Research ArticleAUTOIMMUNITY

Suppression by human FOXP3+ regulatory T cells requires FOXP3-TIP60 interactions

See allHide authors and affiliations

Science Immunology  16 Jun 2017:
Vol. 2, Issue 12, eaai9297
DOI: 10.1126/sciimmunol.aai9297

TIPing the balance of autoimmunity

Individuals with mutations in Foxp3 develop an autoimmune syndrome called immunodysregulation polyendocrinopathy enteropathy X-linked syndrome (IPEX). However, the severity of IPEX varies depending on the specific Foxp3 mutation. Now, Bin Dhuban et al. find that the most common IPEX mutation, p.A384T, disrupts FOXP3 binding to the histone acetyltransferase TIP60 (KAT5), resulting in abrogated regulatory T (Treg) cell suppressive capacity, but maintained repression of proliferation and inflammatory cytokine production. Allosteric modifiers that help stabilize TIP60-FOXP3 interactions by inhibiting the autoacetylation of TIP60 molecules correct this disruption. These data suggest that targeting this interaction may be a therapeutic avenue in treating IPEX and other autoimmune and inflammatory diseases.


CD4+FOXP3+ regulatory T (Treg) cells are critical mediators of immune tolerance, and their deficiency owing to FOXP3 mutations in immunodysregulation polyendocrinopathy enteropathy X-linked syndrome (IPEX) patients results in severe autoimmunity. Different FOXP3 mutations result in a wide range of disease severity, reflecting the relative importance of the affected residues in the integrity of the FOXP3 protein and its various molecular interactions. We characterized the cellular and molecular impact of the most common IPEX mutation, p.A384T, on patient-derived Treg cells. We found that the p.A384T mutation abrogated the suppressive capacity of Treg cells while preserving FOXP3’s ability to repress inflammatory cytokine production. This selective functional impairment is partly due to a specific disruption of FOXP3A384T binding to the histone acetyltransferase Tat-interacting protein 60 (TIP60) (KAT5) and can be corrected using allosteric modifiers that enhance FOXP3-TIP60 interaction. These findings reveal the functional impact of TIP60 in FOXP3-driven Treg biology and provide a potential target for therapeutic manipulation of Treg activity.


FOXP3+ regulatory T (Treg) cells are a CD4+ T cell subset that plays an essential role in the maintenance of immunological tolerance to self and innocuous foreign antigens. Congenital or acquired Treg deficiency in several animal models precipitates autoimmune conditions that can be ameliorated with Treg infusion (1). In humans, Treg deficiency due to FOXP3 mutations results in a severe multi-organ autoimmune condition known as the immunodysregulation polyendocrinopathy enteropathy X-linked syndrome (IPEX) (2, 3). FOXP3, a forkhead (FKH) family transcription factor, plays an indispensable role in the development and regulatory function of Treg cells by orchestrating the transcriptional program responsible for the various aspects of Treg development. FOXP3 acts both as a transcriptional activator of Treg-relevant genes, such as IL2RA and CTLA4, and as a repressor of inflammation-driven genes, such as IFNG and IL-2 (4). Although FOXP3 homomeric dimers can mediate transcriptional regulation through direct binding to DNA, FOXP3 has also been shown to interact with several molecular partners forming FOXP3/protein complexes that are required for certain regulatory processes (5). The activity of the FOXP3 protein in Treg cells is tightly regulated through posttranslational modifications such as acetylation, phosphorylation, and ubiquitination [reviewed in (6)].

Several studies have demonstrated that acetylation of FOXP3 plays a critical role in the stability of the FOXP3 protein and in regulating its ability to access DNA (79). FOXP3 acetylation is mediated by two distinct histone acetyltransferase enzymes (HATs), namely, p300 and Tat-interacting protein 60 (TIP60). The acetylation of FOXP3 by p300 has been shown to prevent its polyubiquitination and proteasomal degradation, thus increasing the stability of FOXP3 (9). TIP60 has also been shown to interact with, and acetylate, FOXP3 (7, 9, 10). In a recent study, Xiao et al. (11) compared the importance of the two HATs in Treg cells by using conditional knockout mouse models lacking p300 or TIP60 in their Treg cells. In this study, the lack of TIP60 in Foxp3+ Treg cells resulted in severe weight loss, dermatitis, splenomegaly, and early death. Disease severity and time course in these Tip60fl/fl Foxp3YFP-Cre mice closely resemble those in the scurfy mouse model where Foxp3 mutations result in multi-organ autoimmunity leading to death 16 to 25 days after birth (11, 12). On the other hand, mice with p300-deficient Treg cells (p300fl/fl Foxp3YFP-Cre) developed normally until 8 weeks of age (11). Moreover, deficiency in TIP60, but not in p300, resulted in a substantial reduction in the frequency of Treg cells in the periphery, suggesting that TIP60 plays an indispensable role in the development and function of Treg cells (11). The role of TIP60 in human Treg cells has not been investigated.

To date, more than 60 different FOXP3 mutations have been reported in IPEX patients and are mainly found in the FKH domain followed by the N-terminal proline-rich domain (3). Although most FOXP3 mutations result in devastating autoimmunity and early death, several mutations have been associated with milder forms of IPEX (3). This clinical heterogeneity suggests that different FOXP3 mutations have distinct effects on the cellular functions of Treg cells, likely reflecting the relative importance of the affected residues in the various interactions of FOXP3. Thus, studying the molecular and cellular impact of various IPEX-causing mutations in primary Treg cells has the potential of highlighting critical pathways that are required for specific aspects of FOXP3 function in human Treg cells.

We have previously demonstrated that a missense mutation (A384T) in the FKH domain of FOXP3 leads to a diminished suppressive function of Treg cells in a severe case of IPEX (13). Here, we extend our analysis of the cellular and molecular impact of the p.A384T mutation of FOXP3 to identify specific protein interactions that are critical for FOXP3 to enable the development of human Treg function. We show that the FOXP3A384T mutation specifically impairs the suppressive function of Treg cells while maintaining several aspects of FOXP3 transcriptional functions including its ability to repress inflammatory cytokines. We further show that the impaired suppressive function of FOXP3A384T Treg cells is associated with a specific disruption of FOXP3 interaction with TIP60 and that enhancing FOXP3-TIP60 restores the ability of FOXP3A384T to drive the development of suppressive function. Our findings highlight how targeting FOXP3-TIP60 interaction can provide a potential avenue for therapeutic modulation of Treg activity in IPEX and other autoimmune and inflammatory diseases.


