Comprehensive Phenotypic Analysis of Diverse FOXN1 Variants

Capsule Summary

Impact of human FOXN1 variants on T cell development determined with transcriptional assays, protein localization studies, mouse models, and reaggregate thymus organ cultures. Variants with benign, partial-, complete- loss and gain- of function outcomes found.

Background

The thymus supports T cell development via two well-defined epithelial cell subsets, cortical and medullary thymus epithelial cells (TECs). TEC development and function is controlled by the transcription factor, Forkhead Box N1 (FOXN1), which positively regulates the expression of nearly 500 genes. The genes include chemokines, cytokines, keratins, Notch ligands and components of the thymoproteosome. Autosomal recessive FOXN1 mutations in humans result in a combined nude and severe combined immunodeficiency phenotype. However, individuals with single allelic or compound heterozygous FOXN1 mutations often have less severe phenotypes. With >400 FOXN1 variants now reported, the impact of many on protein function and thymopoiesis remain unclear. Using functional studies, we classified FOXN1 variants into benign, loss- or gain- of function, and dominant negatives. Several FOXN1 variants reduced the nuclear distribution of the protein, revealing a nuclear localization signal. Mice developed to genocopy the compound heterozygous FOXN1 mutations, identified in a patient, exhibited a transient thymus hypoplasia during embryogenesis that normalized post-natally. To screen the consequences of diverse FOXN1 variants on thymopoiesis, reaggregate thymus organ cultures combined with TAT-Foxn1 fusion protein transduction procedures were developed. These diverse experimental strategies aid in the identification of FOXN1 variants with clinical consequences.

Introduction

The thymus produces the T cells of the immune system, a process that begins during embryogenesis and continues throughout life¹. The magnitude of thymopoiesis is maximal around adolescence, followed by a progressive age-dependent decline in T cell output^{2, 3}. The principal cells needed for maintaining thymus integrity and support T cell development are thymic epithelial cells (TEC)⁴. Such cells are categorized into 2 major subsets, cortical and medullary TECs. These TECs express chemokines, growth factors, and cell surface ligands, including self-peptides embedded in major histocompatibility complex (MHC) molecules to support T cell development^{1, 4}. The presentation of self-peptide/self-MHC complexes by TECs provides selection cues for those T cells expressing T cell receptors recognizing self. This enables the T cells of the immune system to discriminate self from foreign antigens¹. The differentiation, expansion, and function of cortical and medullary TECs is governed by the transcription factor, Forkhead Box N1 (FOXN1)^{5, 6}. FOXN1 expression is induced following the initial specification of the thymus anlage within the 3^rd pharyngeal pouch during embryogenesis^{7, 8}. This occurs during the 7^th week of gestation in humans, which corresponds to day 11 in mouse embryonic development⁹. Upon expression, FOXN1 supports cTEC and mTEC sub-lineage development in both the fetal and postnatal thymus. This is dependent on Foxn1 induced expression of a suite of genes needed for TEC functions. Among these are DLL4, CCL25, Cathepsin L, CD40, PAX1, and HLA/MHC Class II^{4, 10, 11 10, 12, 13}. In addition to its role in the thymus, FOXN1 also maintains the structural rigidity of the hair shaft and keratinization of the nail plate^{14, 15}. There are two defined domains in FOXN1, a centrally located DNA binding region and a carboxy terminal transactivation segment (Fig. 1A)^{10, 16–18}.

Figure 1.

Human FOXN1 mutations with benign-, loss- and gain-of- function activities. A. Human FOXN1 is encoded on chromosome 17q11.2, with exons 2–9 coding for the protein. The OMIM allelic variants and those listed in ClinVar span all 8 coding exons (SNPs = single nucleotide polymorphisms; CNVs = copy number variants). The highest variant frequency, noted with black lines under mutations, exists in the Forkhead Box/DNA binding domain (blue box) followed by the transactivation domain (yellow box). In the figure, several key FOXN1 amino acids changes are listed below the 648 amino acid protein sequence. Among these are autosomal recessive mutations leading to Nude/SCID (in red) and distinct compound heterozygous mutations causing the indicated amino acid changes found in two unrelated individuals (Pt.1 in green; Pt.2 in orange). Several of the single allelic mutations that result in amino acid changes are black in color. B. To study the impact of the human FOXN1 mutations on protein function, selected variants were initially cloned into the murine Foxn1 cDNA. Wildtype Foxn1 or the various Foxn1 variants were transfected into HeLa cells (300 ng plasmid) in combination with a beta5t-luciferase transcriptional reporter construct and beta-galactosidase, the latter for normalization purposes. The levels of transcriptional activity were compared among the variants relative to wildtype Foxn1 (mFoxn1 Wt), shown with the dotted line. P values were determined using one-way ANOVA, with triplicate samples/variant tested, and subsequently verified in 3 independent experiments. C. The effects of the diverse Foxn1 mutations on protein size were determined by Western immunoblotting with anti-Foxn1 antibodies, specific for the amino terminal region. Blots were re-probed with antibodies specific for Gapdh, an endogenous protein as a control for protein loading with the transfectants. Blots are representative of 3 independent experiments.

