Ectopic FOXP3 Expression Preserves Primitive Features Of Human Hematopoietic Stem Cells While Impairing Functional T Cell Differentiation

FOXP3 is the transcription factor ruling regulatory T cell function and maintenance of peripheral immune tolerance, and mutations in its coding gene causes IPEX autoimmune syndrome. FOXP3 is also a cell-cycle inhibitor and onco-suppressor in different cell types. In this work, we investigate the effect of ectopic FOXP3 expression on HSC differentiation and we challenged this approach as a possible HSC-based gene therapy for IPEX. FOXP3-expressing HSC showed reduced proliferation ability and increased maintenance of primitive markers in vitro in both liquid and OP9-ΔL1 co-cultures. When transplanted into immunodeficient mice, FOXP3-expressing HSC showed significantly enhanced engraftment ability. This was due to a pronounced increase in the frequency of repopulating cells, as assessed by extreme limiting dilution assay. Likely underlying the increased repopulating ability, FOXP3 expressing HSC showed significantly enhanced expression of genes controlling stemness features. However, peripheral T cells developed in the FOXP3-humanized mice were quantitatively reduced and hyporesponsive to cytokine and polyclonal stimulation. Our findings reveal unpredicted effects of FOXP3 in the biology of HSC and may provide new tools to manipulate primitive features in HSC for clinical applications. Moreover, they formally prove the need of preserving endogenous FOXP3 regulation for an HSC-based gene therapy approach for IPEX syndrome.

Thus, we have tested in this work the effect of lentiviral vector (LV)-mediated constitutive expression of FOXP3 throughout hematopoiesis by transducing human CD34+ hematopoietic stem progenitors cells (HSPCs) and assessing their differentiation into an implemented NSG-based humanized mouse model.

Modulation of the expression of FOXP3 affects HSPC in vitro maintenance and differentiation.
In order to study the impact of constitutive expression of FOXP3 on human hematopoiesis, we transduced cord blood-derived CD34+ HSPCs by LV-vectors expressing FOXP3 (LV-FOXP3) or a control gene (LV-Ctrl) and a reporter gene (either ΔLNGFR or GFP) (Fig. S1A). We obtained 42 ± 6.4% and 57 ± 5.1% reporter gene positive cells in LV-FOXP3 and LV-Ctrl transduced CD34+ cells, respectively (Fig. 1A). FOXP3 expression was well detectable at the protein level in most but not all ΔLNGFR+ LV-FOXP3 transduced CD34+ cells, likely reflecting a higher limit of detection for the intra-cytoplasmic FOXP3 staining compared to the membrane-bound ΔLNGFR. Indeed, FOXP3 RNA expression was comparable, if not higher, to the endogenous levels observed in Tregs, and indicated a very high FOXP3 expression per transduced cell when considering that only a fraction of the assessed CD34+ population was transduced and thus expressing FOXP3 (on average 40%, see Fig. 1A), while all Tregs homogenously express it (Fig. 1B) (see below for FOXP3 expression in ΔLNGFR sorted CD34+).
We performed a first set of in vitro experiments by culturing transduced HSPCs in liquid culture, clonogenic assays and T cell differentiation assays on OP9-ΔL1 stromal cells, and evaluated the effect of FOXP3 expression on proliferative and differentiation ability of these cells. We observed a significant reduction in the proportion of FOXP3 transduced cells in liquid culture during time, when compared to control transduced cells (transduced cells at day 14: 35 ± 7.5% vs 61 ± 10% of the relative transduction measured at day 3, LV-FOXP3 vs LV-Ctrl) (Fig. 1C), indicating a negative selection of LV-FOXP3 HSPCs. While FOXP3 expression did not significantly affect the mortality rate of cultured HSPC-derived cells at any of the time points assessed during the two weeks culture (Fig. 1D), it significantly reduced their proliferation rate (0.63 ± 0.12 vs 0.91 ± 0.12 proliferation rate relative to untransduced cells -set as 1 -7 days post-transduction, LV-FOXP3 vs LV-Ctrl; Fig. 1E and Fig. S1B). Interestingly, we simultaneously observed increased maintenance of primitive CD34hi and CD34hiCD38-markers 14 days after transduction (14 ± 2.2% vs 9.0 ± 1.4% CD34hi, and 8.0 ± 1.7% vs 5.1 ± 0.84% CD34hiCD38-, LV-FOXP3 vs LV-Ctrl; Fig. 1F). The clonogenic potential of HSPCs was not affected by