The p.A384T FKH mutation preserves many aspects of the Treg phenotype

The impact of various mutations in the FOXP3 gene on the FOXP3 protein expression is variable. Although some mutations abrogate protein expression, other mutations can allow normal or near-normal protein expression (3). Although lack of FOXP3 expression is associated with severe IPEX phenotype, the maintenance of FOXP3 expression does not necessarily predict milder disease. Here, we assessed the impact of the p.A384T mutation on FOXP3 expression levels in two unrelated IPEX patients bearing the same mutation. In both cases, we detected a population of CD4+FOXP3+ cells that is similar in frequency to that seen in age-matched healthy controls (Fig. 1A). However, the levels of FOXP3 protein per FOXP3+ cell [FOXP3 median fluorescence intensity (MFI)] are reduced in both IPEX patients compared with age-matched healthy controls (Fig. 1A). Hence, the p.A384T mutation does not substantially affect the frequency of FOXP3+ T cells, nor does it completely abrogate FOXP3 expression.

Fig. 1 Diminished suppressive potency of primary Treg clones from IPEXA384T patients.

(A) FOXP3 and CD25 expression in CD4+ cells of IPEXA384T patients and age- and sex-matched controls ex vivo. The FOXP3 MFI levels of CD25+FOXP3+ cells were determined in independent assessments and were normalized to the corresponding healthy controls. (B to E) Pools of single cell–derived clones of FOXP3+ and FOXP3 CD4+ T cells were generated from IPEX patients (shown as IPEX 1 and 2) and their corresponding controls (Control 1 and 2). (B) FOXP3 and CD25 expression at the harvest of the clones. (C) Cytokine production assessed by flow cytometry. (D) Proliferation of individual T cell clones in the absence of exogenous rhIL-2. cpm, counts per minute. (E) Suppression of proliferation of allogeneic CD4+CD25 T cells. Clone sample sizes: IPEX 1: FOXP3+ = 5, FOXP3 = 14; Control 1: FOXP3+ = 17, FOXP3 = 19; IPEX 2: FOXP3+ = 110, FOXP3 = 10, Control 2: FOXP3+ = 70, FOXP3 = 41. Data are presented as individual clones with line at median ± interquartile range. All data sets were compared with one-way nonparametric ANOVA (P < 0.0001) and followed by Dunn’s posttest.

We next assessed the impact of the p.A384T mutation on the development of Treg cells. However, this assessment could not be performed directly ex vivo with conventional Treg markers, such as CD25High and CD127Low, owing to the inflammatory conditions of the IPEX samples, which induces the expression of Treg-specific markers, including FOXP3, on activated effector CD4+ T (Teff) cells (1, 14). Thus, a reliable distinction between bona fide Treg cells and activated Teff cells under inflammatory conditions is not possible, precluding the definitive assessment of Treg function ex vivo. To circumvent these issues, we examined the phenotype and function of Treg cells using a single-cell cloning approach, a strategy that has the potential to uncover Treg-intrinsic functional differences (15, 16). In this approach, a phenotypic and functional assessment of clones is performed at the resting phase at the end of a T cell receptor (TCR) activation cycle. This allows activated Teff cells to down-regulate CD25 and FOXP3 while bona fide Treg cells maintain high expression levels of both molecules (15, 16). We generated primary Treg and Teff clones from both IPEX patients and respective healthy controls and analyzed their general phenotype, cytokine production, and suppressive function. In accordance with our observations in peripheral blood mononuclear cells (PBMCs) directly ex vivo (Fig. 1A), FOXP3 expression in IPEX-derived Treg clones from both IPEX patients was significantly lower than that observed in control clones (Fig. 1B). Furthermore, IPEX-derived Treg clones expressed CD25 at reduced levels compared with healthy clones (Fig. 1B). To determine whether the reduced CD25 expression in IPEXA384T Treg cells reflects a quantitative defect in FOXP3, we calculated the linear correlation between the FOXP3 and CD25 MFIs in both the IPEX and control FOXP3+ clone pools. Although FOXP3 and CD25 levels were correlated in both clonal pools, we found that the slope of the linear regression for the healthy controls is two times greater than that of the IPEX (control slope = 2.245 ± 0.284; IPEX slope = 1.094 ± 0.1262; P < 0.0001) (fig. S1). These results suggest that the reduced levels of FOXP3A384T may alter its capacity to act as a transcriptional enhancer.

The p.A384T mutation maintains the repressor function of FOXP3 in Treg cells

FOXP3 potently inhibits inflammatory cytokine transcription and production in Treg cells (17, 18). We examined inflammatory cytokine production in T cells, relative to FOXP3 expression, in PBMCs from both patients and controls. In both IPEXA384T patients and control donors, we found that FOXP3-expressing cells are refractory to the ex vivo production of inflammatory cytokines such as of interleukin-2 (IL-2), interferon-γ (IFN-γ), or IL-17 (fig. S2). This repression of inflammatory cytokines was also observed in IPEX-derived FOXP3+ clones (Fig. 1C). Similar to healthy controls, cytokine secretion in IPEX clones was predominantly restricted to FOXP3 Teff cell clones (Fig. 1C). Furthermore, upon TCR activation in the absence of exogenous IL-2, we found that IPEX-derived FOXP3+ clones are as hypoproliferative as healthy clones, further confirming their inability to produce IL-2 (Fig. 1D). These results show that, despite the p.A384T mutation, FOXP3 maintains its capacity to repress proliferation and cytokine production within FOXP3+ cells. Together, these results show that the p.A384T mutation preserves many transcriptional repressor functions in IPEXA384T Treg cells.

The p.A384T mutation selectively impairs the suppressive function of Treg cells

We next assessed the impact of the p.A384T mutation on the suppressive function of IPEXA384T Treg cells. We measured the ability of individual FOXP3+ and FOXP3 clones from IPEX patients or healthy controls to inhibit the proliferation of allogeneic Teff responder cells isolated from healthy individuals. FOXP3+ clones derived from IPEXA384T patients were severely impaired in their capacity to suppress Teff cells compared with control healthy FOXP3+ clones (Fig. 1E). The overall suppression by FOXP3A384T clones is not different from that of FOXP3 clones, which exhibited no suppression in either patients or controls. Thus, the p.A384T mutation severely impedes the capacity of FOXP3 to orchestrate the Treg cell suppressive function in IPEX CD4+ T cells.

