in our own animal colony. MHC-II KO , Runx3d-YFP 50 , ThPOK-GFP 51 , and Foxp3-GFP knock-in (KI) 52 mice were maintained in our own animal colony, as were OT-I.Rag2 KO and OT-II.Rag2 KO mice 53,54 . The mice were housed on a 12-h light-dark cycle at 20-26 °C with 30-70% humidity. All mice were analyzed without randomization or blinding. All mice analyzed were 6-12 weeks of age and both sexes were used unless mentioned otherwise in the manuscript. All animal experiments were approved by the National Cancer Institute Animal Care and Use Committee and were maintained in accordance with US National Institutes of Health guidelines.
FlipFlop mice. FlipFlop mice with Cd4 and Cd8α gene loci encoding the opposite coreceptor proteins were generated by mating mice with altered Cd8 gene loci together with mice with altered Cd4 gene loci. Mice whose Cd8a gene loci had been altered to encode CD4 coreceptor proteins (Cd8 CD4 ) were previously reported and named 4in8 (ref. 25 ), whereas mice whose Cd4 gene loci were altered to encode CD8 coreceptor proteins (Cd4 CD8 ) needed to be constructed and were named 8in4 to indicate 'CD8 encoded in Cd4 gene loci' . For construction of 8in4 mice, we designed a gene-targeting vector and inserted it into the Cd4 gene of B6 embryonic stem cells by homologous recombination (Extended Data Fig. 1a). The altered Cd4 CD8 gene locus produces CD8.1αβ complexes that could be distinguished from B6-origin CD8.2αβ complexes. Because Cd4 and Cd8 gene loci are closely linked on chromosome 6, a genetic crossover event was necessary to generate a chromosome 6 allele containing both Cd4 CD8 Cd8 CD4 altered gene loci, and we intercrossed mice with a crossover allele on one copy of chromosome 6 to generate FlipFlop mice with Cd4 CD8 Cd8 CD4 crossover alleles on both copies of chromosome 6 ( Fig. 1a right). Quantitative analysis of CD4 and CD8 surface expression. Thymocytes (1 × 10 6 ) were incubated with 1 μg of non-conjugated anti-CD4 (GK1.5 rat IgG2b) or anti-CD8α (53-6.7 rat IgG2b) antibodies for 6 hours at 4 o C. After being washed twice, they were incubated with 0.5 μg of FITC-conjugated anti-rat IgG antibody for 40 minutes at 4 o C. After being washed twice, expression of FITC was analyzed by flow cytometry.

In vitro stimulation of LN T cells. T cells were purified from LN with Pan T cells Isolation Kit (Miltenyi Biotec). LN T cells were stimulated with or without
plate-bound anti-TCRβ (1 μg/ml) and anti-CD28 (1 μg/ml) antibodies for 24 hours at 37 °C. Cultured cells were stained and analyzed by flow cytometry for CD40L and CD69 expression.
RNA-sequencing analysis. CD4 and CD8 T cells were electronically sorted from LN of B6 and FlipFlop mice. Total RNA was prepared from sorted cells with the RNeasy Plus Mini Kit (Qiagen). The quality of RNA was assessed by Bioanalyzer (Agilent), and RNA samples with an RNA integrity number >9 were used. The library was made by using the SMARTer Universal Low Input RNA Kit (Clontech) for sequencing. The sequencing was performed as paired-end 125 bp by using HiSeq2500 equipment (Illumina). Reads were aligned to the mouse genome (mm10) with STAR aligner. Differentially expressed genes (DEGs) were genes whose fold change was more than fivefold and P value was less than 0.05. Visualization of DEGs was shown in a heat map generated with Partek (Partek).
In vitro T reg suppression assay. CD25 + Foxp3 + (GFP + ) T cells were electronically sorted from LNs of B6 Foxp3KI (CD4 T) and FlipFlop Foxp3KI (CD8 T) mice by flow cytometry. Sorted cells were cultured with CD4 + CD25 − cells and irradiated (2,000 rad) splenocytes from B6 Foxp3KI mice in the presence of anti-CD3 antibody (145-2C11; BD Pharmingen, 1 μg/ml) at 37 °C for 3 days. After the culture, [ 3 H]thymidine (400 μCi/ml) was added and incubated for 6 hours at 37 °C. Skin allograft rejection. Tail skins prepared from BALB/c mice were grafted onto the flanks of host mice. Bandages were removed at day 7. Grafts were inspected every 1-2 days and were considered to have been rejected when <20% of the graft remained 55 .
Immunization and ELISA. Mice were immunized intraperitoneally (i.p.) with 100 μg NP-KLH (Biosearch Technologies) mixed with 50 % (vol/vol) imject Alum (Thermo Scientific). Serum and tissues were collected at the appropriate time after immunization. For analysis of NP-binding cells, cells were incubated with NP-PE (Biosearch Technology) and analyzed by flow cytometry. NP-specific antibodies were analyzed using ELISA. Forty-eight-well plates were coated with NP-BSA (Biosearch Technologies) at 4 °C overnight, followed by incubation with serially diluted serum at room temperature for 1 hour. After washing, HRP-conjugated goat anti-mouse IgM (dilution 1:1,000) or IgG1 antibodies (Southern Biotech, dilution 1:2,000) were added to plates and incubated at room temperature for 1 hour. The reaction was developed by incubation with ABTS Peroxidase Substrate (KPL) and was stopped by ABTS Peroxidase Stop Solution (KPL). Plates were analyzed at 405 nm with Fluostar Optima plate reader and software (BMG Labtech).
LCMV infection and viral titer assay. Mice were infected intravenously (i.v.) with LCMV-Armstrong (2 × 10 6 PFU/mouse), and their serum and tissues were collected at day 8 for analysis. Viral titer in the serum from infected mice was assessed using a modified focus-forming assay 56 . Diluted serum (1:100) was incubated with Vero cells (2.5 × 10 4 cells/well) on a 24-well plate at 37 °C for 4 hours. Each well was subsequently overlaid with 0.5% methylcellulose and incubated at 37 °C for 48 hours. Cells were subsequently fixed with 2% formalin/formaldehyde for 30 minutes and then with 0.5% Triton-X for 20 minutes. Fixed cells were stained with anti-LCMV NP antibody (VL-4; Bio X Cell, 5 μg/ml) for 1 hour, followed by staining with anti-rat IgG HRP antibody (Jackson Immunoresearch, dilution 1:1,350) for 1 hour. LCMV foci were visualized using an ImmPACT DAB Peroxidase (HRP) Substrate kit (Vector Labs).