Subcellular localization of FOXP3 in human regulatory and nonregulatory T cells

Authors


Full correspondence: Dr. Michael H. Albert, Dr. von Haunersches Kinderspital der LMU, Lindwurmstr.4, 80337 Munich, Germany

Fax: +49-89-5160-4719

e-mail: michael.albert@med.uni-muenchen.de

Abstract

The transcriptional regulator FOXP3 is an important determinant of regulatory T (Treg) cell development and function and is frequently used to quantitate Treg cells. However, FOXP3 is also expressed in recently activated conventional human T cells. Here, we investigated the FOXP3 expression patterns in Treg and activated T cells at a cellular level. Upon activation, human CD4+CD25 T cells expressed FOXP3 mainly in the cytoplasm, in sharp contrast to human CD4+CD25+ Treg cells, where we found FOXP3 to be predominantly expressed in the nucleus. A GFP-FOXP3-fusion protein shuttled from the nucleus to the cytoplasm in transfected primary human T cells. We identified two novel leucine-rich nuclear export signals in FOXP3. Site-directed mutagenesis of both sequences completely abolished nuclear export of FOXP3 in human T cells. Both export sequences localized to exons affected by alternative splicing. The three isoforms FOXP3Δ2, FOXP3Δ7, and FOXP3Δ2Δ7 localized preferentially to the nucleus. Additionally, forced expression of nucleus-directed FOXP3 induced a Treg-cell-associated gene expression pattern and induced regulatory capacity. These findings should aid in the interpretation of future studies utilizing FOXP3 expression as a Treg-cell marker and shed some light on the molecular mechanisms controlling subcellular FOXP3 localization in human T cells.

Introduction

Naturally occurring CD4+CD25+ regulatory (Treg) T cells play a fundamental role in maintaining immune homeostasis [[1]]. Alterations in the number or functional activity of Treg cells have been observed in several immunological disease states [[2]]. Furthermore, studies in mice have reported beneficial effects of Treg-cell-based immunotherapies in a wide range of disease models. This has quickly prompted the initiation of early phase clinical trials in humans with promising preliminary results [[3-5]]. A stringent definition of the human Treg-cell phenotype is, however, a prerequisite for such studies. Therefore, much effort has been put into characterizing human Treg cells based on the expression of specific markers, ideally on the cell surface. In addition to CD25, other surface markers such as CD62L, GITR, LAG-3, and CTLA-4 were identified to be expressed in Treg cells, albeit not exclusively so [[6-10]]. These markers are also up-regulated in recently activated nonregulatory human T cells, limiting their usefulness as markers [[11-13]]. The same is true for CD127 which was found to be absent on Treg cells but not on conventional T cells. It has recently been shown to be down-regulated upon activation [[11, 14]]. These findings cast some shadow on numerous studies using these surface markers to describe the Treg-cell phenotype in human disease [[15-17]]. The identification of one or more dependable identifier molecules could significantly propel Treg-cell research and clinical application forward.

The transcription factor forkhead box protein P3 (FOXP3) is considered to be the master regulator for the development and function of Treg cells [[18, 19]]. It is highly expressed in Treg cells and generally accepted as the single best marker to identify Treg cells [[19]]. But just as the surface markers named above, FOXP3 expression is not Treg cell-specific but is also up-regulated in nonregulatory CD4+ T cells upon activation [[20]]. Kinetic studies of FOXP3 expression in activated human T cells reveal a transient up-regulation of the protein that is not sufficient to induce a regulatory phenotype in vitro [[11]]. These findings might suggest a differential function of FOXP3 in activated T cells and in Treg cells. The complex nature of FOXP3 is further highlighted by the presence of different splice variants [[20-22]], whose functions remain largely obscure.

One would expect that the cellular functions of FOXP3 are tightly controlled, FOXP3 being such an important molecule for immune homeostasis in general and Treg-cell function in particular. Recently, it has been reported that a substantial subset of murine naturally occurring Treg cells can lose their FOXP3 expression and become auto aggressive, stressing the notion that regulation of FOXP3 expression is an important physiological switch in T-cell immunity [[23]]. Humans with mutations in FOXP3 exhibit the so-called immune dysregulation, polyendocrinopathy, enteropathy, and X-linked (IPEX) syndrome, which is characterized by excessive autoimmunity in infancy and has been associated with both defective regulatory and effector T-cell function in affected individuals [[24]]. In a recent report addressing the molecular basis of IPEX, different mutations in FOXP3 were described affecting both expression level and subcellular localization of the protein [[25]]. This has prompted us to investigate the posttranslational differences in FOXP3 expression between Treg cells and activated non-Treg cells in more detail.

Indeed, in this study, we show that FOXP3 is localized in different subcellular compartments in conventional human CD4+ T cells and Treg cells. The nuclear-cytoplasmic shuttling of FOXP3 can be attributed to two newly identified nuclear export sequences (NESs) and a previously described nuclear localization sequence (NLS) [[26]]. Furthermore, the FOXP3 isoforms FOXP3Δ2, FOXP3Δ7, and FOXP3Δ2Δ7, devoid of either or both of the NESs, preferentially localize to the nucleus.