To further demonstrate a direct causal link between the p.A384T mutation and the observed defective Treg function, we examined the capacity of FOXP3A384T to drive the Treg cell phenotype and function upon ectopic expression in healthy conventional CD4+ Teff cells, as previously described (19). The p.A384T mutation was compared with the proximal R397W mutation, which abrogates the DNA binding capacity of the FKH domain (20, 21). Naïve CD4+CD45RA+CD25 Teff cells from healthy donors were transduced using a bicistronic lentiviral vector expressing both green fluorescent protein (GFP) and either wild-type (WT) or IPEX FOXP3 variants (A384T or R397W) (Fig. 2A). We detected high levels of FOXP3 in cells transduced with FOXP3A384T, like those obtained with FOXP3WT (Fig. 2B). Investigation of inflammatory cytokine production in FOXP3-transduced cells indicated that both FOXP3WT and FOXP3A384T, but not FOXP3R397W or an empty vector, strongly repress the production of IFN-γ, IL-2, and IL-4 in transduced cells (Fig. 2C). Thus, FOXP3A384T preserves the capacity of FOXP3 to repress the inflammatory cytokine profile of transduced cells. However, cells expressing either FOXP3A384T or FOXP3R397W were significantly impaired in their ability to suppress Teff cell proliferation compared with those expressing FOXP3WT (Fig. 2, D and E). Collectively, we show that FOXP3A384T can reprogram CD4+ T cells into phenotypically Treg-like cells that maintain strong transcriptional repression of inflammatory cytokines but are severely defective in suppressive function, consistent with our observations on ex vivo and in primary Treg clones.

Fig. 2 Impaired development of Treg cell function upon overexpression of FOXP3A384T in naïve CD4+CD25 T cells.

(A) Schematic representation of the constructs used in transduction. (B) Expression levels of FOXP3, CD25, and CD127 in transduced cells on day 28 post-transduction. (C) Cytokine production by transduced cells expressing WT or mutated FOXP3. (D and E) Suppressive potency of empty vector (EV)–, FOXP3WT-, FOXP3A384T-, and FOXP3R397W-transduced cells. Transduced, GFP+ cells were cocultured with proliferation dye–labeled CD4+CD25 Teff cells at a 1:4 ratio. Results are representative of 11 independent assessments on seven independent transduction experiments carried out on cells from four different donors. All data sets were compared with one-way nonparametric ANOVA followed by Dunn’s posttest.

FOXP3A384T partially establishes the FOXP3-driven genetic program of Treg cells

To identify FOXP3-regulated genes specifically required for Treg suppressive function, we assessed the capacity of FOXP3A384T to establish the canonical FOXP3 transcriptional signature in developing Treg cells. To this end, we compared the gene expression profiles of primary human CD4+ T cells transduced with empty vector or vectors encoding FOXP3A384T or FOXP3R397W, relative to control FOXP3WT. We first determined a FOXP3-regulated gene signature by comparing the empty vector and FOXP3WT-transduced cells (Fig. 3A). We identified 2306 genes that are either induced or repressed by FOXP3 expression and whose expression constitutes the canonical Treg signature. Principal component analysis (PCA) of these FOXP3-regulated genes indicates that, although this signature is mostly perturbed in FOXP3R397W-transduced cells, it is only partially perturbed in FOXP3A384T-transduced cells compared with that of FOXP3WT-transduced cells (Fig. 3B). Around 63% of the genes regulated by FOXP3WT are similarly induced or repressed by FOXP3A384T (Fig. 3C). In contrast, FOXP3R397W only induces about 10% of the Treg signature. However, we identified 223 genes that are part of the FOXP3-regulated gene signature but are differentially expressed (P < 0.05) in FOXP3A384T-expressing cells relative to FOXP3WT-expressing cells (table S1). Many of these genes, namely, NT5E, FOXO1, S1PR1, LGMN, KLF, DUSP5, and DUSP6, have been described to have a potential role in Treg development and function (Fig. 3D) (2228).

Fig. 3 Partial alteration of the Treg gene signature by the p.A384T mutation.

FOXP3WT-, FOXP3A384T-, FOXP3R397W-, or empty vector–transduced cells were activated with PHA (0.5 μg/ml) for 36 hours and subjected to microarray analysis to compare their gene expression profiles. (A) A volcano plot showing the identification of genes that are modulated by the expression of FOXP3WT relative to empty vector–transduced cells [log2 (fold change) < −0.5 or > 0.5; P < 0.05]. (B) PCA of FOXP3-regulated genes [identified in (A)] in the four indicated conditions. Each circle represents one biological replicate. Component 1 (x axis) explains 66% while component 2 (y axis) explains 10% of the variance. (C) Heat map showing the expression levels of the FOXP3-regulated genes identified in (A). % Similarity to WT was estimated on the basis of the number of FOXP3-modulated genes that are differentially expressed in each condition relative to WT FOXP3WT-transduced cells. Biological replicates for each condition are shown side by side. (D) Selected FOXP3-modulated genes that are significantly differentially expressed (P < 0.05) in FOXP3WT-tranduced cells versus FOXP3A384T-transduced cells. (E) DAVID pathway analysis of 223 FOXP3-modulated genes that are differentially expressed (P < 0.05) in FOXP3WT-transduced cells versus FOXP3A384T-transduced cells.

To highlight the cellular pathways that are most affected by the p.A384T mutation, we performed a Database for Annotation, Visualization and Integrated Discovery (DAVID) pathway analysis on the FOXP3-regulated genes that are altered in FOXP3A384T-expressing cells. We found that this mutation significantly alters the expression of genes involved in multiple pathways affecting lymphocyte differentiation, activation, proliferation, and migration, as well as cytokine production (Fig. 3E), suggesting that this mutation interferes with a potentially critical molecular interaction of the FOXP3 protein that is required for multiple downstream processes.

