A missense mutation (R620W) of protein tyrosine phosphatase nonreceptor type 22 (PTPN22), which encodes lymphoid-tyrosine phosphatase (LYP), confers genetic risk for multiple autoimmune diseases including type 1 diabetes. LYP has been putatively demonstrated to attenuate proximal T and BCR signaling. However, limited data exist regarding PTPN22 expression within primary T cell subsets and the impact of the type 1 diabetes risk variant on human T cell activity. In this study, we demonstrate endogenous PTPN22 is differentially expressed and dynamically controlled following activation. From control subjects homozygous for the nonrisk allele, we observed 2.1- (p < 0.05) and 3.6-fold (p < 0.001) more PTPN22 transcripts in resting CD4+ memory and regulatory T cells (Tregs), respectively, over naive CD4+ T cells, with expression peaking 24 h postactivation. When LYP was overexpressed in conventional CD4+ T cells, TCR signaling and activation were blunted by LYP-620R (p < 0.001) but only modestly affected by the LYP-620W risk variant versus mock-transfected control, with similar results observed in Tregs. LYP overexpression only impacted proliferation following activation by APCs but not anti-CD3– and anti-CD28–coated microbeads, suggesting LYP modulation of pathways other than TCR. Notably, proliferation was significantly lower with LYP-620R than with LYP-620W overexpression in conventional CD4+ T cells but was similar in Treg. These data indicate that the LYP-620W variant is hypomorphic in the context of human CD4+ T cell activation and may have important implications for therapies seeking to restore immunological tolerance in autoimmune disorders.
Genome Wide Association Studies have identified over 60 genetic variants that are thought to confer varying degrees of susceptibility for the T cell–mediated autoimmune disease type 1 Diabetes (T1D) (1, 2). Protein tyrosine phosphatase, nonreceptor type 22 (PTPN22), which encodes lymphoid-tyrosine phosphatase (LYP), is one such candidate gene bearing the strongest T1D-risk association after the HLA class II and INS-IGF2 loci (3, 4). A missense C1858T single-nucleotide polymorphism (SNP) in the PTPN22 coding sequence results in an LYP variant with an arginine-to-tryptophan substitution at position 620 (R620W) that is associated with increased risk for a number of autoimmune conditions, including T1D, rheumatoid arthritis, systemic lupus erythematosus, Graves’ disease, vitiligo, and myasthenia gravis (3, 5–10). Most early studies of PTPN22-associated risk centered around lymphocytes, given their central role in autoimmune pathogenesis. However, to date, no consensus has been reached regarding whether LYP R620W constitutes a gain or loss of function for phosphatase activity (11, 12). Moreover, it remains poorly characterized how the risk variant influences human regulatory T cell (Treg) versus conventional T cell (Tconv) function in a manner that facilitates autoimmune disease development.
Ptpn22 was originally identified in the mouse (13) and later in humans (14) as a cytoplasmic phosphatase predominantly expressed in lymphoid cells. It is composed of a catalytic domain responsible for its phosphatase activity as well as a series of proline rich domains that regulate activity and protein interactions, most notably, with C-terminal Src kinase (CSK) (15). In T cells, LYP dephosphorylates several proximal TCR signaling molecules including LCK, CD3ε, CD3ζ, ζ-chain associated protein kinase 70 (ZAP-70), and VAV to downmodulate TCR signaling and is, therefore, predicted to have a central role in T cell selection, activation, and differentiation (16). It is known that LYP also functions in B cells, as well as myeloid cells (17–21). As such, LYP has been shown to modulate BCR signaling (22), TLR signaling (23–25), NLRP3 signaling (19), and outside-in integrin signaling (26, 27). Hence, LYP serves many functions across the cellular immune landscape, and it is possible that modulation of LYP expression levels may play an important role during immune cell development and activation.
In murine models, Ptpn22 deficiency has been associated with increased Treg numbers in both the thymus and the periphery as well as with protection against the induction of experimental autoimmune encephalomyelitis (28, 29). Similarly, Ptpn22 deficiency resulted in reduced disease severity and incidence in the SKG mouse model of rheumatoid arthritis as well as enhanced tolerance of allogeneic islet transplants (30, 31). Conversely, although Ptpn22 deficiency does not confer overt autoimmune disease, it has been shown to facilitate autoimmunity when combined with other genetic risk factors, such as hyperactive CD45 E613R mutation, BXSB background, and the KBxN arthritis model (32–34). Similar to its known association with human T1D susceptibility, Ptpn22 has been identified as the candidate gene for the insulin-dependent diabetes 18.2 (Idd18.2) diabetes susceptibility locus in the nonobese diabetic (NOD) mouse model (35). Interestingly, whereas silencing Ptpn22 expression in NOD mice resulted in reduced incidence of diabetes (36), expression of the orthologous human risk variant in the endogenous Ptpn22 locus promoted autoimmunity in the NOD strain (37, 38). Altogether, these observations support the notion that LYP likely influences T cell development and selection in the thymus as well as activation in the periphery, suggesting that the autoimmune-associated variant may confer multifaceted effects on host immunity.