Autosomal recessive (AR) FOXN1 mutations exist in humans, rats, mice, and cats, resulting in a nude/severe combined immunodeficiency (SCID) presentation^19–24. The nude phenotype, or alopecia universalis, results from the hair shaft folding over at the level of the sebaceous gland and breaking prior to extrusion through the dermal layer¹⁴. A nude presentation is also observed in patients with HOXC13 mutations, with this transcription factor positively regulating FOXN1 expression in the skin^{14, 25, 26}. AR FOXN1 mutations likewise impact the nails, with a layer of cornified/stratified squamous epithelium forming at the proximal end of the bed instead of the normal translucent layers¹⁴. FOXN1 autoregulates its own expression in the thymus^{10, 27}. The SCID phenotype arises because AR FOXN1 mutations lead to significantly reduced TEC numbers along with their impaired functionality, resulting in a block in T cell development. AR loss of function mutations in FOXN1 and the consequent T cell lymphopenia result in life-threatening viral, fungal, and bacterial infections^{23, 24, 28}. For such affected individuals, an allogenic thymus implant remains the only effective clinical treatment option^{29, 30}. Since the original description of 3 distinct AR FOXN1 mutations affecting just a few individuals, increasing numbers of patients with single allelic and compound heterozygous mutations are being reported^{23, 28, 31}. This increase is partly due to the widespread application of genomic sequencing to identify mutations in individuals identified with low T cell receptor excision circles (TRECs), which reveal reduced T cell output from the thymus^{32, 33}. ClinVar database lists 400 mutations in human FOXN1, which are often designated as of uncertain clinical significance, likely benign, or conflicting interpretations of pathogenicity.

We analyzed the impact of 35 distinct human FOXN1 variants on protein function and thymopoiesis, selecting representative mutations throughout the protein, including those forming truncations. Our functional assays included transcriptional reporter assays and protein localization studies. The findings uncovered FOXN1 variants with functional changes categorized into benign, partial- to complete- loss of activity, dominant negative consequences, and gain of function outcomes. Several amino acid substitutions impacted the nuclear distribution of FOXN1, revealing a nuclear localization sequence within the DNA binding domain. To define the functional consequences of specific FOXN1 mutants on thymopoiesis, we used mouse models to genocopy compound heterozygous variants identified in a patient. The mouse model revealed a transient thymus hypoplasia evident during embryogenesis. Additionally, reaggregate thymus organ cultures coupled with transduction procedures using TAT-Foxn1 fusion proteins were developed to more rapidly screen variants for their impact on thymopoiesis. Taken together, these approaches are enabling a precise determination of the functional consequences of assorted FOXN1 variants on T cell development.

METHODS

Study Approval

The Institutional Review Board at UT Southwestern Medical Center approved this study (IRB# 072010–009 and IRB# 112010–013). M.T.d.l.M, C.S., and C.A.W. acquired clinical information for several patients with FOXN1 variants (all in the U.S.A). Sequence information on de-identified patients were obtained under a protocol declared exempt by Duke University Health System Institutional Review Board. Additional patients with FOXN1 variants were obtained from published reports and/or the ClinVar database^{23, 28, 34}. Animal work described in this manuscript has been approved and conducted under the oversight of the UT Southwestern Institutional Animal Care and Use Committee (APN numbers 2015–101247 and 2015–101163). Mice were housed in a specific pathogen-free facility at UT Southwestern Medical Center. The Foxn1 targeted mice were developed entirely on a C57BL/6 background. While >40 founders screened for each construct, one independent line was expanded for each of the mutants generated by CRISPR/Cas; the Foxn1¹²⁸⁸ mutations (Foxn1-1288 #2; p.P430S) and Foxn1¹⁴⁶⁵ (Foxn1-1465 #43; p.Q489Rfs16). Foxn1⁹³³ #17 (p.D313Tfs) was previously described³¹. Mice used are listed in Supplemental Table 2. All the Foxn1 lines were crossed for 1–2 generations with C57BL/6 mice prior to intercrossing.

Antibodies, oligonucleotides, and plasmids

The antibodies used in the experiments, their sources and the plasmids used for transfections are listed in Supplemental Table 2. Oligonucleotide sequences used for genotyping, PCR reactions, qRT-PCR reactions, and sequencing are listed in Supplemental Table 3. Murine Foxn1 cDNA was cloned as previously described³¹. The EcoRI restriction site within the coding sequence of murine Foxn1 was eliminated by site-directed mutagenesis to facilitate subcloning into various vectors. Site-directed mutagenesis was performed to introduce human variants into the mouse Foxn1 cDNA. Human FOXN1 cDNA was obtained from Addgene (#153137) and subcloned into pCDNA3.1 using HindIII and XbaI restriction sites. For direct comparisons with the mouse Foxn1 constructs, the coding region of human FOXN1 was subcloned into the pCMV-FLAG vector, which was used originally used for murine Foxn1 (Cat. #635668, Clonetech Inc., now Thermo-Fisher). First, site-directed mutagenesis was used to mutate the internal BglII site in the human cDNA. Then primers were used to create new BglII and KpnI restriction sites at the 5’ and 3’ end of human FOXN1. The PCR product was digested and cloned into the BglII/KpnI sites in pCMV-FLAG. Site-directed mutagenesis was used to create the variants listed in the Table 1 and supplemental Table 1. Sanger sequencing was used to confirm the mutations in all the constructs. The pyruvate kinase gene includes the cDNA encoding a full-length pyruvate kinase construct with a Myc-tag at the 5’end. The pcDNAI-mycPKsec13 plasmid was kindly provided by B. Fontura (UTSWMC), with this plasmid sourced from Dr. M. Michael, University of California, San Diego, CA. The plasmid was digested with HindIII and ApaI, and the myc-PK-sec13 insert was subcloned into pcDNA3.1, digested with the same restriction enzymes. This enabled plasmid propagation in ampicillin selection media. The ApaI site was subsequently modified to incorporate a XhoI restriction site. With this, candidate Foxn1 cDNA sequences comprising a NLS were subcloned into the EcoRI/Xho I restriction sites within pcDNA3.1-mycPK plasmid to yield an in-frame extension of pyruvate kinase.