Results

FOXP3 is differentially localized in activated T cells and Treg cells

FOXP3 expression is up-regulated in human T cells upon activation [[11, 20]]. Since FOXP3 has been touted as the key designator of Treg cells, we sought to detect differences in FOXP3 expression in both cell populations on a quantitative and single cell level.

As described by others before, freshly isolated CD4+CD25 T cells did not express significant levels of FOXP3. Upon activation, these T cells transiently up-regulated FOXP3 protein peaking at day 2, before it declined to baseline levels. In activated CD4+CD25+ T cells, FOXP3 markedly increased over time and then returned to preactivation levels (Supporting Information Fig. 3A and B). After 2 days of activation, FOXP3 expression in CD4+CD25 T cells (MFI = 104.4 ± 13.5) was comparable with that of naïve CD4+CD25+ T cells (MFI = 118.0 ± 8.1, n = 6, p = 0.07). However, when added to a proliferation assay with CFSE-labeled responder T cells, these FOXP3 expressing cells had no suppressive capacity (Supporting Information Fig. 3C).

Thus, total FOXP3 protein expression does not allow to discriminate between CD4+CD25+ Treg cells and activated human CD4+ T cells, and activation induced FOXP3 expression does not confer regulatory capacity to these cells. Since FOXP3 acts as a transcriptional repressor in Treg cells [[27]], we hypothesized that it would be located in the nucleus.

When we analyzed FOXP3 expression on a subcellular level by confocal microscopy as well as by immunoblot analysis, we observed a marked difference of FOXP3 distribution in activated human CD4+CD25 T cells and CD4+CD25+ Treg cells. In highly purified FACS-sorted CD4+CD25+ T cells (Supporting Information Fig. 2A), FOXP3 protein co-localized exclusively with the nuclear DAPI staining and was not redirected within the cell upon activation with CD3/CD28 beads. In contrast, FOXP3 rapidly appeared in the cytoplasm of CD4+CD25 T cells upon activation with only little nuclear fluorescence signal (Fig. 1A and B). It was previously shown that Treg cells can be further subgrouped phenotypically and functionally based on the expression of CD45RA [[28]]. Although CD45RA+ and CD45RA Treg cells are both suppressive when freshly isolated, only the CD45RA+ Treg cells were shown to result in homogenous regulatory T-cell lines when expanded in vitro [[29]]. We therefore analyzed these particular Treg-cell subsets, asking whether FOXP3 would localize differentially. After sorting of CD4+CD25+CD45RA+ and CD4+CD25+CD45RA Treg cells (Supporting Information Fig. 2B) and stimulation with CD3/CD28 beads, we found FOXP3 to be mainly expressed in the nucleus of CD45RA+ Treg cells, in contrast to CD45RA Treg cells where it was predominantly residing in the cytoplasm (Fig. 1C).

Figure 1.

FOXP3 is differentially localized in activated conventional T cells and Treg cells. (A) Confocal microscopic analysis of sorted CD4+CD25 and CD4+CD25+ T cells before (0 d) and after stimulation (2 d, 5 d, 8 d). FOXP3 was stained with Alexa488-FOXP3 antibody and nuclear DNA was stained with DAPI. (B) Subcellular fractionation of CD4+CD25 and CD4+CD25+ T cells before (0 d) and after stimulation (2 d, 5 d, 8 d). Immunoblot was probed with anti-FOXP3. Purity of subcellular fractions was confirmed with actin and lamin A/C stainings. Please note a faint cytosolic cross-contamination in the nuclear fraction of CD25 T cells at day 0 and day 5. Blots are representative of three independent experiments. (C) Confocal microscopic analysis of sorted CD4+CD25+CD45RA (CD45RA) and CD4+CD25+CD45RA+ (CD45RA+) T cells 2 and 5 d after stimulation. FOXP3 was stained with Alexa488-Foxp3 antibody. Counterstains were done with DAPI. (D) T cells were transfected with a dual-expression plasmid expressing GFP-FOXP3-fusion protein or GFP together with ΔLNGFR. Nontransfected and LNGFR-selected GFP-FOXP3 and GFP-transfected T cells were stained for intracellular FOXP3 expression with PE-anti-FOXP3 and analyzed 24 h after transfection. Data shown are representative of one of three experiments with similar results. (E) Confocal microscopic analysis of GFP-FOXP3 and GFP-transfected primary human T cells 5 and 24 h after transfection. (F) Confocal microscopic analysis of GFP-FOXP3 transfected HeLa cells 5 and 24 h after transfection. Nuclei were stained with DAPI. Microscopic images are representative of three independent experiments with more than 100 cells analyzed on each slide. Isotype control was used to rule out nonspecific staining (data not shown). Scale bar = 20 μm. C: cytoplasmic fraction; N: nuclear fraction.