FOXP3 constitutive expression, as evaluated by semisolid culture (Fig. 1G). We then assessed the effect of FOXP3 expression on the ability of HSPCs to differentiate along the T cell lineage when co-cultured on OP9-ΔL1 stromal cells 14 . The percentage of CD3+ T cells arising from LV-FOXP3 and LV-Ctrl HSPCs was comparable (11 ± 1.1% vs 8.8 ± 0.86% CD34-CD3+ cells, LV-FOXP3 vs LV-Ctrl), while the CD4 +/CD8+ cell ratio was significantly reduced (0.20 range: 0.13-0.97 vs 0.93 range: 0.25-2.9, LV-FOXP3 vs LV-Ctrl CD4+ /CD8+ ratio), confirming the predominant role of FOXP3 in CD4+ T cells as compared to CD8+ (Fig. 1H). Similar to the data obtained in liquid culture, we observed a tendency for higher percentages of primitive cells, marked as CD34+CD3-, at the end of the co-culture in LV-FOXP3 when compared to control cells (18 ± 3.1% vs 12 ± 3.9% CD34+CD3-cells, LV-FOXP3 vs LV-Ctrl) (Fig. 1H). These data indicate that constitutive expression of FOXP3 might interfere with self-renewal/ proliferation pathways in HSPCs and alter their differentiation into T cells.
Constitutive expression of FOXP3 enhances repopulating ability of HSPCs. We transplanted equivalent starting cell doses of LV-FOXP3 and control transduced bulk CD34+ cells into the NSG-based humanized mouse model (huMice) we established and assessed human engraftment by flow cytometry at 15-18 weeks by the GFP+ (transduced) and GFP-(not transduced) fractions (Fig. S2A). On the basis of the results obtained in vitro, we then assessed a possible advantage in repopulating potential of FOXP3 transduced over the non-transduced HSPCs. If FOXP3 was conferring a competitive repopulating advantage to HSPCs, we would expect an in vivo enrichment of the GFP+ fraction in LV-FOXP3 when compared to LV-Ctrl transplanted huMice. We thus calculated for each huMouse in vivo GFP+ enrichment as fold change between the percentage of GFP+ cells in different engrafted tissues and the relative original percentages of GFP+ cells in infused HSPCs, and found that LV-FOXP3 transplanted huMice displayed a significantly higher fold change compared to controls in bone marrow and spleen (fold change 0.90 range: 0.38-1. advantage of FOXP3-expressing HSPCs in repopulation, which was lost along the T cell lineage. Still, this analysis cannot discriminate between an effect of FOXP3 on the repopulating ability/lineage differentiation at the single cell level and an effect of FOXP3 on the number of repopulating cells in the HSPC population. To test the former hypothesis, we assessed the lineage differentiation and progenitor composition in periphery and bone marrow of LV-FOXP3 and LV-Ctrl transplanted huMice. We did not find major differences in LV-FOXP3 and LV-Ctrl huMice at the level of bone marrow composition and differentiation, beside the already mentioned trend towards reduced T cell differentiation, reflected in a significant accumulation of committed lymphoid progenitors in the bone marrow and lower percentages of CD3+ T cells in both bone marrow and periphery of LV-FOXP3 huMice (5.6 range: 3.7-6.4% vs 3.2 range: 1.4-5.3% CD34+CD38+CD10+CD45RA+ B-NK-T progenitors, LV-FOXP3 vs LV-Ctrl) (Fig. 2B,C). This result indicated that cell intrinsic repopulation and early differentiation was not affected by FOXP3 expression and suggested that FOXP3 rather affected the overall number of repopulating cells at the HSPC population level. In order to further test this latter hypothesis, we transplanted NSG mice with decreasing doses of transduced CD34+ cells and performed extreme limiting dilution analysis 15 . The fitted model  Table S1: logarithm of the fraction of non-repopulated mice (<1% CD45+GFP+ in the BM; Log fraction nonresponding) versus number of CD34+GFP+ cells transplanted per mouse (Dose); slope of the line: log-active cell fraction, dotted lines: 95% confidence interval, down-pointing triangle: cell dose with 0 non-repopulated mice. Bottom, confidence intervals for SRC frequency in the tested group (1/) and p value calculated by ELDA software (http://bioinf.wehi.edu.au/software/elda) are shown. showed highly different active cell fractions in the two groups, with a frequency of SCID Repopulating Cells (SRCs) in the LV-FOXP3 that was significantly higher (on average more than 4 folds) than the one in LV-Ctrl HSPCs (Fig. 2D, Table S1).
