The Lupus Susceptibility Gene Pbx1 Regulates the Balance between Follicular Helper T Cell and Regulatory T Cell Differentiation

The NZM2410 murine strain was derived from the classical (NZB × NZW) F1 model of systemic lupus erythematosus (SLE), which is an autoimmune disease characterized by the production of autoantibodies to nuclear Ags, immune activation, and immune complex–mediated inflammation in various tissues. We mapped three major NZM2410 lupus-susceptibility loci and generated three congenic strains, B6.NZMSle1, B6.NZMSle2, and B6.NZMSle3, each carrying the corresponding NZM2410-derived genomic interval on the C57BL/6J (B6) genome (1, 2). Analysis of immunological properties of each congenic strain in comparison with B6 showed that Sle1 and Sle3 increase T cell activation and effector functions and that Sle1 and Sle2 promote the development of autoreactive B cells and B cell hyperactivity (3).

Sle1 on telomeric chromosome 1 was the locus with the strongest linkage to lupus nephritis, and its expression is required for the development of systemic autoimmunity and pathogenesis in the NZM2410 model (4, 5). Sle1 expression results in the production of autoantibodies specific for chromatin (6) through intrinsic defects in B and CD4⁺ T cells (7). Three Sle1-independent subloci, Sle1a, Sle1b, and Sle1c, contribute to autoimmune phenotypes (8). Sle1a and Sle1c induce the production of activated autoreactive T cells and decrease the number and function of Foxp3⁺ regulatory T cells (Tregs) (9–11). Sle1b is associated with extensive polymorphisms between two divergent haplotypes of the SLAM family, and it regulates B cell (12) and T cell (13) tolerance. Sle1a was mapped to two interacting loci: Sle1a1 and Sle1a2 (14). Sle1a1 contains only one functional known gene, Pbx1, and the lupus-associated allele corresponds to the increased expression of a novel splice isoform, Pbx1-d, in CD4⁺ T cells (15). Pbx1, a member of the three-amino-acid-loop-extension class of homeodomain proteins, is a transcriptional factor that regulates chromatin access of multimeric complexes that include Hox factors, as well as Meis and Prep-1, two other three-amino-acid-loop-extension proteins that regulate chromatin remodeling and coactivator access (16). The interactions of Pbx-Meis/Prep-1 complexes with Hox and non-Hox factors ultimately contribute to gene activation or repression (17). Pbx1-d lacks exon 6 and exon 7 corresponding to the DNA-binding domain and the Hox-binding domain, respectively, which confers this splice isoform a dominant-negative function (18).

Sle1a1 expands the number of activated and autoreactive CD4⁺ T cells and reduces the number of peripheral Tregs (pTregs) in a CD4⁺ T cell–intrinsic manner (15). However, these T cell phenotypes were not sufficient to induce a robust production of autoantibodies in B6.Sle1a.1^NZW/NZW (B6.Sle1a1) mice, which requires the expression of Sle1 in B cells (14). In addition, Pbx1-d expression is associated with abnormal responses to TGF-β and retinoic acid (RA) in murine and human T cells (19). Pbx1 expression is necessary for B cell development (20), but its function in T cells is unknown. The goal of this study was to directly address the role of Pbx1-d overexpression in CD4⁺ T cells. We showed that transgenic (Tg) B6 mice that overexpress Pbx1-d in their CD4⁺ T cells (Pbx1-d–Tg mice) reproduce the phenotypes of B6.Sle1a1 mice, with increased inflammatory functions of CD4⁺ T cells and impaired Treg homeostasis. In addition, Pbx1-d–Tg mice possessed a follicular helper T cell (T_FH) population that expanded in an Ag-specific and T cell–intrinsic manner, with an enhanced capacity to locate in B cell follicles and to promote affinity maturation of Th1-associated Ab isotypes. These results suggest that Pbx1 regulates the balance between pTreg and T_FH maintenance or differentiation and that Pbx1-d contributes to autoimmunity by tilting the balance in favor of T_FHs over Tregs.