In summary, these findings show that the subcellular localization of FOXP3 protein differs between CD4+CD25+ Treg cells and recently activated conventional CD4+ T cells, which allows for discrimination of these otherwise phenotypically inseparable cell populations.

FOXP3 shuttles from the nucleus to the cytoplasm

To exclude that the cytoplasmic expression of FOXP3 protein was only simulated by nonspecific binding of antibody in the cytoplasm of activated T cells, we ectopically expressed GFP-FOXP3-fusion protein in primary human CD4+CD25 T cells. We used a dual-expression plasmid that allows co-expression of the truncated low-affinity nerve growth factor receptor (ΔLNGFR) for positive selection of transfected T cells [[30]].

GFP-FOXP3 and GFP (control) transfected T cells were highly enriched as shown by GFP fluorescence and intracellular FOXP3 staining (Fig. 1D). After initial nuclear expression, exogenously expressed GFP-FOXP3-fusion protein rapidly shuttled from the nucleus to the cytoplasm within 24 h whereas GFP alone evenly distributed within the cell (Fig. 1E). We observed redistribution of FOXP3 also in the Jurkat T-cell line (not shown) but not in Hek293 (not shown) or HeLa cells (Fig. 1F), suggesting that nucleo-cytoplasmic shuttling of FOXP3 is a T-cell-specific process.

In summary, exogenously expressed FOXP3 protein was quickly shuttled from the nucleus to the cytoplasm in primary human T cells. This likely excluded nonspecific antibody binding and indicated that FOXP3 could also contain nuclear export signals in addition to an already known NLS [[26]].

Identification of NESs in FOXP3

Due to the size limitation of nuclear pores, proteins above ∼40 kDa of size use an active transport machinery in the presence of specific transport signals to pass the nuclear envelope in either direction [[31]]. A nuclear import signal (NLS), corresponding to amino acids 414RKKR417, has already been described in the forkhead domain of FOXP3 [[26]]. The differential subcellular localization of FOXP3 in activated T cells versus Treg cells and the translocation of GFP-FOXP3 in transfected T cells described above led us to speculate on the additional presence of at least one NES. Therefore, we examined the amino acid sequence of FOXP3 with regard to leucine-rich NESs [[32]] and identified two separate leucine-rich regions in FOXP3 that fit the criteria for a NES with the consensus Lx1-3Lx2,3LxL, where x is any amino acid and L can be substituted with amino acids M, V, I, or F [[33]]. A detailed alignment of both putative export sequences with known export signals is shown in Fig. 2A. Both export sequences are conserved between mouse and human, except for a conservative amino acid substitution L74V in NES1. A detailed sequence alignment from various species is shown in the Supporting Information Fig. 4. The first NES (NES1) lies in the exon 1/2 boundary zone and is partly destroyed in the FOXP3Δ2 isoform [[20]]. The second NES (NES2) localizes within the leucine-zipper domain and is encoded by exons 6 and 7. Exon 7 is missing in FOXP3Δ2Δ7 and the recently described FOXP3Δ7 isoform [[21, 22]] (Fig. 2B). To assess whether these sequences can indeed act as functional export signals they were analyzed in a model system [[34]]. The corresponding peptide sequences 68QLQLPTLPL76 (for NES1) and 239VQSLEQQLVL248 (for NES2) of human FOXP3 were fused to GFP and expressed in Jurkat T cells. A GFP fusion construct with the NLS of FOXP3 414RKKR417 was used to confirm the assay. GFP itself does not contain any NLS or NES and distributes both in the nucleus and the cytoplasm. When fused with the putative NLS, GFP was visibly enriched in the nucleus. Fusion of either of the two putative nuclear export signals with GFP resulted in expression of the protein in the cytoplasm (Fig. 2C).

Figure 2.

Identification of two NESs in human FOXP3. (A) Sequence comparison of two putative NESs in FOXP3 with known export signals [[33]]. (B) Schematic diagram of FOXP3 coding regions (starting with the first coding exon) and corresponding protein domains. NES1 is encoded by exon 1/2 and lies in the repressor domain of FOXP3. NES2 maps to the leucine-zipper domain of FOXP3 and is encoded by exon 6/7. The NLS localizes to the extreme c-terminal end of FOXP3. (C) The peptides resembling NES1 (aa 68-76), NES2 (aa 239-248), and NLS (aa 414-417) were fused to GFP and expressed in Jurkat T cells. Microscopic images are representative of three independent experiments. Scale bar = 10 μm.

These results show that in addition to an NLS, human FOXP3 contains two putative NESs.

Mutation of NESs in FOXP3 targets the protein to the nucleus

After showing that both putative export sequences were functional in a model system, we next addressed the question whether the identified export sequences were also functional in the context of full-length FOXP3 protein.