The p.A384T mutation impairs the interaction of FOXP3 with the histone acetyltransferase, TIP60

Part of the transcriptional regulation mediated by FOXP3 requires its binding to a number of molecular partners such as FOXP1, RORα, NFAT, HDAC7, HDAC9, and TIP60 (10). We assessed the ability of FOXP3A384T to interact with several of these molecular partners. Although it maintains a normal capacity to bind FOXP1 relative to FOXP3WT (Fig. 4A and fig. S3), FOXP3A384T displays a greatly diminished ability to bind TIP60 (KAT5) (Fig. 4B and fig. S3), a histone acetyltransferase that binds to FOXP3 at the N terminus and leads to FOXP3 acetylation, a prerequisite to its functional activation and stabilization (11). The importance of TIP60 in Treg development has been previously demonstrated in a knockout mouse model where Treg-specific TIP60 deficiency impaired Treg development and caused fatal autoimmune manifestations akin to those observed in the scurfy mouse (11). TIP60 also interacts with another histone acetyltransferase, p300, which synergizes with TIP60 to acetylate FOXP3 (11). In addition, p300 promotes the autoacetylation of TIP60, leading to substrate switching and the dissociation of TIP60 molecules from the TIP60-p300-FOXP3 complex (11). Curiously, we observed a markedly enhanced interaction between p300 and FOXP3A384T protein (Fig. 4C and fig. S3), suggesting that p300 binding may be increased to compensate for the diminished TIP60 binding to FOXP3A384T. These observations indicate that the impaired binding of FOXP3A384T to TIP60 can interfere with certain functional properties of FOXP3 and may underlie the diminished suppressive function seen in human Treg cells with the p.A384T mutation of FOXP3.

Fig. 4 Impaired interaction of FOXP3A384T with TIP60.

(A to C) 293T cells were transfected with EV, FOXP3WT, FOXP3A384T, or FOXP3R397W and FLAG-FOXP1, FLAG-TIP60, or FLAG-p300. Twenty-four hours after transfection, cell lysates were collected and immunoprecipitated, followed by blotting with anti-FLAG or anti-HA HRP. The co-immunoprecipitation of FOXP3 mutants with FOXP1 (A), TIP60 (B), or p300 (C) in representative experiments of at least three separate experiments is shown. IP, immunoprecipitated protein.

Enhancement of FOXP3-TIP60 interaction restores FOXP3A384T Treg cell suppressive function

We have developed TIP60-binding small molecules that act on an allosteric cavity of TIP60. These allosteric modifiers help stabilize TIP60-FOXP3 interactions by inhibiting the autoacetylation of TIP60 molecules, thus delaying the release of TIP60 from the TIP60-p300-FOXP3 complex (29). Using one such TIP60 allosteric modifier, SGF003 (fig. S4), we observed an enhanced interaction between TIP60 and FOXP3WT as early as 30 min after addition of the modifier to 293T cells (Fig. 5A), or to Jurkat T cells (fig. S5), cotransfected with TIP60 and FOXP3. This modifier-mediated enhancement of FOXP3 binding was specific to TIP60, because it did not affect the interaction of FOXP3 with either p300 or FOXP1 (Fig. 5B and fig. S6).

Fig. 5 Enhancement of TIP60-FOXP3 interaction by allosteric modification of TIP60.

293T cells were transfected with HA-FOXP3WT and (A) FLAG-TIP60, or (B) FLAG-FOXP1 (top) or FLAG-p300 (bottom) in the presence or absence of SGF003 (8 μg/ml). (C) 293T cells were transfected with HA-FOXP3A384T and FLAG-TIP60 in the presence or absence of SGF (8 μg/ml). Twenty-four hours post-transfection, cells were washed with PBS, and cell lysates were prepared for immunoprecipitation and Western blot analysis. The effects of SGF treatment on the interaction of FOXP3 with the flagged proteins are shown. Representative blots from one of at least three separate experiments are also shown.

We next asked whether this allosteric modifier could rescue the impaired interaction between TIP60 and FOXP3A384T. SGF003 treatment led to a rapid and marked increase in TIP60 binding to FOXP3A384T (Fig. 5C and fig. S7). Similar improvements in TIP60 interaction with both FOXP3WT and FOXP3A384T were also achieved using a chemically related compound (B7A; fig. S8) (29).

We next examined the effect of these TIP60 modifiers on the suppressive function of murine and human Treg cells in a Treg/Teff coculture suppression assay. We observed that neither SGF003 nor B7A affected the proliferative capacity or the production of inflammatory cytokines, such as IFN-γ, in murine and human responder Teff cells cultured alone (Fig. 6A and fig. S9). However, both SGF (Fig. 6B) and B7A (fig. S9) markedly enhanced the suppressive capacity of murine and human Treg cells, especially at lower Treg/Teff ratios where more than twofold enhancement in Treg suppressive function was achieved (Fig. 6B). Moreover, given the reciprocal relationship between Foxp3 and ROR-γt in the development of Treg and TH17 (T helper cell 17) cells, respectively, we assessed whether the Treg-promoting action of TIP60 modifiers is partly due to stabilizing Foxp3 expression at the expense of ROR-γt expression and IL-17 production. However, no differences in the polarization of murine TH17 cells were observed upon SGF003 treatment (fig. S10). Collectively, these results indicate that the enhancement of Treg suppressive function by TIP60 allosteric modifiers is achieved through direct action on the suppressive program of Treg cells, rather than through the inhibition of the activation, differentiation, and proliferative responses of effector T cell subsets.

Fig. 6 Enhancement of Treg suppressive function by allosteric modification of TIP60.

The suppressive potency of murine Treg (CD4+Foxp3+ splenocytes) or human Treg (CD4+CD25HighCD127Low) was assessed in the presence or absence of SGF003. SGF003 was added at the time of activation. Suppression was measured as the reduction in proliferation of responder Teff cells (murine CD4+FOXP3; human CD4+CD25CD127High). (A) The effect of SGF003 treatment on the proliferative capacity and IFN-γ production of Teff cells cultured in the absence of Treg cells. (B) The suppressive potency of murine and human Treg cells in the presence of SGF003 at various Treg/Teff ratios. The percentages of improvement in Treg suppressive capacity obtained by comparing the suppressive potency of untreated versus SGF003 (8 μg/ml)–treated Treg cells are shown. (C) Effect of SGF003 on the suppressive potency of FOXP3WT-, FOXP3A384T-, or FOXP3R397W-transduced cells cocultured with CD4+CD25CD127High Teff cells at a 1:4 Treg/Teff ratio. The results shown in this figure are representative of two to three independent experiments performed in triplicate. Statistical analyses were performed using t test (A) and one-way ANOVA (B and C).