Pbx1-d–Tg mice were generated at the University of Florida transgenic core using a bicistronic Pbx1-d/GFP cDNA controlled by the CD4 promoter (Supplemental Fig. 1A) on a B6 background. Four Tg lines were obtained, with a Tg copy number of 8–13 in the A886, C855, and A872 lines, and a Tg copy number of 30 in the C861 line (Supplemental Fig. 1B). GFP expression was not achieved in any of the lines. The following primers were used to detect Pbx1-d Tg: forward (spanning exons 5–8): 5′-ATCACAGTCTCCCAGGTGGA-3′, and reverse (in exon 9): 5′-ATCCTGCCAACCTCCATTAG-3′. Pbx1-d expression was restricted to CD4⁺ T cells (Supplemental Fig. 1B, 1C). Primer and TaqMan probe sequences used to measure Pbx1-d message expression were described (21). Mice from all four lines were healthy at least up to 1 y of age. The results reported in this article were obtained using mice from the first three lines, without any difference observed among them. C861 mice were not included because of poor breeding performance. Initial characterization was performed with Tg-negative littermates, which presented phenotypes identical to those of B6 controls. No difference was observed between hemizygous and homozygous lines, indicating that, within the observed range, the Tg copy number was not critical for the phenotypes. The results reported in this article were obtained with homozygous mice. B6, B6.SJL-Ptprc^aPep3^b/BoyJ (B6.Ly5^a), B6.Cg-Tg(TcraTcrb)425Cbn/J (B6.OT-II), and B6.129S7-Rag1^tm1Mom/J (B6.Rag-1^−/−) mice were originally obtained from The Jackson Laboratory. B6.Thy1^a.OT-II mice were graciously provided by Dr. Stephen Schoenberger (La Jolla Institute for Allergy and Immunology). The B6.Sle1a1 and B6.NZM-Sle1^NZM2410/AegSle2^NZM2410/AegSle3^NZM2410/Aeg/LmoJ (TC) congenic mice were described previously (5, 14). B6.Foxp3-enhanced GFP (B6.Foxp^egfp) mice (22) were kindly provided by Dr. Vijay Kuchroo (Harvard Medical School). Pbx1-d–Tg.Foxp3^egfp mice were bred from the A886 line, and Pbx1-d–Tg.OT-II mice were bred from the C855 line. All mice were bred and maintained at the University of Florida in specific pathogen–free conditions. With the exception of the colitis experiment, only female mice were used in this study under a protocol approved by the Institutional Animal Care and Use Committee of the University of Florida.

In vitro–induced Tregs (iTregs) were differentiated as previously described (15). Briefly, CD4⁺CD25⁻ cells were negatively selected from B6 or Pbx1-d–Tg splenocytes using the CD4⁺CD25⁺ Regulatory T Cell Isolation Kit (Miltenyi Biotec); 5 × 10⁵ cells were stimulated with mouse T-activator CD3/CD28 beads (Life Technologies) at a concentration of 1 × 10⁶ beads/ml in the presence of 100 U IL-2, 20 ng/ml TGF-β, (PeproTech), and 0.5 nM RA for 5 d. Th1 and Th17 polarization was performed as previously described (23).

Single-cell suspensions were prepared using standard procedures from spleen, thymus, and mesenteric lymph node (mLN). After RBC lysis, cells were blocked with anti-CD16/32 Ab (2.4G2) and stained in FACS staining buffer (2.5% FBS, 0.05% sodium azide in PBS). Fluorochrome-conjugated Abs, purchased from BD Biosciences, eBioscience, BioLegend, and R&D Systems, were to B220 (RA3-6B2), BCL6 (K112-91), CD4 (RAM4-5), CD8 (53-6.7), CD25 (PC61.5), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD62L (MEL-14), CD69 (H1.2F3), CD95 (Jo2), CD90.1 (HIS51), CD90.2 (53-2.1), Foxp3 (FJK-16s), Ly-77 (GL7), Neuropilin-1 (761705), PD-1 (RMP1-30), IFN-γ (XMG1.2), IL-2 (JES6-5H4), IL-17a (eBio17B7), and IL-21 (FFA21). Follicular T cells were stained as previously described (23) in a three-step process using purified CXCR5 (2G8), followed by biotinylated anti-rat IgG (Jackson ImmunoResearch) and then PerCP-Cy5.5–labeled streptavidin in FACS staining buffer on ice. Dead cells were excluded with fixable viability dye (eFluor 780; eBioscience). Data were collected on an LSRFortessa (BD Biosciences) and analyzed with FlowJo software (TreeStar).

Chimeras were prepared as previously described (9), with the following modifications. Eight- to ten-week-old (B6 × B6.Ly5^a)F1 recipient mice were lethally irradiated with two doses of 452 rad (4 h apart) using an X-RAD 320 irradiator (Precision X-Ray). Donor bone marrow (BM) cells were depleted of mature T cells using CD5 MicroBeads (Miltenyi Biotec). BM cells (1 × 10⁷ cells) mixed 1:1 between the two donors were given to the recipient mice by tail vein injection. Tregs were analyzed in chimeric mice 8 wk later. For T-dependent (TD) immune responses, chimeric mice were immunized with 100 μg NP₂₃-KLH (Biosearch Technologies) in alum 8 wk after BM cell transplantation, and T_FHs were analyzed 1 wk after immunization. CD4⁺ T cells from Pbx1-d–Tg.OT-II, B6.Sle1a1.OT-II, and B6.OT-II mice were purified by negative selection using MACS MicroBeads, and 0.5 × 10⁶ cells were injected into the tail vein of B6.Ly5^a mice, followed by s.c. immunization with 50 μg NP₁₆-OVA in alum 4 h after cell transfer. For mixed adoptive transfer, CD4⁺ T cells from Pbx1-d–Tg.OT-II or B6.Sle1a1.OT-II and B6.Thy1^a.OT-II mice were mixed at a 1:1 ratio, and 1 × 10⁶ cells were injected into (B6 × B6.Thy1^a)F1 recipient mice prior to immunization with NP₁₆-OVA.