These data indicated that FOXP3 expression altered repopulating ability by increasing the numbers of repopulating cells in the HSPC population.
Ectopic FOXP3 expression hampers in vivo T cell differentiation and response. We then analysed the T cell compartment and observed an overall reduction in the T cell differentiation in LV-FOXP3 huMice compared to controls, with decreased percentages of mature thymocytes in both the vector carrying (GFP+) and not-carrying (GFP-) fractions (Fig. 3A). Accordingly, LV-FOXP3 huMice showed decreased percentages of peripheral CD4+ and CD8+ T cells compared to controls, while Treg frequencies were comparable (Fig. 3B,C). In order to verify if constitutive expression of FOXP3 affected the ability of T cells to respond to external cues, LV-FOXP3 and Ctrl huMice were orthotopically injected with breast cancer tumor cells MDA expressing human IL7, IL15 and GM-CSF, which has been previously shown to induce significant human T cells expansion in huMice 16 . By using this method, we tested in vivo T cell response to cytokines and ex vivo response to polyclonal stimulus by the in vivo expanded T cells. Three weeks post-implant, LV-FOXP3 huMice showed reduced peripheral T cell expansion (18 range: 11-47 vs 36 range: 9.9-111 folds increase in %CD3+, LV-FOXP3 vs LV-Ctrl) and smaller spleens (on average 52% and 74% of LV-Ctrl as cellularity and weight, respectively), as compared to Ctrl mice (Fig. 3D). Accordingly, when we assessed the presence of human cytokines in the serum of challenged huMice, we found inflammatory cytokines such as IL1β, IL5, CCL4, IL10 and IL17 significantly reduced in LV-FOXP3 compared to LV-Ctrl huMice (Fig. 3E). When we purified CD4 + cells from the spleen of challenged mice and stimulated them in vitro through the TCR, we observed reduced proliferation of LV-FOXP3 cells compared to Ctrl cells (0.31 ± 0.18 fold the proliferation index of LV-Ctrl on average) (Fig. 3F). Moreover, while the Ctrl CD4+ cells depleted of the Treg fraction (CD4+CD25−) proliferated more than total CD4+ cells as expected, CD4+CD25-LV-FOXP3 cells were still hypo-proliferative (0.20 ± 0.11 vs 1.1 ± 0.15 proliferation index, LV-FOXP3 vs LV-Ctrl) (Fig. 3F), indicating that FOXP3 overexpression intrinsically affected effector T cell proliferative capacity. Overall, these findings indicated that constitutive expression of FOXP3 in the hematopoietic system altered HSPC biology and reduced T cell differentiation and function.

FOXP3 expression affects the expression of key genes in HSPC.
In order to assess the molecular pathways affected by constitutive FOXP3 expression and underlying the observed outcome in HSPCs, we transduced CB CD34+ cells by LV-FOXP3 or LV-Ctrl, magnetically purified NGFR+ and NGFR-populations and assessed the expression of genes involved in maintenance of HSC primitive features and/or in controlling HSC niche. In particular, we assessed by dd-PCR analysis the tumor suppressor/cell cycle regulator p21/CDKN1A and the Transforming Growth Factor β1 (TGFB1) and found them significantly upregulated when FOXP3 was overexpressed compared to control HSPCs (FOXP3: 741 range: 29-1903 vs 0.001 range 0.00-1.3; p21/CDKN1A: 8.1 range: 4.9-16 vs 0.95 range: 0.33-1.2; TGFB1: 9.9 range: 1.6-53 vs 1.1 range: 0.81-1.4 arbitrary units FOXP3+ vs FOXP3-) (Fig. 4). This held true also for the gene encoding for the Matrix Metallopeptidase 9 (MMP9: 4.7 range: 0.28-56 vs 0.10 range: 0.01-0.34 arbitrary units FOXP3+ vs FOXP3-) (Fig. 4). KIT gene, encoding for the stem cell factor receptor, was not affected by FOXP3 expression (Fig. 4). Given the very well described role of p21 and TGFβ1 in controlling human HSC quiescence and self-renewal 17,18 , these data indicate that constitutive FOXP3 expression affects the main molecular pathways regulating HSC biology.