Table 1:

Human FOXN1 variants have diverse functional consequences

Patient Identifiers	DNA Sequence Variant	Protein Sequence Change	Luciferase Reporter Assaysa	Functional Impact	Protein Domain(s) Impactedb	ClinVar Designationc	Designation based current findingsd
Normal	NAe	None	100%	NA	NA
Du Study and current study
Pt.1f	c.933_936dupACCC	p.D313fs169g	0.9%	LOFh	DBD-TAD	Pathogenic	Pathogenic
	c.1089_1103del15	p.W363Cdel5aa	49%	LOF	DBD	Pathogenic	Pathogenic
Pt. 2f	c.1288C>T	p.P430S	92%	No impact	-	Uncertain Significance	Benign
	c.1465delC	p.Q489Rfs60	5%	LOF	TAD	Pathogenic/Likely pathogenic	Pathogenic
Pts. 3–4	c.1465delC	p.Q489Rfs60	5%	LOF	TAD	Pathogenic/Likely pathogenic	Pathogenic
Pt. 5	c.724C>T	p.P242S	85%		N-term	Uncertain significance	Benign
Pt. 6	c.958C>T	p.R320W	2%	LOF	DBD	Uncertain significance	Pathogenic
Pt. 7	c.962A>G	p.H321R	206%	GOFg	DBD	Uncertain significance	Pathogenic
Pt. 8	c.982T>C	p.C328R	57%	Partial LOF	DBD	No entry	Likely benign
Pt. 9	c.1075G>A	p.E359K	112%	GOF	DBD	Uncertain significance	Benign
Pt. 10-Pt. 14	c.1201_1206	p.P401del2aa	58%	Partial LOF	TAD	No entry	Likely Benign
Pt. 15	c.1293delC	p.P432fs118	0.4%	LOF	TAD	No entry	Pathogenic
Pt. 16	c.1418delC	p.P473Hfs77	1.5%	LOF	TAD	Pathogenic	Pathogenic
New Pt. 17	c.1567G>A	p.G523R	84%	No impact	TAD	No entry	Benign
New Pt. 18	c.1628G>A	p.G543E	107%	GOF	TAD	Uncertain significance	Benign
New Pt. 19	c.1364–1367del	p.Y455Cfs94	1.6%	LOF	TAD	Pathogenic/Likely pathogenic	Pathogenic
Bosticardo Study
P1	c.505G>A	p.E169K	78%	No impact	N-term	Uncertain significance	Benign
P2	c.907delG	p.E303Sfs247	1.2%	LOF	DBD-TAD	Uncertain significance	Pathogenic
P4	c.961C>A	p.H321N	5%	LOF	DBD	Uncertain significance	Pathogenic
P5	c.1009delG	p.G337Efs213	Not tested	Predict LOF	DBD-TAD	No entry	Not known
P6	c.1086_1087insA	p.W363Mfs118	Not tested	Predict LOF	DBD-TAD	No entry	Not known
P7	c.1135+5G>C	not reported	Not tested	Predict LOF		No entry	Not known
Giardino Study
P8	c.1168_1168delG	pD390Kfs160	Not tested		DBD-TAD	No entry
P9	c.1201_1216del16	p.P401Afs144	1.4%	LOF	TAD	Pathogenic/Likely pathogenic	Pathogenic
P14	c.1205delC	p.P402Lfs148	Not tested	Predict LOF	TAD	Pathogenic	Pathogenic
P15	c.1206delT	p.L404Cfs146	Not tested	Predict LOF	TAD	Uncertain significance	Pathogenic
P16	c.1315delC	p.L439Cfs111	Not tested	Predict LOF	TAD	Conflicting interpretations	Pathogenic
P18	c.1316delT	p.L439Rfs111	Not tested	Predict LOF	TAD	Pathogenic	Pathogenic
P19	c.1418delC	p.P473Hfs77	1.4%	LOF	TAD	Pathogenic	Pathogenic
P20	c.1420C>T	p.Q474*	Not tested	Predict LOF	TAD	Pathogenic	Pathogenic
P21	c.1584delC	p.L529Wfs21	12.5%	LOF	TAD	No entry	Pathogenic
P22	c.974T>C	p.L325P	1.5%	LOF	DBD	Uncertain significance	Pathogenic
P23	c.1392_1401del10	p.P465Rfs82	1.4%	LOF	TAD	Pathogenic	Pathogenic
P25	c.1850_1854del5	p.Y617Cfs157	32%	LOF	TAD	No entry	Pathogenic
P31-P47	c.763C>T	p.R255*	Not tested	LOF	DBD-TAD	Pathogenic	Pathogenic
P8f	c.340C>T	p.R114X	Not tested	LOF	N-term	Pathogenic	Pathogenic
	Not Reported	p.E139fs	Not tested	Predict LOF	N-term	No entry	Pathogenic
P9f	Not Reported	p.C82*	Not tested	Predict LOF	N-term	No entry	Pathogenic
	c.1049C>T	p.P350L	88%	No impact	DBD	Uncertain significance	Benign
P11	Not Reported	p.T527fs	Not tested	Predict LOF	TAD	No entry	Pathogenic
P12	Not Reported	p.V294I	18%	LOF	DBD	No entry	Pathogenic
P13-P18	c.1392–1401del10	p.P465Rfs82	1.4%	LOF	TAD	Pathogenic	Pathogenic
Rota Study
3 Pts	c.1370delA	p.H457Pfs93	0.6%	LOF	TAD	Pathogenic	Pathogenic
ClinVar
	c.205C>T	p.R69C	91%	No impact	N-term	Benign	Benign
	c.362C>T	p.A121V	89%	No impact	N-term	Benign	Benign
	c.689C>G	p.P230R	103%	No impact	N-term	Uncertain significance	Benign
	c.713G>A	p.G238D	105%	No impact	DBD	Benign	Benign
	c.1664C>T	p.A555V	75%	No impact	TAD	Benign	Likely benign
Invitae
	c.1911C>A	p.Tyr637*	72%	Partial LOF	C-term		Likely benign
Variants developed for functional studies
	c.1584delC	p.D528fs20	12%	LOF	TAD	N/A	Pathogenic
	c.1015–1016TC>GA	p.S339D	83%	No impact	DBD	N/A	Benign
	c.1465–1466delCA, c.1584–1585insC	p.Q489Gfs60	1.4%	LOF	TAD	N/A	Pathogenic
	Site-directed mutations	p.K316A/R320W	5.7%	LOF	DBD	N/A	Pathogenic
	Site-directed mutations	K316A/R320W/K327A	3.5%	LOF	DBD	N/A	Pathogenic
	Site-directed mutations	p.K331A/N334A/K335A	5.5%	LOF	DBD	N/A	Pathogenic
	Site-directed mutations	p.K327A/K331AN334A/K335A	2.1%	LOF	DBD	N/A	Pathogenic
	Site-directed mutations	p.K331A/N334A/K335A/W363Adel5aa	3.5%	LOF	DBD	N/A	Pathogenic