T cells isolated from the peripheral blood of human T1D patients carrying the LYP-620W variant exhibited reduced proliferation and calcium flux in response to polyclonal stimulation (39), and heterozygosity for LYP-620W was associated with reduced IL-10 production after stimulation of memory CD4+ T cells (17). There is conflicting evidence as to whether the LYP-620W variant impairs interactions of LYP with CSK, which are important for regulating TCR signal strength (32, 40, 41). Furthermore, it is unclear whether the LYP/CSK interaction strengthens or weakens inactivation of the Src family kinase LCK (40, 41). Thus, there is a need to investigate the complex effects of the LYP-620W variant. In this study, we used a lentiviral gene delivery system to stably overexpress the LYP-620R nonrisk and LYP-620W risk variants in human Tconvs and Tregs. Our objective was to assess the cell-intrinsic effect conferred by LYP-620W in the CD4+ T cell compartment while mitigating epistatic biological variability inherent in studies of genotype-selected donors. The methods in this study, may also be applied to other cell types, risk variants, or combinations thereof, to facilitate pathway-directed therapies for autoimmune disorders.
Materials and Methods
PTPN22 genotype was determined using a TaqMan Genotype Assay (Thermo Fisher Scientific) for rs2476601 per manufacturer instructions. In this report, we refer to the reverse strand genotype (rs2476601 C > T) to match the PTPN22 gene orientation.
Healthy control donors (median age, 30.9; range, 23.6–38.0 y; n = 8) that were homozygous for the nonrisk (C/C) PTPN22 variant at rs2476601 were selected from the University of Florida Diabetes Institute Study Bank. Fresh venous blood was collected in sodium-heparinized vacutainer tubes (BD Biosciences) after informed consent in accordance with the University of Florida Institutional Review Board approved protocol (IRB201400703). Leukopack blood donations were also obtained from rs2476601 C/C homozygous individuals under Institutional Review Board–exempt protocols from Life South Blood Centers (n = 9) in sodium citrate. All studies involving human samples were conducted in accordance with the Declaration of Helsinki.
Cell enrichment and sorting
For FACS experiments, whole blood or leukopaks were pre-enriched by CD4+ T cell–negative selection with RosetteSep (StemCell Technologies) prior to density gradient centrifugation. CD4+ T cells were labeled with fluorescent Abs against CD4 (RPA-T4), CD45RA (HI100), CD197 (G043H7), CD127 (A019D5), and CD25 (BC96) as previously described (42). Tregs were FACS-purified as CD4+CD25+CD127lo/–, whereas naive Tconvs were FACS-purified as non-Treg CD4+CD127+CD45RA+CD197+, and memory Tconvs were FACS-purified as non-Treg CD4+CD127+CD45RA– on an FACS Aria III (BD Biosciences) as previously described (43).
PTPN22 transcript quantification
Leukopak-enriched CD4+ T cell subsets were used for endogenous PTPN22 expression kinetic studies. RNA was prepared from 106 naive Tconv, memory T cells, and Treg either immediately after sorting or at various time points following activation. Tconv were activated with a 1:1 bead to cell ratio of Dynabeads Human T-Activator CD3/CD28 (Thermo Fisher Scientific), and Treg were activated with 4:1 ExpAct beads (Miltenyi). Cells were supplemented with growth factor cytokines every 2–3 d, specifically 300 U/ml IL-2 for Tregs or 20 U/ml IL-2 for Tconv, and after 7 d, 5 ng/μl IL-7 was also added for Tconv. RNA was Qiashredded and extracted using RNeasy Plus Kit (Qiagen) and reverse-transcribed into cDNA using Protoscript First Strand cDNA Synthesis Kit (New England Biolabs). The expression level of PTPN22 was measured using an absolute quantification real-time PCR approach on a Lightcycler 480 platform (Roche Diagnostics). Each reaction well contained 1μL PTPN22 TaqMan Gene Expression Assay solution (Thermo Fisher Scientific), 2 μL of LightCycler480 Probe Master Mix (Roche Diagnostics), and 3 μL of cDNA in a 20-μL reaction. To create the DNA standards for the quantification real-time PCR assay, the pUC57.LYP620R.fuT2A.eGFP (see Lentiviral vector design and construction) plasmid containing PTPN22 coding DNA sequence (CDS) was serially diluted 10-fold. Λ DNA was used to bring all standards to equivalent concentrations of total DNA (1.2 ng/µL). Six concentrations were prepared ranging from 1.2 fg/µL to 12 ng/µL of plasmid, in addition to a no-template–control consisting of 1.2 ng/µL Λ DNA only. A total of 3 µL of standard DNA was used in reactions as above. Standards were converted to transcript copy number by dividing the mass of plasmid per reaction in grams by the plasmid formula weight, where 650 g/mol was used as the average nucleotide pair formula weight times the plasmid length. Copy number of each standard per well was log-transformed and plotted against the corresponding Ct value to form a standard curve. The standard curve R2 > 0.998 for each batch, and all samples were within range of the standard curves. PTPN22 expression level of each sample was extrapolated from the standard curve as previously done (44).