By targeted mutation of leucine to alanine at specific sites within NES1 (L69/71/74/76A) and NES2 (L242/246/248A) of a GFP-FOXP3-fusion protein (Supporting Information Fig. 1B), we created two single mutant constructs GFP-FOXP3-NES1m and GFP-FOXP3-NES2m and one double mutant GFP-FOXP3-NES1/2m. The NLS in FOXP3 was disrupted by the exchange of lysine 115/116 to glutamate (GFP-FOXP3-NLSm). The different FOXP3 mutants were then expressed in primary human CD4+ T cells and analyzed by confocal microscopy. The percentages of cells with different FOXP3 localization are shown for each construct in the right panel of Fig. 3. FOXP3 expressing cells were grouped into nuclear (N) or cytoplasmic (C) localization according to the brightest fluorescence intensity of GFP in the particular compartment. Cells with equal GFP fluorescence in both compartments were determined as nuclear and cytoplasmic (N/C) localization. As expected, NLS-mutated FOXP3 did not enter the nucleus at any time point (Fig. 3, second panel). Wild-type GFP-FOXP3 (GFP-FOXP3-wt) fusion protein was initially predominantly expressed in the nucleus, but rapidly redistributed to the cytoplasm within 24 h after transfection (Fig. 3, first panel). When expressing the NES-mutant forms, we observed a marked difference compared with wild-type GFP-FOXP3. Both single mutants GFP-FOXP3-NES1m and GFP-FOXP3-NES2m translocated to the cytoplasm with a slower kinetic than GFP-FOXP3-wt (Fig. 3, third and fourth panel). Furthermore, both NES mutants distributed equally in the nucleus and the cytoplasm, compared with FOXP3-wt which accumulated predominantly in the cytoplasm 72 h after transfection. The double mutant however remained in the nucleus as long as the protein was detectable under the transient transfection conditions (Fig. 3, fifth panel).

Figure 3.

Mutation of NESs in FOXP3 targets the protein to the nucleus. GFP-fusion proteins of FOXP3 wild-type (GFP-FOXP3-wt) and with mutated NLS (GFP-FOXP3-NLSm), NES1 (GFP-FOXP3-NES1m), NES2 (GFP-FOXP3-NES2m), and NES1/2 (GFP-FOXP3-NES1/2m) were expressed in primary human T cells and analyzed by confocal microscopy. DNA was stained with DAPI. Scale bar = 20 μm. Microscopic images are representative of five independent experiments with more than 50 cells analyzed on each slide. Graphs with the percentage of cells with different FOXP3 localization are shown on the right. N: nuclear localization; C: cytoplasmic localization; N/C: nuclear and cytoplasmic localization.

In summary, we could demonstrate that the subcellular localization of FOXP3 can be controlled by its nuclear import sequence and two newly described NESs. Disruption of both export sequences almost entirely prevented nucleo-cytoplasmic redistribution.

FOXP3 isoforms FOXP3Δ2, FOXP3Δ7, and FOXP3Δ2Δ7 localize differentially in activated human T cells

In theory, both NESs could have a differential impact on subcellular distribution of the alternatively spliced human FOXP3 isoforms. NES1 and NES2 occupy the exon 1/2 and exon 6/7 boundaries and are consequently both affected by alternative splicing of exons 2 and 7, respectively [[20, 22]]. To test this, the three known FOXP3 isoforms FOXP3Δ2 (lacking NES1), FOXP3Δ7 (lacking NES2), and FOXP3Δ2Δ7 (lacking NES1 and NES2) were fused with GFP and expressed in primary human T cells. As shown in Fig. 4, we observed redistribution of GFP-FOXP3Δ2 (first panel) and GFP-FOXP3Δ7 (second panel) albeit with a slower kinetic than full-length GFP-FOXP3 (Fig. 3, first panel). Similar to the double NES mutant GFP-FOXP3-NES1/2m (Fig. 3, fifth panel), the isoform GFP-FOXP3Δ2Δ7 localized also predominantly in the nucleus (Fig. 4, third panel).

Figure 4.

The three different FOXP3 isoforms FOXP3D2, FOXP3Δ7, and FOXP3Δ2Δ7 localize to the nucleus in transfected primary human T cells. Confocal microscopic images were taken at the indicated time points. DNA was counterstained with DAPI. Scale bar = 20 μm. Images are representative for three individual experiments with more than 50 cells analyzed on each slide. Graphs with the percentage of cells with different FOXP3 localization are shown on the right. N: nuclear localization; C: cytoplasmic localization; N/C: nuclear and cytoplasmic localization.

These results show that the alternative FOXP3 isoforms localize differently than the full-length FOXP3 protein in primary human T cells according to the presence or absence of the respective nuclear export signals.

Transcriptional activity of FOXP3 is dependent on its subcellular localization

FOXP3 is known to act both as a nuclear transcriptional activator and repressor [[35]]. Therefore, one would expect that nuclear localization is essential for its transcriptional activity and that altered subcellular localization as we had observed in activated CD4+CD25 cells would affect its transcriptional activity.