Discussion
HSPCs do not physiologically express FOXP3. However, by studying in this work the effect of constitutive FOXP3 expression in human HSPCs during their hematopoietic differentiation in huMice, we demonstrate that FOXP3 expression, unexpectedly, preserves HSC features by increasing quiescence and repopulating ability. Constitutive FOXP3 expression resulted in reduced in vitro proliferation and a prominent increase in the frequency of SRCs in the HSPC population, and upregulated p21/CDKN1A and TGFB1, well known factors controlling quiescence and self-renewal of HSC 17,18 , and MMP9 genes. Interestingly, both p21/CDKN1A and TGFB1 have been associated to FOXP3 activity in different cell types. FOXP3 has been shown to directly bind to the first intron of the CDKN1A gene and increase its expression 19 , and TGFβ1 has been described by several studies as mediator of the immune suppressive ability exerted by Treg cells 20 . Therefore, our data might call for a direct effect of FOXP3 on p21/CDKN1A and TGFB1 expression also in HSPCs and suggest that they might be the molecular basis of the observed phenotype. Furthermore, a role for MMP9 protein, which belongs to a family of proteolytic enzymes involved in the degradation of the extracellular matrix and associated to tissue remodelling, has been proposed in HSC mobilization, though never fully proven in the human system 21 . Its expression is directly controlled by TGFB1 signaling 22 and its proteolitic activity is required to activate latent TGFB1 itself 23 . MMP9 might thus be involved in the remodelling of HSC niche.
Our data thus indicate that, by affecting the molecular pathways controlling the main primitive features of HSC, ectopic FOXP3 expression improves the maintenance of more primitive HSC in the bulk HSPC population. These findings might open new avenues to improve human HSPC transplantation and gene manipulation.
in LV-FOXP3 (black bars) and LV-Ctrl (grey bars) huMice. (F) In vitro proliferation of sorted CD4+ and CD4+CD25-splenic cells upon anti-CD3/28/2 bead stimulus. Proliferation index is calculated as ratio between the percentage of proliferating cells in the depicted population and percentage of proliferating cells in tot CD4+ LV-Ctrl (dashed line) (n = 3). *p < 0.05 and **p < 0.01 by one-tail Mann Whitney test.
Indeed, induction of transient FOXP3 expression, by nucleofection of either FOXP3 RNA or artificial transcriptional activators to induce activation of the endogenous FOXP3 gene, might be envisaged to improve maintenance of repopulating human HSC. This may address an open need for many clinical applications, such as transplantation of CB-derived HSPC or gene edited HSPC 24,25 . The effect of FOXP3 expression on the number of SRC is less pronounced than the one obtained with the molecules nowadays approaching the clinics for HSPC expansion, such as SR1 and UM171 (4 fold for FOXP3 versus 17 and 13 fold for SR1 and UM171, respectively). Still, both approved compounds need a prolonged in vitro culture of HSPCs to be effective, as shown by the significant drop in the efficacy of SR1 when used for 12 instead of 21 days 26,27 . Instead, at least at the level of gene expression, FOXP3 effect is evident already at 5 days post-transduction. Even if only further experiment will clarify if transient expression will be as efficacious as constitutive in terms of SRC increase, we believe that the reduced culture time needed for ectopic FOXP3 expression might be an advantage compared to state-of-the-art compounds for the expansion of HSPCs, as it might reduce the risk of detrimental effects on HSC biology.
Ectopic FOXP3 expression as tool to specifically modulate HSPC features needs to be transient, since our data also show that, when FOXP3 expression persists, it impairs T cell differentiation and TCR-dependent/independent responses. These findings in the human system are consistent with the data obtained in FOXP3 transgenic mice 28 . Nevertheless, it is interesting to note that also GFP-cells (not carrying the vector and thus not expressing FOXP3) showed a similar, if not more pronounced, phenotype when compared to GFP+ (carrying the vector and expressing FOXP3) cells in huMice transplanted with LV-FOXP3 CD34+. This suggests that the fraction of FOXP3-expressing T cells affect the whole T cell compartment, possibly by altering the cytokine milieu. Consistently, LV-FOXP3 huMice displayed significantly reduced pro-inflammatory cytokines in the serum upon in vivo stimulation. These results clearly indicate that an HSC-based gene therapy approach for IPEX disease is not feasible by gene addiction but rather needs FOXP3 gene replacement preserving endogenous regulation of expression.