Lentiviral vector design and construction
Design of the LYP620R.fuT2A.eGFP construct began with the human PTPN22 full-length isoform CDS containing the nonrisk variant (C at position 1858 of the CDS). Two endogenous EcoRI restriction enzyme sites (GAATTC) within the CDS were synonymously mutated, and the stop codon was removed (Supplemental Fig. 1A, 1B). This was followed by a furin cleavage site, a T2A self-cleaving peptide sequence (45), and an enhanced GFP (eGFP) coding sequence (stop codon included). The LYP620R.fuT2A.eGFP was synthesized in a pUC57 cloning vector by GenScript (Piscataway, NJ). Site-directed mutagenesis by PCR with PFU DNA polymerase followed by DpnI digest was performed to generate the LYP620W.fuT2A.eGFP construct, which contains the PTPN22 C1858T variant that encodes the LYP R620W amino acid change. These constructs were then subcloned into the pFUGW third generation lentiviral vector (LV) (46). pFUGW expressing only eGFP served as the control vector. Lentiviral preparations were conducted as previously described (47). All constructs were validated by restriction enzyme digest and sequence analysis.
FACS-isolated naive Tconv were activated with a 1:1 bead to cell ratio of Dynabeads Human T-Activator CD3/CD28 (Thermo Fisher Scientific), and Treg were activated with 4:1 ExpAct beads (Miltenyi). On day 2, the cultures were infected with 0.5 titrated units of lentivirus per cell in the presence of 8 μg/ml protamine sulfate (Sigma-Aldrich) and spinoculated as previously described (47). The cultures were then allowed to expand and rest for at least 21 d after activation, during which they were supplemented with growth factor cytokines every 2–3 d at dosages of 300 U/ml IL-2 for Tregs and 20 U/ml IL-2 and for Tconvs. After 7 d, 5 ng/μl IL-7 was also supplemented every 2–3 d with IL-2 for Tconvs. Portions of the cultures were FACS-isolated into stably transduced (eGFP+) and mock-transduced (eGFP–) for cytokine and suppression assays. All other assays used the unsorted fractions of the cultures so that the results from stably transduced cells could be normalized to mock-transduced cells.
FACS-isolated transduced (eGFP+) and mock-transduced (eGFP–) fractions were lysed with cold lysis buffer (Cell Signaling Technologies) containing protease inhibitors (Sigma-Aldrich). Protein was quantified via bicinchoninic acid (BCA) assay (Thermo Fisher Scientific). A total of 5 μg of protein was denatured with Laemmli sample buffer at 95°C for 5 min and run on Mini Protean TGX Gels (Bio-Rad Laboratories). Transfers were performed using the Transblot Turbo Transfer System (Bio-Rad Laboratories) and Transblot Turbo Transfer Packs (Bio-Rad Laboratories). Membranes were washed with 1× TBST, blocked with 5% milk in TBST for 1 h, and incubated with primary Ab (rabbit anti-human LYP [clone D6D1H], β-actin [polyclonal], or GFP [clone D5.1]; Cell Signaling Technologies) overnight at 4°C. Membranes were then washed with TBST, and incubated with anti-rabbit IgG-HRP secondary Ab (1:1000; Cell Signaling Technologies) for 1 h at room temperature. Membranes were subsequently washed, developed with Western DuraSignal Substrate (Thermo Fisher Scientific), and imaged on the GeneGnome XRQ (Syngene). Western blot band quantification was performed in ImageJ (48). Serial blots were stripped with Restore Western blot Stripping Buffer (Thermo Fisher Scientific) between probes for 15 min at room temperature per manufacturer protocol.
Stably transduced Tconvs or Tregs, and their mock-transduced internal control cells, were labeled with 2.5μg/ml FuraRed calcium responsive dye (Thermo Fisher Scientific) in 1% BSA/HBSS at 37°C for 10 min, followed by labeling with Live/Dead Near-IR Viability Dye (Thermo Fisher Scientific). Cells were then resuspended in HBSS (with Ca2+) containing 10 μg/ml goat anti-mouse IgG F(ab')2 crosslinking Ab (Jackson ImmunoResearch). Next, cells were warmed to 37°C, and fluorescence at 605 nm by a 405-nm laser and at 660 nm by a 532-nm laser was collected on an LSRFortessa (BD Biosciences) to indicate high and low calcium, respectively (Supplemental Fig. 3A–C). After recording 30 s of basal fluorescence, anti-human CD3 (OKT3) was added at 1 μg/ml, and fluorescence was acquired for 2 min. FlowJo software (FlowJo) was used to analyze the mean fluorescence intensity (MFI) kinetics of FuraRed for the high and low [Ca2+] channels after first gating on live cells by forward and side-scatter morphology and viability dye exclusion. The percent change from mock in fluxed calcium was calculated as the percent difference between curves of the mock and transduced cells (Supplemental Fig. 3D–G).
LV-transduced Tconvs or Tregs and their mock-transduced internal control cells were labeled with viability dye, then washed into assay buffer (RPMI base medium) and plated at 106 cells per well in 96-well V-bottom plates in the presence of 10 μg of anti-mouse IgG crosslinking Ab (Jackson ImmunoResearch). Anti-CD3 (2 μg of OKT3) was added at times –10, –5, and –2 min. At time 0, wells were fixed with 4% formalin for 10 min. Cells were then washed, permeabilized with methanol, and stained with PE-conjugated anti–phospho-ERK (anti-pERK; clone 197G2, Cell Signaling Technology) for 60 min. Data were acquired by flow cytometry on an LSRFortessa (BD Biosciences), and pERK MFI of live cells was assessed using FlowJo software (Flowjo).