To test this hypothesis, we isolated mRNA from transfected and LNGFR-selected primary human CD4+ T cells expressing either GFP-FOXP3-NES1/2m or GFP-FOXP3-wt. Transfected T cells expressing only GFP were used as a reference (Fig. 5A). The mRNA levels of a panel of Treg-cell-specific genes were then measured with quantitative real-time PCR using the human T regulatory phenotyping StellArray Gene Expression System. Changes in gene expression between GFP-FOXP3-NES1/2m and GFP-FOXP3-wt were calculated using the global pattern prediction algorithm using control-transfected T cells expressing only GFP as reference [[36]]. Nuclear overexpression of FOXP3 induced a number of Treg-cell-associated genes such as IL-2 receptor α, CCR4, and Gzmb, and down-regulated proinflammatory cytokine expression, such as IL-2, IL-4, or TNF-α (Fig. 5B). However, expression of other known FOXP3 targets including CTLA-4 and GITR was not affected in the transfected T cells.

Figure 5.

Transcriptional activity of FOXP3 is affected by its subcellular localization. (A) LNGFR-enriched primary human T cells transfected with GFP, GFP-FOXP3-NES1/2m, and GFP-FOXP3-wt were evaluated for gene expression. (B) Fold change in expression of various Treg-cell-specific genes between T cells expressing nuclear (GFP-FOXP3-NES1/2m) and cytoplasmic (GFP-FOXP3-wt) FOXP3 identified using global pattern recognition software analysis. Data shown are representative of one of two experiments.

These data show that different localization of FOXP3 has distinct effects on the expression of known Treg-cell-associated genes implying an important role for nuclear localization of FOXP3 protein in preserving the integrity of the Treg-cell phenotype.

Nuclear FOXP3 induces regulatory capacities in transfected T cells

Having shown that expression of nuclear localizing mutant FOXP3 (GFP-FOXP3-NES1/2m) has downstream gene regulatory effects distinct from FOXP3-wt, we intended to analyze whether forced nuclear localization of FOXP3 would induce regulatory capacities in transfected T cells.

The suppressive activities of FOXP3-expressing T cells were thus determined in CFSE dilution assays. Transfected T cells were positively enriched with LNGFR selection 16 h after transfection and then added to CD3/CD28 bead stimulated CFSE-labeled responder CD3+ T cells. As shown in Fig. 6, GFP-FOXP3-wt transfected T cells displayed only moderate suppressive capacity (23.6 ± 11.7%) as compared with GFP-FOXP3-NES1/2m transfected T cells (48.8 ± 14.8%). Although similarly expressed in the nucleus as the GFP-FOXP3-NES1/2m mutant, no significant suppressive activity was observed in GFP-FOXP3Δ2Δ7 overexpressing T cells (13.2 ± 0.7%).

Figure 6.

Suppressive activities of FOXP3 transfected primary human T cells. (A) T cells were transfected with control vector (expressing LNGFR alone), GFP-FOXP3-wt, GFP-FOXP3-NLSm and GFP-FOXP3-NES1/2m, and GFP-FOXP3Δ2Δ7. Suppressive activity was determined by adding transfected T cells to CD3/CD28-stimulated, CFSE-labeled CD3+ responder T cells at a ratio of 1:1 and evaluating CFSE dye dilution. Nontransfected CD4+ T cells were used as control. Representative data from a single experiment are shown. (B) Data are shown as mean + SD of n = 6 samples pooled from six independent experiments, each performed in triplicate. *p = 0.01. Comparison of means was calculated using the two-tailed Student's t-test.

In summary, these data show that nuclear localization of FOXP3 is a prerequisite for conferring regulatory potential to transfected T cells. Overexpression of the Δ2Δ7 splice variant is however not sufficient to induce a regulatory phenotype.

Discussion

The transcription factor FOXP3 is well accepted to be the master regulator of Treg-cell development and function, and to be the best molecule to identify this T-cell subpopulation [[1]]. However, like other Treg-cell markers it is not exclusively expressed in Treg cells but also up-regulated in conventional T cells upon activation, thus limiting its use for defining the human Treg-cell phenotype [[11, 20, 37, 38]]. In the study presented here, we describe for the first time different subcellular localization of FOXP3 in primary human CD4+CD25+ Treg cells and recently activated conventional CD4+ T cells. It had previously been shown that CD4+CD25+ Treg cells comprise a heterogeneous cell population that can be further subdivided based on expression of CD45RA into naïve and memory Treg cells [[28]]. Within these subpopulations only the CD45RA+ Treg cells can yield stable, homogenous regulatory T-cell lines upon in vitro expansion [[29]]. In line with these observations, we found FOXP3 to be predominantly expressed in the nucleus of activated CD45RA+ Treg cells but not activated CD45RA Treg cells where it is mainly distributed in the cytoplasm. This further strengthens the notion that only the CD45RA+ subset of Treg cells can withstand reversion to a conventional T-cell phenotype after in vitro expansion, making nuclear FOXP3 expression an additional potentially useful marker for in vitro expanded Treg cells. This differential localization of FOXP3 can be attributed to two newly described NES and one previously reported NLS and results in differential downstream gene expression.