Considering the impaired interaction of TIP60 with FOXP3A384T, we hypothesized that SGF003-mediated enhancement of this interaction would rescue the suppressive function of FOXP3A384T-engineered Treg cells. Treatment of FOXP3A384T-transduced cells with SGF003 resulted in a significant restoration of their suppressive function to levels comparable to those observed in untreated FOXP3WT-transduced cells (Fig. 6C). In agreement with our observations ex vivo, SGF003 also enhanced the suppression exerted by FOXP3WT-transduced cells albeit to a lesser extent than it did for FOXP3A384T-transduced cells (Fig. 6C). These findings highlight an important role for the FOXP3-TIP60 interaction in the suppressive function of Treg cells and suggest that this interaction may be required for driving a genetic program that is critical for suppressive function by Treg cells.

Allosteric modification of TIP60 improves the FOXP3-mediated transcriptional program in FOXP3A384T Treg cells

To gain insight into specific genes and pathways that may be downstream of the TIP60-FOXP3 interaction and likely involved in the development of Treg suppressive function, we evaluated the effects of SGF003 treatment on the gene expression profile in FOXP3A384T-engineered Treg cells. To this end, cells transduced with an empty vector, FOXP3WT, or FOXP3A384T were activated in the presence or absence of SGF003, and their gene expression profiles were examined. In line with its enhancement of FOXP3-TIP60 binding and its positive impact on Treg suppressive function, we found that SGF003 induces a marked shift in the expression levels of FOXP3-regulated genes in cells transduced with FOXP3A384T. Specifically, SGF003 renders the gene expression profile of FOXP3A384T-tranduced cells more similar to that of FOXP3WT-transduced cells, as indicated by PCA (Fig. 7A). Focusing on the 223 FOXP3-regulated genes that are significantly altered in untreated FOXP3A384T-transduced cells compared with FOXP3WT-transduced cells (table S1), SGF003 treatment modulates the expression of the majority of these genes (128 of 223) to levels that are comparable to untreated FOXP3WT-transduced cells (Fig. 7, B and C, and table S2). This SGF003-responsive gene set includes several genes that have previously been described to play roles in Treg development and function and include LGMN, FOXO1, DUSP6, NT5E, and CEACAM1 (22, 23, 25, 28, 30), as well as several other genes whose role in Treg cells remains unknown. Further work is needed to elucidate the potential roles of these proteins in Treg suppressive function and the mechanisms of their regulation by the TIP60-FOXP3 complex.

Fig. 7 Improvement in the transcriptional activity of FOXP3A384T by enhancement of TIP60-FOXP3 interaction.

FOXP3WT-, FOXP3A384T-, or empty vector–transduced cells were activated with PHA (0.5 μg/ml) for 36 hours in the presence or absence of SGF003 (8 μg/ml), followed by genome-wide microarray analysis. (A) PCA of FOXP3-regulated genes (identified in Fig. 3A) in the four indicated conditions. Component 1 (x axis) explains 65% whereas component 2 (y axis) explains 15% of the variance. (B) Heat map showing the impact of SGF treatment on the 223 FOXP3-regulated genes that are differentially expressed (P < 0.05) in untreated FOXP3WT-transduced cells versus FOXP3A384T-transduced cells. (C) The impact of SGF treatment on the expression level of 30 Treg-associated genes (shown in Fig. 3C) that are differentially expressed in FOXP3A384T-transduced cells.

Therapeutic restoration of FOXP3-TIP60 interaction promotes Treg cell function and disease protection in mouse models of colitis and collagen-induced arthritis

To assess the therapeutic potential of TIP60 modifiers in enhancing Treg function and reducing inflammation in vivo, we first examined the benefit of allosteric TIP60 targeting in controlling inflammation in a dextran sodium sulfate (DSS)–induced colitis model. Administration of the TIP60 allosteric modifier (B7A) rescued mice from established colitis as evidenced by weight regain (Fig. 8A) and preservation of colon length (Fig. 8B). Moreover, compared with the transmural inflammation, goblet cell destruction, and extensive thickening of colons in dimethyl sulfoxide (DMSO)–treated mice (mean histologic score of 3.2 ± 1.3), B7A therapy preserved colonic histology (mean histologic score of 1.6 ± 1.0, P < 0.01), with minimal inflammation, edema, or goblet cell injury (Fig. 8C). The beneficial effects of the treatment in this inflammatory condition are apparent despite treating after the onset of the pathology and are associated with a significant increase in the frequency of Foxp3+ Treg cells (Fig. 8D).

Fig. 8 Therapeutic effects of TIP60 allosteric modification in DSS-induced colitis.

Colitis was induced by adding 4% DSS to pH-balanced tap water of study mice. Disease progression was assessed by daily monitoring of body weight, stool consistency, and fecal blood. B7A (4 mg kg−1 day−1, 7 days) or DMSO control was injected intraperitoneally 1 week after DSS administration. The effects of B7A treatment on (A) body weight, (B) colon length, and (C) gut histology (hematoxylin and eosin–stained paraffin sections) are shown. (D) Frequency of FOXP3+ Treg cells in the spleen and lymph nodes of B7A-treated compared with DMSO-treated mice. n = 10 to 12 mice per group.

These modifiers specifically affect Foxp3+ Treg cells and their suppressive function without affecting the activation, proliferation, or cytokine production of ex vivo Teff cells in vitro (Fig. 6A and fig. S9), under polarizing TH17 conditions in vitro (fig. S10A), or in mesenteric sites in vivo (fig. S10B). These data demonstrate that TIP60 modifiers protect from disease progression and pathology by enhancing Foxp3+ Treg cell function without influencing the development of pathogenic Teff cells involved in gut pathology.

To confirm these therapeutic effects in an antigen-driven setting, we used a well-established collagen-induced arthritis (CIA) model where CIA was induced in DBA mice through immunization with bovine type ll collagen. Treatment with B7A after disease onset led to a markedly improved arthritis score (fig. S11). These results further highlight the potential therapeutic application of TIP60 allosteric modifiers for the enhancement of Treg-mediated immune regulation in inflammatory diseases.