Proliferation and suppression assays
LV-transduced (eGFP+) and internal control mock-transduced (eGFP–) cells were labeled with Cell Trace Violet (Thermo Fisher Scientific) cell-tracking dye and activated with either beads, autologous APCs, 10ng/ml phorbol 12-myristate 13-acetate (PMA) plus 1 μM ionomycin, or K562 cell line artificial APCs (aAPCs) which do not express MHC class I or class II molecules but do express murine Fc’ receptor (CD64) and human CD86 (49). In brief, beads were used at a 1:1 bead to cell ratio as described in Cell transduction above. Autologous APCs, generated from PBMCs that were CD3-depleted (RosetteSep) and exposed to 3000 Rad of γ irradiation, were cultered 1:1 with T cells in the presence of 2μg/ml anti-CD3 and 1μg/ml anti-CD28. To generate aAPCs, K562 cells (human chronic myeloid leukemia line) expressing murine CD64 and human CD86 were incubated with various ratios of mouse anti-human CD3 (OKT3) and isotype Ab, and then irradiated with 3000 Rad. The ratios of isotype to OKT3 used were 1:0, 9:1, 1:1, and 0:1, such that the resulting aAPCs were loaded with 0, 10, 50, and 100% anti-CD3, respectively. aAPC were similarly labeled to include mouse anti-human CD4 (OKT4) in ratios of 1:0:0, 1:0.5:0.5, and 0:0.5:0.5 (isotype:OKT3:OKT4), such that the resulting aAPCs were loaded with 0, 5, 25, and 50% each activating Ab. These were cultured at a 1:4 aAPC to T cell ratio.
Suppression assays were performed as previously described (50). Briefly, titering proportions of Tregs labeled with Cell Proliferation Dye eFluor670 (CPD670, Thermo Fisher Scientific) were cocultured with autologous Cell Trace Violet–labeled responder Tconv (Tresp) and unlabeled APCs ranging from 1:1:1 to 0:1:1 Treg/Tresp/APC ratios. Suppression assay cultures were set up in triplicate in round-bottom 96-well plates and activated with 2 μg/ml anti-CD3 and 1 μg/ml anti-CD28. Two versions of the suppression assays were performed. The first, which was designed to assess the response of LYP-modulated Tconv to Treg suppression, consisted of a combination of stably transduced and mock-transduced Tconv (not eGFP-sorted) being suppressed by unmodulated Treg (mock-transduced sorted eGFP–). The second, which was designed to assess suppression by LYP-modulated Treg, consisted of unmodulated Tconv (mock-transduced sorted eGFP–) being suppressed by stably transduced Treg (sorted GFP+).
Proliferation and suppression assay data were acquired on an LSRFortessa (BD Biosciences), and analyze using FlowJo software (Flowjo). Live T cells were gated by forward and side-scatter morphology, followed by viability dye exclusion. Proliferation of live-gated cell-tracking dye+ cells was assessed by expansion index. Suppression was calculated as the percent decrease in Tresp proliferation when cocultured with Tregs as compared with when cultured without Tregs.
Surface marker activation kinetics
Stably transduced (eGFP+) and internal control mock-transduced (eGFP–) cells were activated with irradiated autologous APCs at 1:1 in the presence of 2 μg/ml anti-CD3 and 1 μg/ml anti-CD28. The cultures were harvested at 0 h, 4 h, 24 h, 48 h, 72 h, and 144 h and stained with Live/Dead Near-IR viability dye (Thermo Fisher Scientific). Next, the cells were washed into 2% FCS in PBS and labeled with CD278 (C398.4A, BioLegend), CD69 (FN50, BioLegend), CD4 (RPA-T4, BioLegend), CD226 (11A8, BioLegend), CD25 (BC96, BioLegend), CD279 (EH12.2H7, eBiosciences), CTLA4 (L3D10, BioLegend). The cells then were fixed in 1% formaldehyde in PBS. Data were acquired on an LSRFortessa (BD Biosciences) and was analyzed using FlowJo software (Flowjo). Live lymphocytes from CD4+ T cell cultures were gated by forward and side-scatter morphology, followed by viability dye exclusion before extracting surface marker expression. Fluorescence minus-one controls were used to set gates for marker positivity.
Stably transduced (eGFP+) T cells were activated with irradiated autologous APCs at 1:1 in the presence of 2 μg/ml anti-CD3 and 1 μg/ml anti-CD28. Supernatants were collected at the specified time points, and cytokines were detected via multiplexed bead-based ELISA (42).
Data were analyzed by one-way ANOVA with Bonferroni post hoc analysis or by two-way ANOVA with Bonferroni posttest, and graphs were prepared using GraphPad Prism software v6. The p value < 0.05 was considered significant. Subject matching was used to compare donors’ cells across experimental conditions (biological replicates).