The novel finding that human FOXP3 can be found predominantly in the cytoplasm of activated T cells is seemingly in contrast to a report by Allan et al., who showed FOXP3 in the nuclear fraction of these cells [[11]]. However, we also detected a smaller amount of FOXP3 in the nucleus and more importantly the cytosolic fraction was not analyzed in the study by Allan et al. [11]. In transfected primary human T cells, we observed rapid redistribution of GFP-FOXP3-fusion protein from the nucleus to the cytoplasm. This seems to be in contrast with a recent report by van Loosdregt et al. who found transfected Flag-tagged FOXP3 protein to be selectively expressed in the nucleus [[39]]. However, these localization studies were performed in Hek293 cells. But nuclear export of FOXP3 appears to be cell-type specific, as we also did not observe redistribution in nonhematopoietic cell lines such as Hek293 or HeLa. Cytoplasmic relocalization of FOXP3 might be a T-cell-specific phenomenon. Chen et al. have recently reported an activation-dependent shift of FOXP3 from a cytoplasmic/perinuclear pattern to a more dense nuclear localization in murine Treg cells and in FOXP3-transduced Jurkat cells [[40]]. This also does not contradict our findings of redistribution of FOXP3 in the opposite direction from the nucleus to the cytoplasm, as Chen et al. explored FOXP3 distribution in CD4+ T cells and Jurkat cells only shortly after activation (<5 h). At these time points, we also found FOXP3 protein to be still expressed in the nucleus.

Relocalization of FOXP3 in primary human T cells led us to speculate on the presence of at least one NES in FOXP3. We then identified two NESs (here referred to as NES1 and NES2) in human FOXP3 that match with the consensus export signal Lx1-3Lx2,3LxL, previously reported to be present in many nucleo-cytoplasmic shuttling proteins [[33]].

In theory, both NESs could affect the localization of the alternatively spliced human FOXP3 isoforms NES1 and NES2 [[20, 22]] and alternative splicing may offer one explanation for differential FOXP3 localization in human Treg cells and conventional T cells. Indeed, we found marked differences between the subcellular localization of full-length FOXP3 and the three known FOXP3 isoforms FOXP3Δ2, FOXP3Δ7, and FOXP3Δ2Δ7 in transfected primary T cells. While full-length FOXP3 rapidly shuttled from the nucleus to the cytoplasm, the isoforms FOXP3Δ2 and FOXP3Δ7, lacking either of the two NESs translocated to the cytoplasm with slower kinetics. The isoform FOXP3Δ2Δ7, lacking both export sequences, localized to the nucleus. However, this did not result in the induction of a functional Treg-cell phenotype in transfected cells in our study. In line with our observations, the FOXP3Δ7 isoform lacking NES2 was recently found to be preferentially expressed in CD4+CD25+ Treg cells but only after activation [[21]]. In resting Treg cells and in activated conventional T cells, full-length FOXP3 was predominant. So while differential expression patterns of alternatively spliced FOXP3 isoforms could theoretically contribute to the development or function of Treg cells, whether this is in fact the case and what are the exact mechanisms involved still remains to be determined in further studies. Alternative splicing might be only one aspect in regulating subcellular distribution of FOXP3. In conventional CD4+CD25 and CD4+CD25+ Treg cells, we found differential distribution of full-length FOXP3, despite the presence of both NESs. This argues for the existence of additional cell type specific mechanisms that regulate subcellular distribution of FOXP3.

One would suggest that nuclear localization of FOXP3 is a prerequisite for its function as a transcriptional regulator. Indeed, when analyzing Treg-cell-specific genes, we observed marked differences in transfected T cells expressing either nuclear-trapped FOXP3 (FOXP3-NES1/2m) or wild-type FOXP3, suggesting the functional importance of its nuclear localization for preserving Treg-cell marker expression. This is further supported by the fact that humans with nondeleterious germline mutations in FOXP3 which lead to cytoplasmic but not nuclear expression have severely deficient Treg-cell function and display an IPEX phenotype even when normal quantitative levels of the protein are expressed [[25]]. Furthermore, when analyzing the functional properties of nuclear-trapped full-length FOXP3, we observed increased suppressive activity compared with wild-type FOXP3 and FOXP3Δ2Δ7-transfected T cells.