To assess the function of effector T cells (T_Effs) in vivo, 4 × 10⁵ purified CD4⁺CD25⁻ T_Effs from 2-mo-old Pbx1-d–Tg and B6 mice were injected into the tail vein of B6.Rag-1^−/− mice. The function of Tregs was evaluated in coinjections of 1 × 10⁵ CD4⁺CD25⁺ Tregs from Pbx1-d–Tg or B6 mice along with 4 × 10⁵ CD4⁺CD25⁻ B6 T_Effs. Recipient mice were monitored for clinical signs of colitis for up to 8 wk, and body weight was monitored weekly. The recipient mice that lost >15% of body weight or showed overt clinical signs of disease were sacrificed. Colon histology was ranked blindly in a semiquantitative fashion, as previously described (9).

To stain splenic germinal centers (GCs) in B6.Ly5^a mice that received Pbx1-d–Tg.OT-II or B6.OT-II CD4⁺ T cells, 7-μm-thick OCT-embedded cryosections were prepared on Superfrost slides, washed with PBS, fixed with 4% paraformaldehyde for 15 min at 4°C, and subsequently permeabilized with 0.1% Triton X-100 for 5 min at 4°C. The sections were washed with cold PBS and incubated overnight in the dark at 4°C with anti-CD45.2 conjugated to PE, anti-GL7 conjugated to Alexa Fluor 488, and anti-IgD conjugated to allophycocyanin (all from eBioscience). Sections were washed three times with cold PBS, mounted with Cytoseal, and covered with a glass coverslip. The stained sections were analyzed using an EVOS FL digital inverted fluorescence microscope (Fisher Scientific, Waltham, MA); images were captured using 10×, 20×, and 40× objectives, keeping all microscope conditions and software settings identical for all treatments and controls.

For TD responses, 8- to 10-wk-old mice were immunized i.p. with 100 μg NP₂₃-KLH in alum and boosted at weeks 2 and 6. Serum samples were collected 1, 3, 5, and 7 wk after the first immunization. NP-specific Abs were measured by ELISA using plates coated with NP₄-BSA or NP₂₅-BSA (Biosearch Technologies), followed by incubation with 1:1000 diluted serum samples and developed with alkaline phosphatase–conjugated goat anti-mouse IgG1, IgG2a, IgG2c, IgG3, or IgM (Southern Biotech). All samples were run in duplicate. Anti-dsDNA and anti-chromatin IgG were measured, as previously described (6), in sera diluted 1:100, and relative units were standardized using serial dilutions of a positive serum from TC mice, setting the 1:100 dilution reactivity to 100 U. Serum anti-nuclear Abs (ANAs) were measured by applying 1:40 diluted sera to fixed Hep-2 cells (Inova) and revealed with anti-mouse IgG-FITC (Invitrogen).

Splenic CD4⁺ T cells from Pbx1-d–Tg.Foxp3^egfp, B6.Sle1a1.Foxp3^egfp, and B6.Foxp3^egfp mice were enriched by negative selection using MACS MicroBeads and then GFP⁺ and GFP⁻ cells were sorted with a FACSAria cytometer with a purity ≥ 95%. MicroRNAs (miRNAs) were isolated and reverse transcribed using Life Technologies reagents. Quantification of miRNA expression was performed using TaqMan MicroRNA assays (Life Technologies) and the Applied Biosystems StepOne real-time PCR machine. U6 small nuclear RNA was used as internal control. Comparisons were made using the 2^−ΔΔCt method, and values were normalized to the B6 average values.

Putative human PBX1 binding sites in the ∼2 kb flanking miR-10a, miR-21, and miR-155 precursors were identified using Jaspar3 (set for 75% accuracy). We generated luciferase reporter constructs containing the 798-bp, 1071-bp, and 850-bp regions upstream of miR-10a, miR-21, and miR-155, respectively (Fig. 8A), cloned into the pGL4.25 Luciferase vector (Promega). PBX1-b or PBX1-d expression plasmids were generated by inserting cDNAs (15) into the pHAGE-CMV-MCS-IzsGreen plasmid. The luciferase reporter constructs and the pGL3-Basic plasmid were each cotransfected with PBX1-b or PBX1-d expression plasmid into HEK293T cells. Cells were harvested after 48 h. After lysis with Passive Lysis Buffer (Promega), luciferase activity was measured using a dual-luciferase reporter assay system (Promega), according to the manufacturer’s instructions, with a Lumat LB 9507 luminometer (Berthold Technologies).

Differences between groups were evaluated by two-tailed statistics: unpaired t tests or Mann–Whitney U tests depending on whether the data were normally distributed, paired t tests for mixed BM chimeras, χ² tests to compare distributions, and two-way ANOVA tests for time-course experiments. Unless specified, graphs show mean and standard deviations of the mean.