PTPN22 expression varies by lymphocyte subset and activation state
We examined endogenous PTPN22 expression in human naive, memory, and regulatory CD4+ T cell subsets at rest and following in vitro activation. CD4+ T cells were sorted from leukopacks determined to be homozygous for the T1D nonrisk allele, which encodes LYP-620R. Because the transcriptional profile of activated cells is dramatically altered from that of the quiescent state, such that the standard array of housekeeping genes is broadly upregulated, we employed absolute quantification to assess PTPN22 expression. At rest, naive Tconv expressed significantly less PTPN22 (387 ± 83 transcripts per nanogram total RNA) than memory Tconv cells (741 ± 169 transcripts per nanogram total RNA, p < 0.05) and Tregs (1014 ± 120 transcripts per nanogram total RNA, p < 0.001) (Fig. 1A). Following polyclonal activation, all three subsets rapidly increased PTPN22 expression, peaking at 24 h (Fig. 1B). Naive CD4+ Tconv increased 12.00 ± 5.45–fold from the quiescent state whereas memory Tconv increased 4.01 ± 0.89–fold (p < 0.001 versus naive) and Treg increased 2.85 ± 1.31–fold (p < 0.01 versus naive). Following the initial increase, PTPN22 expression decreased and plateaued around day 7. Taken together, we found PTPN22 expression to differ across T cell subsets at rest and in response to activation.
Overexpression of PTPN22 decreases TCR signaling
Next, we examined whether LYP-620R and -620W may differentially modulate T cell function in Tconv and Treg populations. Naive CD4+ Tconv and Treg were sorted from PBMCs from normal healthy control donors that were homozygous for the T1D nonrisk PTPN22 allele (C/C at rs2476601 encoding LYP-620R). Cells were then activated and transduced with LV to express bicistronic constructs of the nonrisk LYP-620R or the risk LYP-620W variant with an eGFP reporter following a 2A element, or with eGFP alone for a vector control condition (Supplemental Fig. 1A-B). The cultures were expanded, followed by an extended rest period for a total of 21–28 d to allow for reversion of activation programs. We confirmed stable overexpression of PTPN22 transcripts and LYP protien for each variant (Supplemental Figs. 1C–E and 2).
The proximal signaling events resulting from TCR ligation lead to assembly and activation of the linker for activation of T cells signalosome, followed by diverging downstream signaling pathways. These include calcium signaling and MAPK/ERK pathways (51). We sought to determine the functional impact of the nonrisk and risk LYP variants on these downstream signaling pathways in Tconv and Treg. We first assessed Ca2+ flux by comparing stably transduced cells to nontransduced (eGFP–) internal culture control cells (Supplemental Fig. 3). As expected, LYP-620R expressing Tconv exhibited 25.8 ± 3.7% less Ca2+ flux than nontransduced Tconv, whereas transduction with eGFP alone had no effect (Fig. 2A). A similar effect was observed for LYP-620R Tregs in which Ca2+ flux was diminished by 40.3 ± 13.3% compared with nontransduced Tregs (Fig. 2B). Conversely, LYP-620W expression only resulted in 4.9 ± 1.6% and 7.9 ± 2.7% less Ca2+ flux in Tconvs and Tregs, respectively (Fig. 2A, 2B). These differences were not different from T cells transduced with eGFP alone. Thus, the risk LYP-620W variant less effectively downregulated TCR-induced Ca2+ flux.
We next assessed the impact of LYP modulation on the MAPK/ERK pathway via pERK signaling following TCR ligation. As with Ca2+ signaling, we observed decreased pERK relative to mock-transfected cells in both Tconvs and Tregs when LYP-620R was overexpressed (Fig. 2C, 2D). Once again, the impact of LYP-620W on TCR signaling was diminished in both subsets when compared with LYP-620R. Taken together, these results indicate that the LYP risk variant has a diminished capacity to regulate TCR signaling activity.
The risk variant of PTPN22 permits increased T cell activation
In response to Ca2+ flux and pERK signaling, T cells express receptors on their surface to direct activation. We further explored the impact of LYP variants by examining the expression of a set of these activation markers. Stably transduced T cells were activated with autologous APCs, and surface expression of activation markers was assessed at 0, 4, 24, 48, 72, and 144 h by flow cytometry. As expected, CD69, ICOS, CD25, CD226, and PD-1 were observed to be dynamically expressed following Tconv activation (Supplemental Fig. 4A–D) (52–56). Likewise, Tregs also exhibited activation-induced expression kinetics for CD69, ICOS, CD226, and CTLA-4 (Supplemental Fig. 4E, 4F) (57). Interestingly, although most activation marker kinetics involved an intensification followed by a reduction of surface expression, the kinetics of CD226 on Tregs waned before increasing to a level higher than baseline (Supplemental Fig. 4H). When normalized to internal mock-transfected controls, the activation marker kinetics on both Tconvs (Fig. 3A–E) and Tregs (Fig. 3F–J) were blunted by overexpression of LYP-620R, with the exception of CD226 on Tconv (Fig. 3D). Similar to what was observed with Ca2+ and pERK signaling, cells expressing the LYP-620W variant were less impacted, supporting the notion that the R620W variant confers a loss of function for LYP in terms of modulating of T cell activation.