^aLuciferase reporter assays performed in the current study with beta5t using human FOXN1 variants

^bDBD = DNA binding domain; TAD = transactivation domain; N-term = Amino terminal region of protein

^cBased on autosomal recessive presentation

^dShaded box reveals updated designation from current findings using ClinVar terminology

^eNA = not applicable

^fYellow shaded box defines patients with compound heterozygous FOXN1 mutations

^gIn the murine cDNA, the ACCC insertion causes a D313Tfs12

^hLoss-of-function = LOF; Gain-of-function =GOF

Lymphocyte and epithelial cell preparations

Thymocytes and peripheral T cell populations were processed for flow cytometric analyses as published previously^{35, 36}. Thymocyte subsets and peripheral T cells were analyzed for the cell surface expression of various proteins including CD3, CD4, CD8, CD11b, CD11c, CD19, CD25, CD44, CD45, CD69, B220, NK1.1, αβ TCR, γδ TCR, and Ter119. Antibody sources are documented in Supplemental Table 2 . Experimentally, the lymphoid tissues were placed in FACS buffer (PBS with 2% FBS). Single cell suspensions were made by mashing the tissues through a stainless-steel mesh (100 Mesh, 0.0045 inch woven #316). The cell suspension was collected, washed, counted and aliquots stained with antibodies for 20 minutes. For the characterization of the stromal cell populations in the thymus, the tissue was digested Liberase^™ (Roche) containing DNase I (Roche) using procedures described³⁷. For embryonic tissues obtained between e12-e17.5, the lobes were separated and independently processed for thymocyte or TEC isolations. In the case of thymocyte preparations, single cell suspensions were prepared by mashing one lobe under a nylon membrane. Such cells were stained with antibodies detecting CD4 and CD8 and the other markers as described above. To characterize mesenchymal, endothelial, and TECs, the whole tissue or one of the two lobes, if sufficiently large, was processed in 100 μl of Liberase^™ (Roche) containing DNase I (Roche), with the digestion performed for 2–10 min. The single cells were stained with antibodies against CD45 (Tonbo Scientific), MHC II (I-A/I-E) (Tonbo Scientific), EpCAM (eBioscience), BP-1 (eBioscience), and UEA-1 (Vector Laboratories). Samples were analyzed on both FACSCaliber^™ and FACSCantoII^™ machines (BD Bioscience). FlowJo software (Tree Star Inc.) was used for data analysis. Thymic epithelial subsets were analyzed by electronically gating for CD45⁻EpCAM⁺ expression and selecting for UEA1⁺BP1⁻ to define medullary TECs or BP1⁺UEA1⁻ for cortical TECs. MHC class II high and low cells were used to discriminate between the subsets within cortical and two medullary TECs.