The requirement for FOXP3+ Treg cells in the maintenance of tolerance in the periphery in humans is compellingly illustrated by the neonatal onset and the severity of IPEX. However, despite major efforts, relatively little is known about the core mechanisms via which FOXP3 orchestrates Treg cell development and suppressive function. In this study, we took advantage of a naturally occurring IPEX-causing mutation to determine critical interactions of FOXP3 that are required for the proper development of human Treg cells. Our results show loss of suppressive function by Treg cells bearing the A384T mutation in the FOXP3 gene. This functional defect is associated with a disruption in the molecular interaction of FOXP3 with the histone acetyltransferase TIP60 and can be corrected by small molecule–mediated enhancement of FOXP3-TIP60 interaction and consequential bolstering of Treg suppressive function.

FOXP3 interacts with TIP60 and p300 to form a TIP60-p300-FOXP3 complex, in which FOXP3 is acetylated, resulting in enhanced FOXP3-mediated transcriptional activity (11). This tripartite complex lasts until TIP60 is autoacetylated, resulting in the dissociation of the complex (11). The structural mechanisms through which the p.A384T mutation interferes with the TIP60-FOXP3 interaction remains to be investigated. Although it is likely that this point mutation impairs this specific interaction through steric or electrostatic effects, it is also possible that the p.A384T mutation impairs FOXP3-TIP60 interaction through promotion of rapid autoacetylation of TIP60, leading to unstable TIP60-FOXP3 complex formation.

The acetylation of the FOXP3 protein mediated by TIP60-p300 cooperation is thought to play an important role in the stability of the FOXP3 protein by preventing its ubiquitination and degradation (9, 11, 31). Xiao et al. have shown that mice with a Treg-specific deletion of p300 (p300fl/fl Foxp3YFP-Cre) develop a significantly milder phenotype compared with those with a Treg-specific TIP60-deficiency (Tip60fl/fl Foxp3YFP-Cre) despite a significant reduction in Foxp3 acetylation in both models (11, 31). Mice with TIP60-deficient Treg cells succumb to severe autoimmunity very early in life, whereas mice with p300-deficient Treg cells develop normally until 8 weeks of life and only show moderate autoimmune manifestations (11, 31). Thus, the severe phenotype associated with TIP60-deficient Treg cells in mice and the complete abrogation of suppressive function in FOXP3A384T cell observed in the present study suggest that TIP60 plays a critical role in the development of Treg function through mechanisms other than acetylation of FOXP3. One possibility is that TIP60 is recruited by FOXP3 to facilitate the activation of target genes that are critical for the development of the suppressive function of Treg cells. Another possibility is that the molecular partnership between TIP60 and FOXP3 involves noncatalytic functions that are essential to Treg activity. Previous studies have demonstrated examples of such nonenzymatic roles of several histone deacetylases. For instance, a splice variant of HDAC9 that lacks the catalytic domain was found to mediate transcriptional regulation in cardiac myocytes through the recruitment of transcriptional co-repressors (3234). Similarly, HDAC3 and HDAC5 have been demonstrated to exhibit nonenzymatic functions by participating in the scaffolding of transcriptional regulatory complexes (35, 36). Future studies may reveal novel nonenzymatic roles of TIP60 in Treg biology.

The FKH domain of FOXP3 is crucial for both the nuclear localization and the repressor functions of FOXP3 (20). However, unlike the R397W mutation, the p.A384T mutation does not affect any of the residues that are directly involved in any of the known molecular functions or overall structure of the FKH domain (21). Moreover, previous reporter-gene studies have indicated that the p.A384T mutation, unlike the R397W mutation, does not abrogate the DNA binding capacity of the FOXP3 protein to the il2 promoter. Other studies have suggested that IL-2 transcription is repressed in the presence of p.A384T FOXP3 (20, 37), an observation fully consistent with our current results showing suppressed IL-2 transcription or secretion in primary and FOXP3-transduced CD4+ T cells.

The association of FOXP3 with TIP60 is thought to be mediated by the N-terminal domain of FOXP3 (7). However, in this study, we show that this interaction is disrupted by the FKH-located p.A384T mutation in the opposite end of the protein. This could be explained by previous studies where FOXP3 has been shown to assemble as an antiparallel homodimer, a configuration that is required for its function (38). The structural characteristics of this homodimeric FOXP3 render its C-terminal DNA binding domain spatially adjacent to the N-terminal repressor domain. Thus, specific characteristics of the FOXP3 homodimer may be altered by the p.A384T mutation, indirectly affecting FOXP3 functional partnerships involving the N-terminal domain while leaving other centrally located interactions unaffected; this is the case for FOXP3-FOXP1 binding that involves the leucine zipper and zinc finger domains of FOXP3. Such intramolecular interactions between domains and mutations have not been previously described for FOXP3 and may contribute to the complex phenotypes observed in other hypomorphic mutations of FOXP3 (3, 39).

Although the canonical Treg phenotype and suppression of inflammatory cytokine production is maintained, the p.A384T mutation differentially alters a portion of the FOXP3 transcriptional signature in T cells and is unable to establish Treg suppressive activity. These observations suggest that FOXP3 regulates Treg phenotype and function through distinct molecular mechanisms, allowing cells to develop a characteristic Treg phenotypic profile independently of suppressive function. Previous studies in mice where the Foxp3 gene was mutated to partially disrupt the Foxp3 protein have shown that Foxp3 disruption allows the development of Treg cells that exhibit the hallmark Treg phenotype and cytokine repression but lack suppressive capacity (40, 41). Although these studies have concluded that Foxp3 is not required for the development of Treg phenotype, the Foxp3 protein in these models was not completely abrogated. It is likely that the residual Foxp3 protein, particularly from the N-terminal domain, was sufficient to regulate the transcription of genes responsible for Treg phenotype and cytokine repression but was too excessively disrupted to orchestrate the genetic program responsible for the development of suppressive function in these models. This broad dichotomy of FOXP3 function revealed in our study highlights the benefit of carefully characterizing how different IPEX-derived FOXP3 mutations affect Treg phenotype and function. Insights into the molecular mechanisms underlying the role of FOXP3 in Treg cells will have a critical impact on the understanding of Treg cells in humans and their role in autoimmune diseases.

One limitation of the current study is that the defective interaction of FOXP3A384T with TIP60 has yet to be recapitulated in primary Treg cells from IPEX patients. The extreme rarity of IPEX cases coupled with the small blood sample sizes that can be obtained from those patients preclude examination of protein-protein interactions using the standard co-immunoprecipitation approaches. Given the sequence homology of the FKH domain in mouse and human FOXP3, it will be interesting to investigate the role of TIP60 in Treg cells in murine models carrying the A384T mutation.