To address whether the overexpression of Pbx1-d in CD4⁺ T cells was sufficient to induce their activation and loss of tolerance, we generated Pbx1-d–Tg mice that express Pbx1-d driven by the Cd4 promoter (Supplemental Fig. 1). Pbx1-d Tg overexpression in CD4⁺ T cells resulted in the production of serum ANAs with the characteristic homogenous nuclear staining pattern (Fig. 1A), as well as anti-dsDNA and anti-chromatin IgG in 7–12-mo-old mice (Fig. 1B, 1C). A similar nuclear staining pattern was observed in the serum of B6.Sle1a1 mice, although it had a lower intensity, which is consistent with a low level of anti-dsDNA and anti-chromatin IgG (Fig. 1B, 1C), as we reported previously for this strain (14, 15). This suggested that Pbx1-d overexpression results in the production of autoreactive CD4⁺ T cells that are sufficient to induce humoral autoimmunity against nuclear Ags. B6.Sle1a1 mice were characterized by the expansion of CD44⁺CD62L⁻ effector memory T cells (T_EMs) relative to CD44⁻CD62L⁺ naive T cells (T_Ns) (15). Aged Pbx1-d–Tg mice also presented a skewed T_N/T_EM ratio (Fig. 1D). We examined the consequence of Pbx1-d overexpression in CD4⁺CD25⁻ T_Effs in vivo with the experimental colitis model that we used with B6.Sle1a1 T cells (9). B6.Rag1^−/− mice that received Pbx1-d–Tg T_Effs showed greater body weight loss and more severe colitis than the mice that received B6 T_Effs (Fig. 1E, 1F). These results suggest that Pbx1-d expression is sufficient to induce an intrinsic activation in CD4⁺ T cells. However, we observed a similar ability of Pbx1-d–Tg or B6 CD4⁺CD25⁺ T cells to suppress B6 T_Effs in the same colitis model (data not shown), suggesting that the in vivo functions of Pbx1-d–Tg Tregs were less affected than those of B6.Sle1a1 Tregs (14).

To assess the role of Pbx1-d in Tregs compared with conventional CD4⁺ T cells, we bred the EGFP-Foxp3 reporter construct to B6.Sle1a1 and Pbx1-d–Tg mice. Pbx1-d expression in Pbx1-d–Tg CD4⁺ T cells was ∼10,000-fold higher than in B6 T cells and was similar in Foxp3⁺ and Foxp3⁻ cells (Fig. 2A, 2B). Pbx1-d expression in Pbx1-d–Tg T cells was also much higher (∼100-fold) than in B6.Sle1a1 mice in Foxp3⁺ and Foxp3⁻ T cells (Fig. 2A, 2B). As previously reported for B6.Sle1a1 mice (14, 15), the frequency of CD4⁺Foxp3⁺ Tregs was reduced in Pbx1-d–Tg mice, with the largest difference in the mLN (Fig. 2C). In addition, in vitro iTreg polarization by TGF-β and RA was reduced in Pbx1-d–Tg CD4⁺CD25⁻ T cells compared with B6 cells (Fig. 2D). These results indicate that, as found in B6.Sle1a1 mice, Pbx1-d overexpression in CD4⁺ T cells impairs iTreg differentiation by interfering with the response to RA in the presence of TGF-β.

To further evaluate the impact of Pbx1-d on Treg differentiation, we reconstituted lethally irradiated (B6 × B6.Ly5^a)F1 mice with T cell–depleted BM cells mixed from Pbx1-d–Tg and B6.Ly5^a mice (Fig. 2E). Pbx1-d–Tg BM yielded a decreased percentage of CD4⁺Foxp3⁺Nrp-1⁻ pTregs compared with B6.Ly5^a BM in the thymus of the recipient mice, whereas the proportion of CD4⁺Foxp3⁺Nrp-1⁺ thymic Tregs (tTregs) was similar between the two genotypes (Fig. 2F–H). However, in these conditions, the proportions of Tregs in the peripheral organs of recipient mice were similar for BM-derived cells of either genotype (data not shown), possibly as a result of the relatively short time between BM transfer and phenotype readout. Together, these data indicate that Pbx1-d Tg overexpression in CD4⁺ T cells impairs the induction or maintenance of pTregs in a cell-intrinsic manner.