Immunohistochemistry and Western blotting

Thymic tissue was fixed in 4% paraformaldehyde (PFA) and used for imaging. H&E staining was performed under standard protocols and imaged on an Aziovert 200M inverted fluorescent microscope. For immunofluorescence staining, HeLa cells were grown on coverslips in 24-well plates followed by transfections. Forty-eight hours later, the cells were washed in PBS and then fixed with 2% PFA) for 10 minutes. These cells were permeabilized with 0.05% Triton X-100 and incubated with anti-mouse Foxn1 (1:100, Santa Cruz Biotech, CA) or anti-myc (9E11,1:100, Abcam Antibodies) along with Phyllandoin 488, all done at 4°C overnight. Cells were then washed with PBS and stained with donkey anti-mouse 594 and DAPI for 1 hour and mounted with ProLong^™ Gold anti-fade mountant (Thermo Fisher Scientific). Images were taken on a Leica confocal microscope and/or Keyence and EVOS microscopes. For quantification of nuclear and cytoplasmic occupancy, 7 –10 random fields were taken for each construct, in each field number of cells with cytoplasmic occupancy were counted and divided with total number of transfected cells. Cytoplasmic occupancy means when Foxn1 protein can be seen outside the nucleus as well. In case of WT the probability of Foxn1 being also present in cytoplasm is less than 1%. For western blotting, 5 × 10⁴ HeLa cells/well were plated in a 24 well plate. The cells were transfected with 500 ng to 1 μg of the various expression vectors containing either human or murine FOXN1 using Fugene HD reagent (Promega, Madison, WI). Twenty-four hours post-transfection, the cells were washed with PBS and lysed in RIPA buffer³⁸. In our protocols, RIPA comprised 50 mM HEPES, pH 7.4, 150 mM NaCl, 0.5% w/v sodium deoxycholate, 0.1% SDS. This was supplemented with protease inhibitors (aprotinin, leupeptin, benzamidine, PMSF) and 1U/ml benzonase (digests DNA). Proteins were resolved on SDS-PAGE gels and transferred onto PVDF membranes for Western blotting. Immunoblotting was done with the indicated antibodies followed by secondary antibodies linked to horseradish peroxidase, as listed in Supplemental Table 2³⁹.

Luciferase reporter assays

A luciferase reporter containing the promoter of Psmb11 gene was generously provided by Drs. S. Zuklys and G. Holländer (University of Oxford, Oxford, UK)¹⁰. The Psmb11 luciferase reporter construct (1 μg) was co-transfected into HeLa cells (2 × 10⁴ cells/well) along with the indicated expression containing murine or human FOXN1 wild type or mutants using 3 μl of Fugene HD reagent per transfectant (Promega, Madison, WI). A separate construct containing beta-galactosidase (0.1 μg), which is used as an internal control, was included in the transfections. Forty-eight hours post-transfection, the cells were washed in PBS, and 200 μl of Passive lysis buffer (1X) was added to the wells (Promega Incorp., Madison, WI). The plate was placed at −80°C for 30 minutes. The plate was thawed, and luciferase activity measured using luciferase assay kit (Promega). Luciferase activity was normalized to beta-galactosidase activity.

TAT-fusion protein production

The TAT-mFoxn1 wildtype expression vector was previously described⁴⁰. With this plasmid, the T7 sequence of the pET45b vector was modified by adding 4 nucleotides (GAGA) to improve protein expression⁴¹. Site-directed mutagenesis was used to create the c.961C>A nucleotide substitution, introducing the p.H321N amino acid substitution. The plasmids were transduced into BL21-DE3 pLysS bacteria. Colonies were grown O/N in Luria broth containing ampicillin (50 μg/ml). The O/N cultures were used to inoculate 500 ml of LB using a 1/100 dilution of bacteria. After 4 hrs. at 37°C, IPTG was added to 0.5 mM final, and the bacteria were cultured for an additional 5 hrs. at 30°C. The cells were pelleted, washed once in PBS, and the pellet frozen at −80°C. The pellet was thawed and resuspended in 45 ml of 50 mM Tris-Cl, pH8.0, pH 8.0, 1 mM 2-ME supplemented with the protease inhibitors, aprotinin, leupeptin, phenyl methyl sulfonyl fluoride and benzamidine. DNase was added at 10 μg/ml along with lysozyme at 20 μg/ml. The cell lysate was incubated at RT for 30 min, and subsequently processed with an Emulsiflex C5 high pressure homogenizer for 6–8 min at 600 psi, with samples kept on ice. TX-100 was added to a final concentration of 1% (v/v). The lysate was cleared by centrifugation at 7500 g for 30 min. The residual supernatant was applied to 4 ml of Ni²⁺-NTA beads according to the manufacturers’ instructions (Qiagen Corp., Germantown, MD). After washing with 20 columns of 50 mM NaH₂PO₄, 0.3 M NaCl, 20 mM Imidazole, the proteins were batch eluted sequentially with 3 column volumes of 0.75, 0.1, 0.2 and 0.3 M imidazole in the wash buffer. Aliquots of the samples were resolved by SDS-PAGE and Coomassie staining was used to visualize TAT-Foxn1 (Supplemental Fig. 5). A 90–95% pure source of TAT-Foxn1 was detected in the 0.15 and 0.2 M imidazole eluates. These fractions were pooled and concentrated with a Pierce protein concentrator and buffer exchanged with 50 mM NaH₂PO₄, 0.3 M NaCl, pH 8.0. The purified protein (0.1–3 μg) was added to the cultures.