In summary, our studies identify FOXP3-TIP60 interaction as a critical requirement for the instruction of suppressive function in human Treg cells. It further uncovers a mechanism by which a small-molecule modifier of TIP60 can increase human Treg cell function without directly impairing T effector cell responses, providing potential avenues for targeted therapeutic manipulation of Treg cells under various disease conditions.


Human samples

The two IPEX patients examined both carried the p.A384T missense mutation in the FKH domain of FOXP3, as determined by genetic sequencing. Patient #1 presented in the first few weeks of life with severe multi-organ failure, rash, enteritis, and endocrine abnormalities, leading to death at the age of 2 months (13). Patient #2 suffered from diabetes, enteritis, atopic dermatitis, and neutropenia (42), and underwent bone marrow transplantation at age 19. Informed consents were obtained, and the study was performed in agreement with the ethical review board of McGill University. In both cases, the p.A384T mutation was identified after sequencing of the FOXP3 gene (Immunology Diagnostic Laboratory, Seattle Children’s Hospital, Seattle, WA, USA).

Multiparametric flow cytometry and cell sorting

For all flow cytometric analyses, samples were labeled with a fixable viability dye to facilitate live-cell gating (eBioscience). The anti-human FOXP3 antibody clone used was 236A/E7 (eBioscience); all other antibodies were obtained from BD Biosciences. Cell sorting was performed using a FACSAriaIIu cell sorter (BD Biosciences). Samples were acquired on a FACSCanto or LSRII-Fortessa analyzer (BD Biosciences), and analysis was performed using FlowJo software (TreeStar).

Generation of CD4+T cell clones

Primary clones were obtained by short-term expansion and subjected to phenotypic and functional assays as previously described (15, 16). Briefly, CD4+CD25High (top 1%) and CD4+CD25Neg T cells were isolated from PBMC and dispensed as single cells by fluorescence-activated cell sorting (FACS).

The cells were activated in the presence of allogeneic irradiated feeders, anti-CD3, and rhIL-2. IL-2 medium was replenished every 2 to 4 days, and cells were restimulated after 11 days. After 22 days of culture, each clone was subjected to micro-sized functional and phenotypic assays.

FOXP3-GFP expression vectors

cDNAs encoding WT or mutated FOXP3-GFP fusions [provided by T. Torgerson as per Lopes et al. (20)] were subcloned in a modified pLVX-Tight-Puro Vector (Clontech), whose promoter was replaced with the EF1α promoter, and the pPGK-Puro sequence was replaced with an IRES (internal ribosomal entry site)–GFP sequence. Appropriate cDNAs were inserted upstream of IRES-GFP. Correlation of GFP and FOXP3 expression was confirmed for each vector after primary T cell transduction.

Lentiviral transduction of primary T cells

Lentiviral pseudoparticles were obtained after four-plasmid transfection of 293T cells using Lipofectamine (Invitrogen), as previously described (19). Viral supernatants were collected 48 and 72 hours after transfection and immediately used for transduction of preactivated primary CD45RA+CD25CD4+ T cells. Infection rates ranged between 10 and 25%. Seven days after transduction, each cell line was FACS-sorted for GFP expression. IL-2 was replenished every 3 days, and cells were expanded for a total of 28 days after initial activation to obtain sufficient amounts of cells for analysis. GFP expression was verified, and cells were re-sorted, if necessary, before analysis.

Intracellular cytokine staining

For measurement of cytokine production, cells were restimulated for 24 hours in the presence of soluble anti-CD3 (clone OKT3; 30 ng/ml) and rhIL-2 (200 IU/ml) and, in the case of clones or transduced cells, with allogeneic irradiated PBMCs (1:4 T cell/feeder ratio). The cultures were activated with phorbol 12-myristate 13-acetate (PMA) (25 ng/ml), ionomycin (1 μg/ml) (Sigma-Aldrich), and GolgiStop (1:1000) (BD Biosciences) for 4 hours; fixed and permeabilized (eBioscience permeabilization kit); and subjected to intracellular cytokine staining.

Suppression assays

All suppression assays performed in this study assessed the capacity of Treg cells to suppress the proliferative response of carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled, FACS-sorted, allogeneic CD4+CD25Neg Teff cells obtained from healthy donors. In all suppression assays in this study, Treg cells were cocultured at the indicated ratios with 104 Teff cells in the presence of 4 × 104 healthy violet proliferation dye 450 (VPD450)–labeled, irradiated PBMCs and soluble anti-human CD3 (30 ng/ml), as previously described (13, 15, 16). When assessing the suppressive potency of GFP+-transduced cells, the CFSE was substituted with the eFluor 610 Proliferation Dye (eBioscience), and the assay was set with 3 × 104 Teff cells per well and a 1:4 Treg clone/Teff cellular ratio in the presence of 1.2 × 105 irradiated PBMCs. Proliferation was determined for each coculture by FACS by measuring the proportion of cells that have diluted the proliferation dye after 96 hours of coculture compared with Teff cells cultured alone.

TH17 differentiation

CD4+Foxp3 splenocytes were FACS-isolated from 7.5-week-old male C57BL/6 mice and activated with soluble α-CD3 (1 μg/ml) in the presence of irradiated antigen-presenting cells. Cells were stimulated with transforming growth factor–β (1 ng/ml) and IL-6 (50 ng/ml) in the presence or absence of SGF003 (8 μg/ml). Cytokine secretion was assessed at 72 hours post-activation after 5 hours of stimulation with PMA (25 ng/ml), ionomycin (1 μg/ml) (Sigma-Aldrich), and GolgiStop (1:1000) (BD Biosciences).

Proliferation assays

Individual clones (2 × 103 cells) were stimulated in vitro in triplicate in the presence of TCR stimulation and IL-2, as previously described (15). Proliferation was assessed by pulsing the cultures with 3H-thymidine (0.5 μCi/ml) for the last 20 hours of the 5 days of culture. 3H-thymidine incorporation in individual wells was then measured using scintillation liquid.