Although the importance of FOXP3 in regulatory T-cell development and function is well described, its molecular action in human T cells still remains largely obscure. The complexity of FOXP3 biology is highlighted by the fact that conventional human T cells transiently up-regulate FOXP3 expression upon activation without acquiring Treg-cell function [[11]]. In line with this observation, conversion of human CD4 T cells into Treg cells by ectopic FOXP3 expression has been shown more convincingly for mouse T cells [[37, 41]] than for human T cells where constant, supranormal, and activation-independent expression of constitutively active FOXP3 is required [[42, 43]]. Even though our data do not elucidate the functional properties of cytoplasmic FOXP3, inactivation of transcriptional activity of FOXP3 by nucleo-cytoplasmic relocalization may be a plausible mechanism by which the Treg-cell phenotype inducing capabilities of FOXP3 in activated and transfected T cells are abolished. Transcriptional inactivation by nucleo-cytoplasmic relocalization has been well described for other members of the forkhead transcription factor family. Relocalization of FoxO factors to the cytoplasm are achieved by complex cooperation of posttranslational modifications and binding partners, thereby terminating their transcriptional function [[44]]. Whether FOXP3 actually realizes additional effector functions in the cytoplasm needs to be addressed in future studies. With regards to this, it was shown recently that the ΔFKH mutant, which does not enter the nucleus, can block NF-κB activation in Jurkat T cells and primary human T cells to a similar extent as full-length FOXP3 [[45]]. This further supports the notion that FOXP3 might indirectly influence transcription in the cytoplasm.

With the demonstration of different splice variants, a further complexity of FOXP3 has evolved in human cells. Depending on the presence or absence of exon 2 and exon 7, FOXP3 can in principle fulfill distinct functions [[46]]. Despite differences between interaction partners and downstream gene regulation, the different export kinetics and localizations of FOXP3 isoforms might provide further fine tuning of transcriptional regulation. Thus, alternative splicing of FOXP3 might be an important mechanism enabling different aspects of FOXP3 function in T cells. Although we could demonstrate that nuclear expression of FOXP3 is a prerequisite for induction of regulatory potential, the FOXP3Δ2Δ7 isoform failed to inhibit lymphoproliferation compared with the double NES mutant FOXP3-NES1/2m. Therefore, the missing exons are probably essential for transmitting regulatory potential. This is in agreement with a recent report from Mailer et al., showing similar results for FOXP3Δ2Δ7 in murine cells [[47]].

This study shows for the first time that subcellular localization of FOXP3 can help to distinguish between recently activated conventional T cells and bona fide Treg cells. This finding should aid in the correct interpretation of future studies utilizing FOXP3 expression as a Treg-cell marker and might argue for a differential function of FOXP3 in activated T cells and Treg cells. Furthermore, this study sheds some light on the molecular mechanisms controlling subcellular FOXP3 protein localization in human T cells, which might be of great importance in preserving the Treg-cell phenotype. Future studies will have to focus on regulatory pathways that alter expression of FOXP3 splice variants, on cytoplasmic and nuclear interaction partners of FOXP3 regulating its nuclear-cytoplasmic shuttling and on the posttranslational regulation of this important determinator of Treg-cell function.

Materials and methods

Cell culture

Peripheral blood mononuclear cells were obtained from healthy donors in accordance with the local ethical committee of the LMU München. Naturally occurring CD4+CD25+ T cells and CD4+CD25 T cells were purified by immunomagnetic beads using the Treg isolation kit II (Miltenyi Biotec, Bergisch Gladbach, Germany) or by FACS sorting using a MoFlo sorter (Cytomation, Fort Collins, CO, USA). Purity of sorted CD4+CD25+ cells was always greater than 97% based on CD25 expression and greater than 93% based on FOXP3 expression. CD4+ and CD4+CD25 T cells for transfection experiments were enriched by negative selection (Miltenyi Biotec). Jurkat T-cell clone E6-1 (ATCC, Manassas, VA, USA) and primary T cells were maintained in RPMI 1640 supplemented with 2 mM glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, and 10% FCS (PAA-laboratories, Pasching, Austria) in a 5% CO2 incubator at 37°C. T cells were stimulated with CD3/CD28 beads (Dynal Biotech, Oslo, Norway) at a ratio of 4:1 (cells:beads). Mixed lymphocyte reaction (MLR) was carried out as described previously [[48]]. Freshly isolated Treg cells and stimulated conventional T cells were added to a secondary MLR consisting of CFSE-labeled responder CD3+ cells from the original donor and allogeneic mitomycin-treated CD3-depleted stimulator APC at a ratio of 5 × 104 to 2.5 × 105 cells per 96-well (Cellstar) in complete media. Proliferation of CFSE-labeled effector T cells was measured after 5 days in culture. T cells that had undergone at least one cell division were designated as alloreactive T cells.