Reaggregate thymus organ cultures (RTOC)

Normal and hypoplastic fetal thymic lobes were isolated between e12.0–12.5 or e13–13.5 gestational ages, as indicated in the figure legends. The tissues were collected in thymus organ culture media (TOC). TOC comprised RPMI containing 10% fetal calf serum, and supplemented with HEPES, l-glutamine, sodium pyruvate, penicillin, streptomycin, 5 × 10⁻⁵ M 2-mercaptoethanol and non-essential amino acids. Pooled or individual lobes were washed with PBS. Embryonic day13–13.5 lobes were processed as described⁴². For e12.-12.5 thymus lobes, the tissues were trypsinized for 1–2 min in 0.25% trypsin, 0.02% EDTA at 37°C. Digestions were stopped by washing in TOC media. The cells were resuspended in volumes <250 μl/20 lobes and an aliquot counted. The cells were stained with antibodies specific for TECs (EpCam-FITC) and mesenchymal cells (Pdgfra-PE). After washing, the cells were sorted into three populations; TECs, mesenchymal and a pool of the remaining cell types; early thymus progenitors (ETPs), endothelial cells, macrophages (EpCAM⁻Pdgfra⁻). Sorting was performed on an Aria Zelda FACS machine. RTOC was performed by reaggregating the 3 cell populations, EpCAM⁺ cell (~20–40%), Pdgfra⁺ (~30–50%) and EpCAM⁻ Pdgfra⁻ (~30%) in a 1.5 ml microfuge tube. In certain experiments, specific cell subsets were substituted from the hypoplastic lobes with those from controls. For TAT transduction assays, between 0.1–3 μg of pure protein was added to the TECs and incubated for 30 min prior to reaggregations. For reaggregating the cells, the pool of different cell subsets was centrifuged consecutively for 5’ and 10’ at 1000 rpm and 2000 rpm, respectively. After the 2^nd spin, the supernatant was aspirated until 2–4 μl of the aggregated cell pellet remained. This was placed on ice for 10 minutes. After gently dispersion, the mixture was drawn into a pulled glass pipette and placed as a single drop onto a Millipore nitrocellulose filter, sitting on top of a foam sponge (Gelfoam; 2 mm thick) in a 6-well tissue culture dish. The reaggregates were cultured for 10 days at 37°C in a 7.5% CO₂ environment. The cells were processed for flow cytometry using previously described procedures³.

Statistics.

Statistical analyses were performed using GraphPad Prism version 9.4.1 (GraphPad Inc.). Differences between 2 groups were evaluated using two-tailed Student’s t tests. A P value of less than 0.05 was considered statistically significant. Ns denotes not significant. For comparisons between multiple samples, one-Way ANOVA was used. Since certain Foxn1 mutant thymuses had significantly lower cell numbers, Brown-Forsythe and Welch tests were applied.

RESULTS

Functional characterization of diverse human FOXN1 variants reveals benign and loss and gain-of-function consequences

The clinical impact of FOXN1 mutations in patients include the classic Nude/SCID phenotype, SCID without alopecia, or a mild T cell lymphopenia that normalizes over time^{23, 31, 43}. An expanding number of FOXN1 variants are now reported, with 400 listed in ClinVar (Fig. 1A, Table 1). Presenting in patients as either autosomal recessive, single allelic or compound heterozygous variants, these FOXN1 mutations result in diverse amino acid substitutions, truncations, internal deletions and/or frameshifts in the protein (Fig. 1A). The first set of human FOXN1 variants studied corresponded to those in patients with low peripheral T cell counts, reported by the several groups listed (Table 1)^{23, 28, 31, 44}. We undertook a systematic approach to define the impact of these diverse FOXN1 variants on protein function with luciferase reporter assays. The human FOXN1 variants were initially introduced into the murine cDNA (designated as Foxn1) as each amino acid impacted was conserved between these species. Cloning into the murine cDNA was first done as we had also developed several mouse lines genocopying the human variants³¹. Expression vectors containing Foxn1 variants were transfected into HeLa cells along with the beta5t-luciferase reporter construct, a well-defined transcriptional target of Foxn1 (Fig. 1B)¹⁰. Constructs containing beta-galactosidase were included for normalization. Statistically significant functional differences were noted with the variants relative to the wild-type Foxn1 control, whose activity is denoted with the dotted line (Fig. 1B). Among these were complete- and partial- loss-of function variants (Fig. 1B, Table 1). The complete loss of activity was defined as <10% normal activity (Table 1). The loss of activity for several variants concurred wherein the mutation led to a reduced protein size, seen by Western blotting with anti-Foxn1 specific antibodies (Fig. 1C, upper panel; Supplemental Fig. 1B). The same membranes were also probed with anti-Gapdh antibodies, confirming similar protein extractions from the cells (Fig. 1C, lower panel; Supplemental Fig. 1B). Foxn1 p.D313Tfs12 had the lowest molecular mass of 34 kDa due to a frameshift causing a premature stop codon prior to the DNA binding domain (Fig. 1C). Foxn1 p.W363Cdel5aa, lacking a 5 amino acid stretch at the end of the DNA binding domain, had a molecular mass slightly less than the wild-type control (75 kDa) (Fig. 1C). Many FOXN1 variants had normal transcriptional activities (Fig. 1B, Table 1). In addition, two gain of function variants were found, FOXN1 p.H321R and p.E359K, defined as activities >110% wildtype values. Foxn1 p.H321R exhibited a 2-fold higher activity compared to the wildtype control (Fig. 1B, Table 2). For comparative purposes, we also tested several previously designated benign variants (p.R69C, p.A121V, p.230R, p.G238D, p.A555V) along with some listed as of “uncertain significance” (p.H321N, p.L325P, p.350L) in ClinVar (Table 1, Supplemental Fig. 1A). The results from such transcriptional reporter assays enabled the classification of numerous variants into either pathogenic or benign (Table 1 and Supplemental Table 1).