Gene expression analysis

Naïve CD4+ cells were transduced with FOXP3WT, FOXP3A384T, FOXP3R397W, or an empty vector in two separate experiments. Transduced cells were sorted 14 days after transduction based on GFP expression and were restimulated with soluble anti-CD3 (30 ng/ml) and irradiated PBMCs (3×) for 14 more days. Cells were then activated with phytohemagglutinin (PHA) (0.5 μg/ml) in the presence or absence of SGF003 (8 μg/ml). RNA was isolated using the RNeasy Mini Kit from Qiagen as per the manufacturer’s instructions. Samples were run on a human Clariom S chip (Affymetrix), and the resulting raw expression data were extracted, annotated, Robust Multi-array Average–normalized, and background-adjusted using the pd.clariom.s.human, clariomshumantranscriptcluster.db, and oligo packages in R. A modified analysis of variance (ANOVA) for microarray analysis (eBayes function from the Limma package) was used to compare across all conditions for each gene. A FOXP3 signature was obtained by identifying genes that are significantly up-regulated [log2 (fold change) > 0.5, P < 0.05] or down-regulated [log2 (fold change) < −0.5, P < 0.05] in FOXP3WT-transduced cells compared with empty vector–transduced cells. Within the FOXP3 signature, the altered FOXP3 signature in FOXP3A384T-transduced cells was identified by comparing them to FOXP3WT-transduced cells (P < 0.05). Of the defective FOXP3 signature in FOXP3A384T-transduced cells, significantly improved (P < 0.05) gene expressions were identified in SGF003-treated FOXP3A384T-transduced cells compared with untreated FOXP3A384T-transduced cells. Visualizations (volcano plot and heat maps) were generated using the gplots package in R. PCAs were based on the FOXP3 signature genes and PCA plots were drawn using PC1 (principal component 1) and PC2, which accounted for >75% of the variance in gene expression. Gene ontology overrepresentation analysis of selected gene lists was performed using DAVID. Functional annotation clustering was performed against the Biological Process Gene Ontology database (GOTERM_BP_FAT).

Cell culture and transfection

293T cells were grown in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum and antibiotics (1% penicillin/streptomycin; Invitrogen) at 37°C in a humidified incubator with 5% CO2 (v/v). Transient transfection was carried out when cells were grown to 80% confluence using a mixture of DNA and Fugene6 (Roche) according to the manufacturer’s instructions. Twenty-four hours after transfection, cells were washed twice with phosphate-buffered saline (PBS), and cell lysates were then prepared for immunoprecipitation and Western blot analysis.


Cells were lysed in modified radioimmunoprecipitation assay (RIPA) buffer [20 mM tris-Cl (pH 7.5), 2 mM EDTA, 420 mM NaCl, and 1% NP-40]. After centrifugation, the soluble fractions were collected and incubated with anti-hemagglutinin (HA) agarose (Sigma) at 4°C for 2 hours. The precipitates were then washed three times with modified RIPA buffer and boiled for 5 min in SDS loading buffer. Samples were analyzed by SDS–polyacrylamide gel electrophoresis, transferred to nitrocellulose membrane (Millipore), and probed with anti-FLAG horseradish peroxidase (HRP) (Sigma). Immune complexes were detected using Immobilon Western Chemiluminescent HRP Substrate (Millipore). Densitometry analysis of Western blots was performed using the ImageJ software. All immunoprecipitation experiments were performed at least three times with similar findings.

Induction of colitis in mice

Freshly prepared 4% DSS (w/v) (MP Biomedicals) was added daily for 5 days to pH-balanced tap water of study mice (12 mice per group). Colitis was assessed by the daily monitoring of body weight, stool consistency, and fecal blood, as previously described (43). The effects of the TIP60 modifier B7A were assessed by administration of B7A (4 mg kg−1 day−1, 7 days, intraperitoneally) or DMSO control, 1 week after beginning DSS administration.

Statistical analysis

Unless otherwise indicated, statistical analysis was performed using Prism 5.04 software from GraphPad Software (La Jolla, CA). For all distributions, the D’Agostino and Pearson omnibus test was used to assess normality. Parametric or nonparametric tests were applied consequently. For multiple comparisons, one-way ANOVA was performed, and posttest P values are indicated. For single comparisons, P values were calculated by t test. Only for comparisons reaching significance (α = 0.05) is the P value shown.


Fig. S1. Reduced expression of FOXP3A384T alters its ability to induce CD25 expression.

Fig. S2. Cytokine repression is maintained by FOXP3A384T.

Fig. S3. Unmodified blots from Fig. 4.

Fig. S4. Model for the mechanism of action of TIP60 modifiers in the TIP60-p300-FOXP3 interaction in Treg cells.

Fig. S5. Enhanced FOXP3-TIP60 interaction in response to SGF003 treatment in Jurkat cell line.

Fig. S6. Unmodified blots from Fig. 5 (A and B).

Fig. S7. Unmodified blots from Fig. 5C.

Fig. S8. Enhancement of TIP60-FOXP3 interaction by B7A.

Fig. S9. Improvement of Treg suppressive capacity by B7A-mediated enhancement of TIP6-FOXP3 interaction.

Fig. S10. TH17 differentiation is not altered by SGF003 treatment.

Fig. S11. Treatment of CIA with B7A.

Table S1. Differentially expressed genes in FOXP3A384T-transduced cells.

Table S2. SGF treatment improves the expression of several FOXP3-regulated genes.

Raw data


Acknowledgments: We thank H. Mason, E. Yurchenko, M. Kornete, M. da Silva Martins, J. Grenier, and E. Sgouroudis for discussions and technical support. Funding: Financial support for this study came from Canadian Institutes of Health Research (CIHR) grant MOP67211 (C.A.P.), CIHR grant MOP84041 (C.A.P.) from the New Emerging Team in Clinical Autoimmunity: Immune Regulation and Biomarker Development in Pediatric and Adult Onset Autoimmune Diseases, the Canada Research Chair program (C.A.P.), NIH grant K08-AI-063267 (T.R.T.), USIDnet grant N01-AI30070 (T.R.T.), NIH grant PO1 AI073489 (M.I.G.), and the Abramson Family Cancer Research Institute (M.I.G.). Author contributions: E.H., K.B.D., and C.A.P. designed the research. K.B.D., E.H., Y.N., Y.X., W.H., R.I., and F.A. performed the experiments. K.B.D. and C.A.P. wrote the manuscript. S.S. performed the microarray analysis. M.B.-S., B.M., H.O., T.R.T., and M.I.G. contributed reagents and samples. Y.X., A.B., B.L., C.S., and M.I.G. developed the TIP60 modifiers used in the study. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: The microarray data have been deposited in the Gene Expression Omnibus database (accession number GSE99236).
View Abstract

Navigate This Article