The PTPN22 risk variant permits proliferation in Tconv, but not in Treg
We demonstrated that the two LYP variants differentially modulate TCR signaling and expression of T cell activation markers. We next asked whether overexpression of the LYP variants would differentially modulate T cell proliferation. Activation-induced proliferation of Tconv and Treg cells was assessed by dilution of cell-tracking dye following in vitro activation. Transduced cells (eGFP+) were cultured with mock-transduced (eGFP–) internal culture control cells. The cultures were activated with either anti-CD3 and anti-CD28 with irradiated APCs, anti-CD3 and anti-CD28 bound to microbeads, or with PMA and ionomycin. Whereas expression of the eGFP reporter did not affect proliferation, overexpression of LYP-620R reduced APC-mediated proliferation of Tconvs and Tregs by 35.32 ± 15.30% and 19.80 ± 17.26%, respectively (Fig. 4A, 4B). In Tconvs, the LYP-620W variant was once again deficient in its ability to downregulate activation, as it did not impede proliferation to the same degree that was observed for the LYP-620R variant (Fig. 4A). However, the LYP-620W variant was capable of blocking Treg proliferation to a similar degree as the LYP-620R variant (Fig. 4B), indicating a differential impact of PTPN22 on Tconvs and Tregs.
As expected, mitogenic activation by PMA and ionomycin was not altered by LYP overexpression, as all transduced Tconvs and Tregs proliferated to a comparable degree as their internal mock control cells (Fig. 4A, 4B). Unexpectedly however, proliferation was also unaltered in transduced cells that were activated by microbeads (Fig. 4A, 4B). To assess whether this was because of TCR signal strength, we repeated the experiment with K562 aAPCs loaded with titrated ratios of activating and isotype Abs. The aAPCs did not induce proliferation when loaded with 0% anti-CD3 (100% isotype), whereas 10% and 50% anti-CD3 induced incrementally more proliferation, with a plateau from 50 to 100% anti-CD3 (Supplemental Fig. 5A). In all cases, the proliferation of transduced cells was similar to their internal controls (Supplemental Fig. 5B). Similar results were found when CD4 was included in the activation signal to allow for LYP-LCK interactions (Supplemental Fig. 5C, 5D) (30, 58). This suggests that optimized signal strength, as well as fluid membrane interface are not sufficient to recapitulate the differential inhibition of proliferation by the two LYP variants, and that other factors supplied by natural APCs (e.g., adhesion molecules, soluble factors, and/or costimulatory and coinhibitory ligand and downstream signaling interactions) may be required (30, 58).
Overexpression of PTPN22 influences T cell cytokine production
CD4+ effector T cells modulate host immune responses after encountering cognate Ags, in part, through the production of an array of cytokines. Thus, we sought to assess the impact of LYP variant expression on the secretion of cytokines important in driving Th1- and Th2-associated immunity. Sorted transduced Tconvs were activated with APCs, anti-CD3, and anti-CD28, and culture supernatants were assayed for cytokines over a 72-h time period. The net accumulation of IL-2 in this system is a function of activation-induced secretion and of consumption via the IL-2R because the high-affinity component, CD25, is dramatically upregulated in this time period (Supplemental Fig. 4) (59). We found that the net accumulation of IL-2 was not altered by overexpression of the LYP-620R variant, but was increased at the 24-h time point by the LYP-620W variant (Fig. 5A). The secretion of IL-4, IL-5, IL-9, IL-10, and IL-13 was robustly inhibited for the first 48 h by expression of either LYP variant. At the72-h time point, inhibition occurred to a somewhat lesser degree by LYP-620W, thought this difference was not significant between variants (Fig. 5B–F). Finally, and somewhat surprisingly given the documented Th1 signature implicated in several of the autoimmune disorders that are associated with the rs2476601 risk variant (60, 61), IFN-γ secretion was not altered by overexpression of either variant of LYP (Fig. 5G). This result is in agreement with studies of Ptpn22 knockout mice (28). Taken together, these data demonstrate that both LYP variants result in similar levels of effector cytokine secretion.
Treg-mediated suppression is altered by overexpressed PTPN22
Our data demonstrated that PTPN22 is more highly expressed in Treg as compared with Tconvs (Fig. 1A). We, therefore, hypothesized that the PTPN22 risk variant may influence disease risk by altering Treg-mediated suppression. Hence, we assessed whether transduced Treg were deficient in their suppressive capacity and whether transduced Tconv were refractory to Treg suppression. To determine the effect of overexpressed LYP variants on Tresp, irradiated autologous APCs and titrated proportions of Treg were cocultured with transduced Tconv (eGFP+) and mock-transduced (eGFP–) internal control Tconv. We found that, similar to our previous results (Fig. 4A), overexpressed LYP-620R significantly repressed Tresp proliferation whereas repression by LYP-620W was dramatically reduced across the Treg/Tresp range (Fig. 6A). However, the relative suppression of transduced Tresp did not differ significantly from their internal mock controls (Fig. 6B). This suggests that the Tconv-intrinsic susceptibility conferred by LYP-620W is due to reduced control over activation, proliferation, and effector mechanisms, rather than a defect in the ability of effector cells to be suppressed by Tregs.
Finally, we assessed the effect of the LYP variants on Treg-suppressive capacity by coculturing titrated amounts of sorted transduced (eGFP+) Treg with irradiated autologous APCs and untransduced Tresp. We found suppression by Tregs overexpressing LYP-620R to be similar to eGFP reporter Tregs, whereas suppression by Tregs expressing the risk LYP-620W variant was slightly increased (Fig. 6C). Consistent with our previous result (Fig. 4B), Treg proliferation in the suppression assay coculture was similar for LYP-620R and LYP-620W (Fig. 6D). This indicates that the difference in suppressive capacity was not due to differences in Treg proliferation, and suggests a Treg intrinsic impact of LYP-620W to enhance suppressive function.