Plasmids

The cDNA encoding full-length FOXP3 and FOXP3Δ2 isoform was reversely transcribed from mRNA of primary human CD4+ T cells and amplified with primer pair P1/P2. Primer sequences are shown in the Supporting Information Fig. 1A. GFP-FOXP3 fusion was carried out by PCR cloning. The PCR products of the first PCR reactions using full-length FOXP3 as template with primer pairs P1/P4 and P2/P3 was combined and fused in a second PCR reaction with primer pair P1/P2. Point mutations were introduced with the following primer combinations: P1/P6 and P2/P5 for GFP-FOXP3-NLSm, P1/P8 and P2/P7 for GFP-FOXP3-NES1m, and P1/P10 and P2/P9 for GFP-FOXP3-NES2m. The double mutant GFP-FOXP3-NES1/2m was created by introducing point mutations with the primer pairs P1/P8 and P2/P7 using GFP-FOXP3-NES2 as template. The human isoforms FOXP3Δ7 and FOXP3Δ2Δ7 were cloned by PCR using full-length FOXP3 or FOXP3Δ2 as template with the primer pairs P1/P12 and P2/P11. PCR products were then fused in a second PCR reaction using primer pair P1/P2. The different GFP-FOXP3-fusion variants were ligated into the XbaI/NotI site downstream of the CMV promoter of pCDH-CMV-MCS-EF1-ΔLNGFR [[30]] and verified by sequencing (Eurofins MWG Operon, Ebersberg, Germany).

Transfection

Primary human CD4+ T cells and Jurkat T cells were nucleofected with 5 μg of plasmid DNA at a cell density of 1 × 107 cells per 100 μL nucleofector solution (human T-cell nucleofector kit VPA-1002, Lonza, Cologne, Germany) with the Amaxa nucleofector device using program U-14 (Lonza). After nucleofection, cells were immediately transferred into prewarmed hTC culture medium (Lonza) supplemented with 2 mM glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, 10% FCS (PAA-laboratories). LNGFR-positive cells were enriched with anti-LNGFR microbeads 5 h after nucleofection (Miltenyi Biotec) as previously described [[30]]. HeLa cells were transfected using Lipofectamine2000 (Invitrogen, Paisley, UK).

Suppression assay

Suppressive activities of FOXP3-transfected primary human T cells was determined by adding transfected T cells to CD3/CD28 bead stimulated CD3+ CFSE-labeled responder T cells at a ratio of 1:1. Responder T cells were activated by CD3/CD28 beads at a ratio of 40:1. Proliferation was measured after 4 days in culture. Gates were set on live CD4+LNGFR T cells.

Gene expression analysis

Nucleofected and LNGFR-positive selected T cells were stimulated for 48 h with CD3/CD28 beads at a ratio of 4:1. Total RNA was isolated using QIAshredder and RNeasy Mini Kit (Qiagen, Hilden, Germany) from 1 × 106 cells. A total of 400 ng RNA was reverse transcribed to cDNA with QuantiTect (Qiagen) using oligo-dT primer. Gene expression analyses were done with the human T regulatory phenotyping StellArray Gene Expression System from Lonza with the Bio-Rad iCyclerIQ platform using SYBR Green (Bio-Rad, Munich, Germany). Changes in gene expression were analyzed with the global pattern recognition algorithm [[36]].

Flow cytometry

T cells were stained for cell surface markers with allophycocyanin-anti-CD4 (eBioscience, San Diego, USA), PC5-anti-CD25 (Beckman Coulter, Fullerton, CA, USA), PE-anti-LNGFR (Miltenyi Biotec), prior to intracellular staining of FOXP3 with Alexa488-anti-FOXP3 or PE-anti-FOXP3 (eBioscience) according to the manufacturer's instructions. FACS analysis was carried out on a FACS Canto II (BD Biosciences).

Fluorescence microscopy

For fluorescent microscopy, 2.5 × 105 cells were stained with A488-anti-FOXP3 clone PCH101 (eBioscience). Nuclei were stained with DAPI (Sigma-Aldrich, Hamburg, Germany). A488-anti-rat IgG2a (eBioscience) was used as isotype control. Cells were centrifuged on object slides at 200 × g for 3 min, mounted in Aqua-Poly/Mount (Polysciences, Eppelheim, Germany) and analyzed on an Olympus IX81 confocal microscope using the UPLanSApo 60×1.35 oil objective (Olympus, Hamburg, Germany).

Western blot analysis

Subcellular fractionations were performed with the nuclear/cytosol fractionation kit from BioVision (Mountain View, CA, USA). Samples were separated on a 10% SDS-PAGE and transferred to nitrocellulose membranes. Blocking was performed with 1% BSA in PBS. Membranes were probed with anti-FOXP3 antibody clone 259D (Biolegend, San Diego, USA), antiactin and antilamin A/C (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Actin was used as a loading control for cyctosolic extracts. Lamin was used as a loading control for nuclear extracts. Lamin was preferred to the transcription factor SP1, which is also frequently used as a marker for nuclear extracts, because it was recently reported to be also located in the cytoplasm [[49]]. Secondary goat antimouse antibody was purchased from Pierce, Rockford, USA.

Acknowledgements

This project was supported in part by the Wilhelm-Sander-Stiftung (to M.H.A.) and by the “Mehr LEBEN fuer krebskranke Kinder - Bettina - Braeu-Stiftung” (to M.H.A.).

Conflict of interest

The authors declare no financial or commercial conflict of interest.

Abbreviations
NES

nuclear export sequence

NLS

nuclear localization sequence

Ancillary