Human FOXN1 frameshift variants have a species-specific dominant negative function

Most FOXN1 mutations listed in ClinVar are monoallelic, with only 4 individuals to date reported with compound heterozygous mutations^{23, 28, 31}. A recent study identified 3 individuals with the same FOXN1 c.1370delA frameshift variant with a dominant negative functional effect⁴⁴. Interestingly, there are a significant number of frameshifts lised in ClinVar, some of which are listed in Table 1. The FOXN1 c.1370delA frameshift mutation creates a p.H457Pfs93 change, resulting in a de novo amino acid sequence of 92 amino acids ending with a premature stop codon at amino acid 550 (Supplemental Fig. 1A–B). To determine how different single allelic human FOXN1 frameshift variants impacted the function of the wildtype protein, the transcriptional assays were done with wildtype human FOXN1 tested in the presence of increasing amounts of the different variants. Wildtype human FOXN1 activity was suppressed to less than 5% normal levels when equivalent amounts of p.H457Pfs93 were expressed, consistent with a dominant negative consequence (Fig. 2A). In contrast, the same mutation created in murine Foxn1 only resulted in a 40% inhibition of wildtype Foxn1 activity (Fig. 2A). Of note, the murine c.1370delA was considered a partial dominant negative per a previous report⁴⁴. The species comparisons were repeated with the FOXN1 c.1465delC frameshift deletion (p.Q489Rfs60). Human FOXN1 p.Q489Rfs60 acted as a dominant negative against the wildtype, when transfected at a 1:1 ratio (Fig. 2A). Murine Foxn1 p.Q489Rfs16 only acted as a competitive inhibitor, suggesting a species selective difference (Fig. 2A).

Figure 2.

Human FOXN1 variants with defined frameshifts following the DNA binding domain form a species selective dominant negative activity. A. To assess and compare the impact of single allelic frameshift variants on the activity of both the human and mouse FOXN1 proteins, transcriptional reporter assays were performed in co-transfection experiments. Increasing amounts of the indicated variants were transfected alone or in combination with the wildtype construct (1:2, 1:1 and 2:1 ratios) along with fixed levels of the beta5t-luciferase reporter construct. The consequence of human and mouse FOXN1 H457Pfs93 and Q489Rfs60 variants on the wildtype protein were compared (human = black lettering; mouse = green). Luciferase-based transcriptional reporter assays were performed as in Fig. 1B. B. Several human FOXN1 variants with nucleotide deletions in or after the region encoding the DNA binding region were independently titrated against control FOXN1. Human FOXN1P401Afs144 and P473Hfs77 exhibited dominant negative activity, defined as <20% wildtype activity at a 1:1 expression ratio. Human FOXN1 D528fs20 and E303Sfs247 acted as competitive inhibitors, defined to be ~50% control activity when transfected at a 1:1 ratio. C. A distinct frameshift was engineered in human FOXN1 (c.1465–1466del/c1584–1585insC), forming a p.Q489fs60 amino acid change. This construct (Seq2DN) was tested for dominant negative functions against the wildtype control as in 2B. D. A cartoon diagram illustrates the amino acid sequence differences between the human FOXN1 frameshifts with/without dominant negative activity and the corresponding murine variants, none of which were dominant negative. Notably, equivalent nucleotide frameshift variants in the human and mouse FOXN1 resulted in distinct premature stop codons at amino acid 549 and 504, respectively. De novo amino acid sequences formed with the 1 or 2 nucleotide deletions were also distinct in the human versus mouse protein (Human = red or gray; Mouse = brown or green). Such sequence differences enabled a mapping of the critical regions needed for human FOXN1 dominant negative activity, indicated in the dotted boxed area. Human FOXN1 variants with dominant negative activity are indicated as DN in the figure. E. The amino acid sequence alterations in the different human and murine FOXN1 mutants were aligned by sequence homology. Human Q489Rfs60 and Q489Gfs60, created with a distinct frameshift and engineered with the same termination codon at aa 550, both had dominant negative activities. These two variants had distinct amino acid sequences after amino acid 489, indicated within the red dotted box. The dominant negative activity correlated with the presence of a truncated protein containing substitutions of several key Asp acid residues (Dx₄DxDx₃DxD) required for transactivation activity. For A-C, luciferase activities were normalized to beta-galactose and p-values were determined using one way ANOVA. The experiments are representative of at least 2 independent experiments/graph.