Because Pbx1-d–overexpressing CD4⁺ T cells showed an activated phenotype (Fig. 1D), we next examined their cytokine production. Ex vivo Pbx1-d–Tg CD4⁺ T cells showed a significantly increased IL-21 expression (Supplemental Fig. 2A, 2B) compared with B6 mice, but there was no difference in the production of IL-2, IL-17A, or IFN-γ (Supplemental Fig. 2A, 2C–E) or the frequency of IL-17A⁺ or IFN-γ⁺ CD4⁺ T cells (data not shown). Under in vitro Th1-polarization conditions, a small, but significant, difference was observed, with Pbx1-d–Tg CD4⁺ T cells producing more IFN-γ than B6 cells, whereas there was no difference for IL-17A (Supplemental Fig. 2F, 2G). These results prompted us to investigate T_FH differentiation in Pbx1-d–Tg mice. T_FHs secrete high levels of IL-21, and IL-21 signaling upregulates Bcl6 expression, which is required for T_FH development and GC formation (24–27). At steady-state, young (6–8 wk) Pbx1-d–Tg and B6.Sle1a1 mice showed an increased cellularity in the mLN, where the numbers and percentages of GC B cells were also increased compared with B6 mice (Fig. 3A–C). The percentages and absolute numbers of CD4⁺PD1⁺CXCR5⁺Bcl6⁺Foxp3⁻ T_FHs were significantly higher in the mLN, but not in the spleen, from Pbx1-d–Tg mice (Fig. 3D–F). A similar trend was observed for B6.Sle1a1 T_FHs. In contrast, the number of CD4⁺PD1⁺CXCR5⁺Bcl6⁺Foxp3⁺ follicular regulatory T cells (T_FRs) was significantly reduced in the spleen of Pbx1-d–Tg mice (Fig. 3F). We next tested the differentiation of T_FHs and T_FRs 7 d after administering a TD Ag to chimeric mice reconstituted with a 1:1 mix of Pbx1-d–Tg and B6 BM (Fig. 3G). The number and percentage of Pbx1-d Tg-derived T_FHs and T_FRs were increased significantly compared with B6-derived cells (Fig. 3H, 3I). The discrepancy between the unchanged frequency or fewer Pbx1-d–Tg T_FRs in unmanipulated mice (Fig. 3E, 3F) on one hand and the increased number and frequency of Pbx1-d Tg-derived T_FRs in chimeric mice (Fig. 3H, 3I) on the other hand is likely due to homeostatic expansion that favors all Bcl6-expressing Pbx1-d Tg-derived T cells. The numbers of B6 and Pbx1-d Tg-derived total CD4⁺ T cells were similar before immunization, indicating an Ag-driven enhancement of T_FH differentiation by Pbx1-d overexpression. These steady-state and immunization results are consistent with Pbx1-d overexpression in CD4⁺ T cells increasing T_FH differentiation in a cell-intrinsic manner.

To track T_FHs in an Ag-specific manner, Pbx1-d–Tg mice were bred with B6.OT-II mice, which express an MHC class II–restricted TCR specific for OVA (28). Pbx1-d–Tg.OT-II or B6.OT-II CD4⁺ T cells were transferred into B6.Ly5^a mice that were subsequently immunized with NP-OVA. By day 5 after immunization, Pbx1-d–Tg.OT-II and B6.OT-II total CD4⁺ T cells were expanded to a similar extent, and there was no difference for the recipient CD45.1⁺ T cells (Fig. 4A). However, CD45.2⁺ Pbx1-d–Tg.OT-II CD4⁺ T cells contained a higher frequency and number of T_FHs (Fig. 4B, 4C) than did the B6.OT-II CD4⁺ T cells. Pbx1-d–Tg.OT-II or B6.OT-II transferred CD4⁺ T cells produced very low numbers of T_FRs (Fig. 4A). Homing of primed T_FHs to B cell follicles is a critical step to generate GCs. The number of transferred Pbx1-d–Tg.OT-II CD4⁺ T cells within the B cell zone was significantly higher than the number of B6.OT-II CD4⁺ T cells, whereas there was no difference in the number of transferred T cells in the T cell zone (Fig. 4D, 4E). In addition, CXCR5 and Bcl6 expression was significantly higher in the transferred Pbx1-d–Tg.OT-II T cells than in the transferred B6.OT-II CD4⁺ T cells (CXCR5: 483 ± 18.14 versus 430.20 ± 10.14, p < 0.001; Bcl6: 507.20 ± 23.21 versus 436.60 ± 13.86, p < 0.05). At day 7 after immunization, there was a similar frequency of Pbx1-d–Tg.OT-II and B6.OT-II T_FHs (Supplemental Fig. 3A) and a similar number of these cells in the spleen (Supplemental Fig. 3B), although the number of Pbx1-d–Tg.OT-II T_FHs was still higher in the mLN (Supplemental Fig. 3B). Interestingly, the number of total Pbx1-d–Tg.OT-II T cells was increased in both spleen and mLN (Supplemental Fig. 3C). This expanded population was largely absent from the GC (Supplemental Fig. 3D), suggesting that they may correspond to extrafollicular Th cells, a population that is expanded in lupus-prone mice (29).

We performed the same experiment with B6.Sle1a1.OT-II CD4⁺ T cells, which generated a similar frequency and number of T_FHs, as well as a similar number of CD4⁺ T cells in the B cell follicles than B6.OT-II CD4⁺ T cells 5 d after immunization (Fig. 4F, 4G). At day 7 after immunization, the number and frequency of T_FHs were also similar between cells of Sle1a1 or B6 origin (Supplemental Fig. 3E, 3F). However, the total number of Sle1a1.OT-II CD4⁺ T cells was higher than the total number of B6.OT-II T cells in the spleen, and there was a similar trend in the mLN (Supplemental Fig. 3G). Therefore, these results suggest that a high level of Pbx1-d expression in CD4⁺ T cells increased Ag-specific T_FH differentiation, as well as early entry in B cell follicles, whereas the more modest increased expression of Pbx1-d in Sle1a1 CD4⁺ T cells increased their Ag-specific expansion.