In conclusion, we discovered here a previously undescribed effect of FOXP3 on the HSPC compartment, which may unveil new pathways controlling HSC biology and could be exploited for clinical applications. Moreover, this work formally proves the need of maintaining endogenous FOXP3 gene regulation, such as by gene editing approaches or the use of lentiviral vectors containing endogenous FOXP3 regulatory regions, in HSC-based gene therapy for IPEX disease.

Methods
Cells and gene modification. Human cord blood HSPCs CD34+ were either purchased (Lonza) or purified by magnetic positive selection (Miltenyi Biotec) from healthy donors after informed consent approved by San Raffaele Ethics Committee and accordingly to Helsinki declaration. HSPCs were seeded and transduced in StemSpan medium (StemCell Technologies) supplemented with rhSCF, rhTPO, rhIL6 and rhFlt3-L (Peprotech) with 2 × 10 8 TU/ml (MOI 200) LV transducing units, as previously described 29 , after 24hrs of prestimulation. Self-inactivating lentiviral vector constructs (Fig. S1A) (Figs 2 and 3). Transduction, as assessed by reporter gene expression by flow cytometry, and expression, as assessed by Q-PCR on FOXP3 cDNA, efficiencies were comparable between the two sets of vectors and are shown as pooled in Fig. 1A,B.
Humanized mouse model. 2-5 days old NSG (NOD.Cg-Prkdc scid Il2rg tm1 WjI/SzJ, JAX mouse strain) mice were sublethally irradiated (1.5 cGy) and injected intrahepatically with 10 5 CD34+ 5-7 hours later. CD34+ cells were counted at time of seeding, before transduction, and control and test mice were transplanted with equivalent starting cell doses. Mice were kept in sterile conditions in ventilated cages in SPF animal house and received irradiated food and, for 1 month after irradiation, Gentamycin (0.3mg/ml) in the drinking water. Animals were sacrificed at 15-18 weeks, unless otherwise specified, and peripheral blood, bone marrow, thymus and spleen were harvested. Mice showing < 1% hCD45+GFP+ cells in the bone marrow were excluded from further analyses. Cytokine producing tumor xenograft was performed and analysed as previously described 16 . Briefly, 4 × 10 6 mammary carcinoma MDA3.1 cells (9/10 MDA-MB231 and 1/10 MDA-MB231 stably expressing human IL7, IL15 and GM-CSF) were implanted orthotopically in the mammary fat pad of huMice 15 weeks post CD34+ injection. Mouse health was monitored for three weeks, at the end of which huMice were sacrificed and lymphoid organs (see above) harvested. All protocols involving animals followed the Decreto Legislativo number 116 dated January 27th 1992 from the Italian Parliament and have been evaluated and approved by the San Raffaele Ethics Committee (IACUC protocol 488 and 632) as following the 3R principles.
In vitro assays. Liquid culture of CD34+ cells and CFC assay were previously described 32 . Briefly, CD34+ were cultured in StemSpan or IMDM 5% fetal bovine serum medium (StemCell Technologies) in presence of rhSCF (100ng/ml), rhIL6 (20ng/ml), rhFlt3-L (100ng/ml) and rhTPO (20ng/ml) (all from Peprotech) for liquid culture, while for CFC assay were plated at 10 3 cells/ml in semi-solid medium (MethoCult H4434 Classic -StemCell technologies) and scored by microscope 14 days later. Proliferation and mortality of liquid cultures were assessed at 3, 7, 11 and 14 days post-transduction. OP9-ΔL1 stromal cell line was kindly provided by I. Schmutz and M. Cavazzana and co-culture was performed as described for 21 days 14 . In vitro proliferation of T cells was assessed by cell proliferation dye efluor670 (eBioscience) staining of total splenocytes or magnetically purified CD4+ and CD4+CD25-cells, where indicated, upon stimulation by antiCD3/CD28/CD2 beads (Treg suppression inspector beads -Miltenyi Biotec) following manufacturer's instructions.
Antibodies purchased from BD Biosciences, eBiosciences and BioLegend. All acquisitions performed by FACSCanto II (Beckman Coulter) and analyzed by FlowJo Software.