To date, much of the literature has attempted to categorize rs2476601, the PTPN22 C1858T autoimmune-associated missense mutation, as a loss- or gain-of-function variant (11, 12). Because the LYP-620W variant confers risk for T cell–mediated autoimmune diseases (3, 5–8), we were interested in its allotypic effects on TCR signaling and function in the Tconv and Treg subsets. We first evaluated endogenous PTPN22 expression in primary human T cells homozygous for the T1D nonrisk allele encoding LYP-620R. We then employed a lentiviral gene delivery system to induce constitutive overexpression of the T1D nonrisk variant, LYP-620R, or T1D risk variant, LYP-620W, in primary human CD4+ Tconv and Tregs. There are a few advantages to this approach. First, it enables expansion of primary cell material that stably expresses either the nonrisk or risk variant for long-term assays. Second, by isolating LYP variant overexpression to CD4+ T cells, the complicating effects of the LYP variant from other immune subsets are removed so that CD4+ T cell–intrisic effects can be studied. Finally, it controls for epistasis and the associated biologic variability by assessing the functional differences of the LYP variants within the same subjects.
Our observation that endogenous expression of PTPN22 varies by T cell subset and is dynamic throughout activation suggests that SNP-related functional effects are multifaceted and may differ across cell types in a manner that is subject to temporal control and activation state in the periphery. Importantly, our assay measured both splice variants which encode the 85-kDa and 105-kDa LYP isoforms that are differentially expressed in T cells at rest and after activation, respectively (14), thereby measuring total endogenous PTPN22 expression. Because of the minor allele frequency of rs2476601, we were unable to assess PTPN22 expression kinetics of the 620W variant using the methods in this study. We will pursue this question using scRNA-sequencing technology, which requires many fewer cells, thus enabling time course examination of genotype-selected, cryopreserved PBMCs. We next demonstrated that, as expected, overexpression of LYP-620R decreases T cell Ca2+ flux, pERK signaling, surface expression of activation markers, and proliferation in both Tconvs and Tregs. However, with the exception of Treg proliferation, overexpression of LYP-620W had little effect on TCR activation-induced responses. We therefore conclude that LYP-620W is hypomorphic in terms of CD4+ T cell responses to TCR activation.
Interestingly, proliferation was only affected by PTPN22 modulation in the context of APC activation, indicating a requirement for a natural immune synapse or specific costimulatory signaling. Burn et al. showed that primary human T cells with the endogenous LYP-620W variant had increased pERK1/2 induced by LFA-1/ICAM-1 outside-in signaling as compared with T cells expressing the LYP-620R variant (27). This resulted in stronger integrin-mediated adhesion that was resistant to shear forces, which may also apply to cell-cell interactions. Thus, whereas interactions between autoreactive T cells and APCs may normally be transient (62), expression of the LYP-620W variant may stabilize the immune synapse allowing for more efficient activation and initiation of autoreactivity. Similar experiments comparing LYP variants in the APC are necessitated to determine if the reported effect on synapse stability occurs bidirectionally.
Treg proliferation was blunted by LYP overexpression, but contrary to Tconvs, Treg proliferation was not differentially impacted by the two LYP variants. This was the case when cultured alone with APCs as well as with APCs and Tconvs in the context of suppression assays. Overall suppressive capacity was not affected for LYP-620R Tregs and was enhanced for LYP-620W Tregs. The data presented in this study further indicate that certain mechanisms of suppression, perhaps IL-2 consumption or regulatory cytokine production, may not be affected by LYP expression, while other mechanisms, such as contact-dependent coinhibitory receptor interactions, may only be affected by the LYP-620W variant. It was previously shown that Tregs from Ptpn22-deficient mice had improved suppressive function because of increased and prolonged LFA-1 interactions between Tresps and Tregs (26). This is in line with the finding that LYP-620W is a loss of function variant with respect to outside-in integrin signaling, which results in sustained LFA-1 interactions (27). We previously showed that Tregs expressing a higher affinity TCR are more potent suppressors (63). Sustained Treg interactions conferred by LYP-620W may be functionally similar to higher affinity TCR engagement, resulting in enhanced suppressive capacity per cell (Fig. 6C). Nevertheless, enhanced Treg-suppressive capacity does not support a mechanism for autoimmune pathogenesis. In fact, suppression assays from rs2476601 T/T individuals were recently shown to exhibit lower suppression than those from C/C subjects (64). Thus, we posit that progression to autoimmunity may occur when the LYP-620W deficit in restraint of Tconv proliferation overcomes the LYP-620W enhancement of Treg suppression.
In terms of cytokine production, IL-2 accumulation was not altered in T cells overexpressing LYP-620R, but was increased by the LYP-620W variant. This finding is consistent with observations of the murine Pep-R619W ortholog knock-in (37). Although further studies are required to determine if this is attributable to increased production or reduced consumption, this result may corroborate the observed differential effects of LYP variants on Tconv proliferation. The secretion of effector cytokines (IL-4, IL-5, IL-9, IL-10, and IL-13) was inhibited by overexpression of either variant, but this effect appeared to wane in LYP-620W Tconvs by the 72-h time point. IFN-γ production by Tconvs was not altered by overexpression of either LYP variant. This is in contrast to observations in a murine model of Ptpn22 overexpression, which exhibited decreased IFN-γ production in effector T cells (65). We also did not observe increased IFN-γ production in LYP-620W–expressing cells relative to LYP-620R–expressing cells. Conversely, Anderson et al. recently reported increased IFN-γ production and activation marker expression in PTPTN22-deficient human CD4+ T cells after 48 h of reactivation (66). Those results suggests that a loss-of-function variant may also result in enhanced IFN-γ secretion. Although the reason for these observed differences between experimental model systems is not clear at this point, we posit that constitutive overexpression and CRISPR knockout systems impact LYP stoichiometry with TCR signaling molecules in an opposing manner. As such, cytokine regulation may not be affected in the same manner as activation and proliferation.