We then examined the effect of Pbx1-d overexpression on T_FH differentiation in a competitive setting by cotransferring B6.Thy1^a.OT-II and Pbx1-d–Tg.OT-II (1:1) CD4⁺ T cells into (B6 × B6.Thy1^a)F1 recipient mice. Five days after primary NP-OVA immunization (Fig. 5A), the percentage of total B6.Thy1^a.OT-II CD4⁺ T cells was higher than that of Pbx1-d.OT-II–Tg CD4⁺ T cells in the spleen and mLN, and the same trend was observed for Sle1a1.OT-II CD4⁺ T cells (Fig. 5B). A higher percentage of T_FHs was only found with Pbx1-d Tg origin in the mLN (Fig. 5C). We next asked whether Pbx1-d overexpression influenced the recall Ag response in the same model (Fig. 5A). In contrast to the primary response, the percentages of Pbx1-d.OT-II–Tg and Sle1a1.OT-II total CD4⁺ T cells (Fig. 5D) and T_FHs (Fig. 5E) were higher than those of corresponding cells of B6 origin in the spleen and mLN at 3 d after secondary immunization. Moreover, the relative expansion of the T_FH subset was significantly greater in cells from Pbx1-d Tg or Sle1a1 origin compared with cells of B6 origin in the recall compared with the primary response (Fig. 5F). Finally, the percentage of Pbx1-d–Tg–activated T cells (Fig. 5G, 5H) and CD44^hi memory CD4⁺ T cells (Fig. 5I, 5J) was enhanced, which we did not observe during the primary response. Taken together, these results indicate that Pbx1-d overexpression conferred a cell-intrinsic competitive advantage during Ag-specific T_FH differentiation that was enhanced during recall responses.

To assess the effect of Pbx1-d overexpression on affinity maturation and class switching, B6 and Pbx1-d–Tg mice were boosted 2 and 6 wk after primary immunization with NP-KLH. Pbx1-d–Tg mice produced significantly more high- and low-affinity anti–NP-IgG3 and high-affinity anti-NP IgG2a and IgG2c than did B6 mice (Fig. 6A, 6B). Moreover, the affinity of anti-NP IgG2a and IgG2c, as measured by the NP₄/NP₂₅ ratio, was significantly increased in Pbx1-d–Tg mice after boosting (Fig. 6C), suggesting that Pbx1-d overexpression increases affinity maturation of Th1-associated Abs.

Numerous studies implicated miRNAs in Treg and T_FH differentiation and function (30). We analyzed the effect of Pbx1-d overexpression on candidate miRNAs selected for their role in this process (miR-10a, miR-17, miR-19a, miR-19b, miR-20a, miR-21, miR-92, miR-155, miR-181a, and miR-181b) in Foxp3-GFP⁺ and Foxp3-GFP⁻ CD4⁺ T cells from young Pbx1-d–Tg.Foxp3^egfp, B6.Sle1a1.Foxp3^egfp, and B6.Foxp3^egfp mice. miR-10a, miR-21, and miR-155 expression was 2–3-fold higher in Pbx1-d–Tg or B6.Sle1a1 Tregs than in B6.Foxp3-GFP⁺ Tregs (Fig. 7A). The same result was obtained for Pbx1-d–Tg Foxp3-GFP⁻ conventional CD4⁺ T cells, whereas expression of miR-21 and miR-155 was more variable in B6.Sle1a1 Foxp3-GFP⁻CD4⁺ T cells, although with differences going in the same direction (Fig. 7B). The expression of miRNAs from the miR-17–92 and miR-181 families was significantly increased in Foxp3-GFP⁺ Tregs from B6.Sle1a1 mice but not in Tregs from Pbx1-d–Tg mice, and there was no consistent difference in Foxp3-GFP⁻ conventional T cells between strains (Supplemental Table I). These findings indicate that Pbx1-d affects the expression of specific miRNAs that are associated with CD4⁺ Treg and T_FH differentiation and function, either directly or indirectly.

To address whether the increased expression of miR-10a, miR-21, and miR-155 in Pbx1-d–Tg and B6.Sle1a1 CD4⁺ T cells was maintained in mice with clinical lupus phenotypes, we analyzed TC mice, which carry the Sle1a1 allele of Pbx1 (5). TC mice present spontaneously expanded subsets of T_FHs (Fig. 7C, 7D). miR-10a, miR-21, and miR-155 were expressed at higher levels in TC CD4⁺ T cells starting at 2 mo of age, before the mice produce autoantibodies and show overt signs of autoimmune activation (Fig. 7E). These results suggest that the upregulation of these three miRNAs driven by Pbx1-d overexpression in Tg mice contributes to the autoimmune phenotype in lupus mice that express the Pbx1-d allele.

miR-21 expression is regulated by the PBX1/HOX9 complex in leukemia (31), suggesting that Pbx1 could directly regulate the transcription of miR-10a, miR-21, and miR-155. An in silico analysis identified several putative PBX1 binding sites for each of the three miRNAs that are overexpressed in Pbx1-d–Tg T cells. We cotransfected HEK293T cells with one of three luciferase constructs containing the upstream regulatory region for miR-10a, miR-21, or miR-155 that contained the most proximal putative PBX1 binding sites (Fig. 8A) and plasmids expressing either the normal allele PBX1-b or the lupus allele PBX1-d cDNA. As shown in Fig. 8B, PBX1-b and PBX1-d increased the transcription of each of the three miRNAs, with higher levels obtained for PBX1-d than for PBX1-b for miR-10a and miR-155. These data strongly support transcriptional regulation of miR-10a, miR-21, and miR-155 by PBX1.

This study examined the role of Pbx1-d expression in CD4⁺ T cells relative to the phenotypes induced by the Sle1a1 lupus-susceptibility locus. Sle1a1 increases CD4⁺ T cell effector functions and reduces the number and function of Tregs, resulting in the production of autoreactive CD4⁺ T cells (9, 11). Pbx1-d overexpression in the CD4⁺ T cells of B6 mice resulted in autoantibody production and T cell activation, as well as fewer pTregs and reduced iTreg differentiation in vitro. These results are fully consistent with Pbx1-d overexpression being responsible for Sle1a1 CD4⁺ T cell activation and impaired iTreg differentiation or maintenance (9). However, the most striking phenotype of Pbx1-d–Tg mice was the expansion of T_FHs and associated TD immune responses. BM chimera and adoptive transfer studies showed that Pbx1-d overexpression resulted in a cell-intrinsic Ag-specific expansion of T_FHs. This phenotype was accentuated in recall responses, which represent the chronic stimulation by autoantigens better than primary responses. This suggests that Pbx1-d expression contributes to lupus by promoting T_FHs at the expense of iTreg differentiation, resulting in the production of class-switched affinity matured autoantibodies that are the hallmark of the Sle1 susceptibility locus and lupus pathogenesis.

SLE patients present high levels of circulating T_FH-like cells that correlate with disease activity (32, 33), and functional T_FHs were found in the kidneys of patients with lupus nephritis (34). T_FHs are the limiting factor for GC size and function, and there is abundant evidence from several mouse models that unrestricted T_FH expansion leads to systemic autoimmunity (35, 36). Multiple studies also found an association between decreased numbers or impaired function of Tregs in lupus patients and mouse models (37). The differentiation of effector CD4⁺ T cell subsets is a dynamic and plastic process, and cross-talk between the T_FH and Treg pathways was described (30, 38). Although rapid progress has been made in recent years in mapping out the molecular pathways leading to Treg (39) and T_FH (40) differentiation, many questions still remain, including the mechanisms by which the balance between these two subsets is maintained. We propose that PBX1 is involved in this process through its ability to regulate gene expression in a cell-specific manner through complex interactions with its cofactors. The overexpression of dominant-negative Pbx1-d leads to an imbalance in T cell homeostasis; a modest Pbx1-d overexpression limits the spontaneous expansion or maintenance of the Treg subset in B6.Sle1a1 mice, whereas a high overexpression of Pbx1-d expands T_FHs in Pbx1-d–Tg mice. Homeostatic expansion in the context of BM chimera revealed a role for Pbx1-d–Tg overexpression in Treg maintenance. Similarly, a recall response in a competitive setting revealed a much stronger response of T_FHs expressing Pbx1-d either as a Tg or endogenously. This could be due to a better survival of memory T_FHs after the primary response or to a better ability to respond to the challenge, two possibilities that we will explore in future experiments.

Interestingly, we found that the imbalance between Tregs and T_FHs was enhanced in the mLN. We previously associated Pbx1-d expression with an impaired response to RA [(15) and this study], and the major effect of RA in vivo occurs in GALT. Therefore, it is possible that the anatomy of T_FH expansion in Pbx1-d–Tg mice reflects the relative response to RA and commensal microbiota, an issue that will be explored in future studies. Pbx1-d led to T_FH expansion despite a relatively expanded T_FR subset, which is reminiscent of a recent study of mice with a specific deletion of PTEN in Foxp3⁺ T cells, in which T_FH expansion occurred as a result of impaired Treg functions but in the presence of unchanged or expanded T_FR numbers (41). In addition to Pbx1-d expression level, other factors may contribute to the differences between B6.Sle1a1 and Pbx1-d–Tg mice. B6.Sle1a1 mice express Pbx1-d in other cell types, such as mesenchymal stem cells, leading to an increased production of proinflammatory cytokines (21). Pbx1 also regulates the production of inflammatory cytokines in macrophages in response to apoptotic cells (42). Although we showed that the Sle1a1 T cell phenotypes are cell intrinsic (15), we cannot rule out that Pbx1-d expression in other cell types may have additional modulatory effects on CD4⁺ T cell differentiation.