Faced with inconsistent reports in the current literature (12, 17, 39), our data nonetheless support the notion that the risk LYP-620W variant is generally a loss-of-function variant in Tconv, as TCR signaling, activation, and proliferation are less affected by its overexpression compared with LYP-620R. Still, the fundamental question of whether the LYP-620W functional decrease results from abrogated phosphatase activity or altered localization remains outstanding. Although altered localization may also expose a distinct set of targets to dephosphorylation, it likely also disengages LYP-620W from the TCR signalosome, where its modulatory capacity is specifically tied to weaker TCR signaling (67, 68). In this regard, differences in activation threshold, signal strength, and costimulation between Tconvs and Tregs may make them differentially amenable to LYP modulation (69, 70). In the context of recurrent autoreactive T cell responses, which tend confer weaker low-affinity TCR signaling in memory T cells (62, 71), the net effect may be the outgrowth of autoreactive Tconvs relative to Tregs. Future studies will examine the nature of our observed loss of function.
Despite the advantages of the lentiviral gene delivery methods used in this study, there are caveats to consider when interpretating our results. First, although constitutive overexpression induced by our LV constructs is useful for unmasking differences between variants, it does not provide equivalent kinetics to those observed for endogenous PTPN22 expression. Hence, more careful control of expression may be required for mechanistic studies of temporal LYP activity. Second, in this experimental system we are examining the effect of LYP variant expression in the context of endogenous LYP-620R. In that regard, these data are analogous to comparing homozygous nonrisk to heterozygous risk LYP variants, with the caveat of continual and mild overexpression (3- to 6-fold higher LYP protein expression, Supplemental Fig. 1E) from the LV construct. A quantitative PCR assay could have been designed for a more thorough analysis of exogenous versus endogenous expression. This is an important consideration because amount of LYP relative to its binding partners, most notably CSK, influences the degree of dephosphorylation of TCR signaling (32, 72). Further studies are required to determine how this modest LYP overexpression relative to CSK impacts the biological outcome. Third, for efficient integration to occur, we transduce actively dividing cells. We are therefore unable to examine LYP functional aspects during primary activation, for example to study naive versus memory or to determine the impact on polarization. Finally, we have only examined the 620W effect in the full-length isoform. Several isoforms of LYP are expressed in human T cells, and variation in phosphatase activity, cellular localization, and association with autoimmune disease have been reported (73, 74).
Considering the various roles for LYP in modulating immune responsiveness of T cells, B cells, APCs, neutrophils, and more (11, 17, 23, 24, 29, 75), it is not surprising that the effects of the LYP-R620W are complex, even within the CD4+ T cell compartment. Furthermore, there is a need for understanding the role of LYP in B and T cell central tolerance. Indeed, it is likely that the phenotypes observed in this study from T cell LYP variant expression would have additional implications in vivo and that cell-specific downstream effects of the PTPN22 SNP (rs2476601) may work in concert for altered effector function. Hence, there is a need moving forward for additional investigation of SNP-mediated in vitro versus in vivo effects. We plan to address these questions through ongoing efforts to build isogenic cellular systems to model multiple cell-cell interactions using gene-editing technologies (76). It is clear that rs2476601 contributes susceptibility for a number of autoimmune diseases including T1D (3, 4), likely related to functional consequences within the T cell compartment. We demonstrated, in this study, that the risk variant imparts a hypomorphic phenotype marked by impaired ability to downregulate T cell activation and effector function, supporting the notion that PTPN22 represents an important target for therapies aimed at preventing or reversing autoimmune disorders, including T1D.
This work was supported by grants from the JDRF (2-2012-280 and 2-PDF-2016-207-A-N), the Foundation for the National Institutes of Health (P01 AI042288, R01 DK106191, and T32 DK108736), and the National Institutes of Health Human Islet Research Network (UC4 DK104194 and UG3 DK122638). The authors declare that no conflicts of interest exist pertaining to the contents of this manuscript.
D.J.P. researched/analyzed the data and wrote the manuscript; P.S.L., L.D.P., L.Z., and Z.H. researched the data and reviewed/edited the manuscript; C.E.M., C.H.W., and M.A.A. contributed to discussion and reviewed/edited the manuscript; and T.M.B. conceived of the study and reviewed/edited the manuscript.
The online version of this article contains supplemental material.
Abbreviations used in this article
coding DNA sequence
C-terminal Src kinase
mean fluorescence intensity
conventional T cell
type 1 diabetes
regulatory T cell
Todd M. Brusko is the guarantor of this work and as such, assumes full responsibility for the ethical acquisition and reporting of data. The other authors have no financial conflicts of interest.