MicroRNAs are essential mediators of Th cell plasticity, including the balance between Tregs and T_FHs (30). Numerous studies documented different miRNA profiles in human and murine SLE (43–45), and causal relationships between overexpression of either miR-21 or miR-155 and lupus were established (46, 47). We showed that increased expression of miR-10a, miR-21, and miR-155 occurs in CD4⁺ T cells before TC mice exhibit autoimmune manifestations. Importantly, increased Pbx1-d expression in the B6.Sle1a1 congenic and Pbx1-d–Tg CD4⁺ T cells showed the same pattern of miRNA expression, implicating Pbx1-d as a primary driver regulating the intrinsic expression of these three miRNAs in T cells, either directly or indirectly. The existence of PBX1 binding sites in the distal promoter of each of these three miRNAs that are sufficient to increase transcription in the presence of Pbx1 supports this hypothesis. If Pbx1-d functions as a dominant-negative transcription factor, it suggests that Pbx1-b is a negative regulator of these three miRNAs, especially miR-10a and miR-155. However, Pbx1 can bind to DNA indirectly through numerous cofactors (48, 49). Detailed studies will be necessary to unravel the molecular mechanisms by which Pbx1 isoforms regulate the expression of miR-10a, miR-21, and miR-155.

The exact role of each of these miRNAs in T cells overexpressing Pbx1-d could provide important cues on the mechanisms by which Pbx1 regulates T cell differentiation. miR-10a is overexpressed in B cells from lupus patients (44), and miR-10a stabilizes Tregs (50) and prevents the conversion of pTregs into T_FHs by targeting Bcl6 and its corepressor Ncor2 in a TGF-β and RA–dependent manner (51). Therefore, it was unexpected to find consistently high levels of miR-10a expression in the CD4⁺ T cells, including Foxp3⁺ T cells, of the lupus congenic mice characterized in this study. Moreover, the level of miR-10a was similar between Pbx1-d–Tg and TC CD4⁺ T cells, arguing that Pbx1-d overexpression is the main determinant of this dysregulation. miR-21 is elevated in lupus (46), and it regulates aberrant T cell responses by blocking PDCD4 expression (45). miR-21 expression has not been directly linked to either Tregs or T_FHs, but it is activated by STAT3 in T cells (52), and it is possible that its increased expression in Pbx1-d–Tg CD4⁺ T cells is secondary to their increased IL-21 expression. miR-155 is a central modulator of T cell functions (53). Receptor activation increases miR-155 expression in T cells, which regulates effector subset differentiation and maintenance. miR-155 deficiency was associated with decreased Th1 responses, as well as low Treg numbers, which fits with Pbx1-d Tg–expressing T cells showing high miR-155 expression, skewed Th1 humoral responses, and a reduced Treg population. Specific deletion of miR-155 in T cells showed an intrinsic requirement of miR-155 for T_FH differentiation (54). miR-155 deficiency normalized B cell functions and autoantibody production in FAS-deficient mice (47); however, the role of miR-155 expression in T cells was not addressed in this model. Our results demonstrate consistent high levels of miR-155 expression in the CD4⁺ T cells of the lupus congenic mice examined in this study, including Foxp3⁺ and Foxp3⁻ Pbx1-d–Tg CD4⁺ T cells. This is consistent with a model in which miR-155 favors T_FH differentiation over Treg differentiation and demonstrates that Pbx1-d overexpression is sufficient to drive this process. Interestingly, we found overexpression of the miR-17–92 cluster in Sle1a1 Tregs but no difference in Pbx1-d–Tg Tregs or in non-Tregs in either strain. miR-17–92 prevents pTreg differentiation (55), which fits with our findings that Tregs are more affected in Sle1a1 T cells than in Pbx1-d–Tg T cells. Moreover, miR-17–92 overexpression in lymphocytes leads to autoimmune phenotypes in mice (56), suggesting that it may be a mechanism by which Pbx1-d overexpression contributes to autoimmunity.

In conclusion, the phenotypes and CD4⁺ T cell differentiation in Pbx1-d–Tg mice are similar to those of B6.Sle1a1 mice, which establishes Pbx1 as the gene responsible for the Sle1a1 phenotype. It also highlights critical roles for lupus development in regulating Treg and T_FH differentiation and developing humoral autoimmunity. Pbx1-d–Tg mice provide a novel model to investigate T cell–intrinsic mechanisms that regulate Treg/T_FH homeostasis. Further studies are required to identify the Pbx1/Pbx1-d–regulated genes in T cells and their contribution to CD4⁺ T cell differentiation and effector functions.

We thank Dr. Edward Chan for advice on miRNAs. We also thank Leilani Zeumer, Nathalie Kanda, Shun Lu, and Yuxin Niu for outstanding technical help.

The authors have no financial conflicts of interest.