To define the basis of the species-specific dominant negative activity, an additional series of human FOXN1 frameshift variants were cloned and tested in the reporter assays. Among the variants tested were FOXN1 c.1201–1216del (p.P401Afs144), c.1293del (p.P432fs118), c.1364–1367del (p.Y455Cfs94), c.1392–1401del (p.P465Rfs82), c.1418del (p.P473Hfs77), c.1584delC (p.D528fs20) and c.907delG (p.E303Sfs247) (Fig. 2B, Supplemental Fig. 2A–D). Additionally, a premature stop codon (FOXN1 c.1138C>T, p.E379stop) was introduced immediately after DNA binding domain to determine the minimal length of protein needed for a dominant negative effect. All the human FOXN1 frameshift variants tested were dominant negative except for c.1584delC (p.D528fs20) and p.E379stop (Supplemental Fig. 2E). None of the equivalent murine Foxn1 variants were dominant negative. The human FOXN1 frameshift variants with dominant negative activity formed de novo amino acid sequences of varying lengths (Supplemental Fig. 2A). To determine if this de novo sequence was responsible for the dominant negative activity, a distinct frameshift was engineered in human FOXN1 to change the de novo coding sequence. This was done by introducing a 2-nucleotide deletion at c.1465–1466 and adding a new nucleotide insertion downstream to maintain the same stop codon position at amino acid 550 while changing the codon usage (Human FOXN1 p.Q489Gfs60) (Fig. 2C). Human FOXN1 p.Q489Gfs60 showed dominant negative activity, revealing that the de novo sequence itself was not causal to the phenotype. Consequently, various FOXN1 frameshifted constructs were aligned to characterize commonalities among the dominant negatives (Fig. 2D). This alignment revealed that the dominant negative effect required the expression of a 549 amino acid protein along with missense mutations at several Asp amino acid acids, previously shown essential for transactivation activity (Fig. 2E)¹⁸. The murine Foxn1 frameshifted variants were distinct as an earlier stop codon was formed at amino acid position 505. To assess whether the species-selective dominant negative function was related to the length of the frameshifted protein, a human FOXN1 c.1370delC (p.H457Pfs93) dominant negative construct was titrated against wildtype murine Foxn1. Human FOXN1 p.H457Pfs93 construct did not act as a dominant negative against murine Foxn1 (Supplemental Fig. 2F). Taken together, these results establish that multiple human FOXN1 frameshift variants form a species-specific dominant negative. This requires an intact DNA binding domain followed by a truncated protein that forms missense mutations at the key Asp residues responsible for transactivation activity. Such results suggest that any single allelic FOXN1 frameshift variant formed after the DNA binding domain and prior to the transactivation region can be considered a dominant negative.

Nuclear localization sequence exists within the DNA binding domain of FOXN1

Wildtype FOXN1 is entirely nuclear in location, and due to its molecular mass (75 kDa) can only enter by a nuclear localization signal (NLS). However, no reported NLS has been identified in FOXN1. Given the many FOXN1 missense variants with amino acid substitutions throughout the protein, we screened 19 different variants for those with potential effects on the nuclear distribution of the protein (Supplemental Fig. 3A–B). For this, HeLa cells were transfected with selected murine Foxn1 variants that matched the human nucleotide mutations. The nuclear versus cytosolic distribution of the Foxn1 variants were compared by immunohistochemistry (IHC). An antibody specific for the NH₂-terminal domain enabled an analysis of every variant. DAPI staining was used to delineate the nucleus while FITC-labeled Phalloidin visualized the actin cytoskeleton (cytosol). Four Foxn1 variants severely or slightly impacted nuclear localization; p.D313Tfs, p.R320W, p.L325P and p.C328R (Fig. 3A–B, Supplemental Figure 3A–B). All these variants were localized before or within the DNA binding domain (Fig. 1A). Foxn1 p.D313Tfs, which lacks the DNA binding and transactivation domains, was entirely cytosolic in all transfected cells. Foxn1 p.R320W, p.L325P and p.C328R had a cytosolic localization in smaller percentage of transfected cells (Fig. 3A–B). Foxn1 p.P430S, which lies outside the DNA binding domain, retained a nuclear location. The results suggest that a nuclear localization sequence (NLS) exists within the highly conserved DNA binding domain (Fig. 3C). To localize the sequence(s) mediating nuclear localization, different subregions of Foxn1 (aa 271–455, aa 271–378, aa 271–344, aa 271–378, aa 314–344, aa 331–395, and aa 415–497) were cloned, in-frame, to the 3’ end of the gene encoding pyruvate kinase (Fig. 3D, Supplemental Fig. 4A). Pyruvate kinase is a cytosolic protein that can redistribute to the nucleus if a NLS is added to its COOH-terminal end, such as that present in sec13⁴⁵. Pyruvate kinase-Sec13 was localized in the nucleus (Fig. 3E, Supplemental Fig. 4C). A pyruvate kinase fusion protein containing the Foxn1 region encompassing amino acids 271–455 also entered the nucleus (Supplemental Fig. 4). Consequently, shorter regions were tested, revealing that an amino acid stretch between 314–344 was sufficient to enable nuclear entry (Fig. 3E, Supplemental Fig. 4C). A pyruvate kinase fusion protein containing amino acids 331–395 had only partial entry while 414–497 of Foxn1 did not enter the nucleus, and instead exhibiting a cytosolic distribution (Fig. 3E, Supplemental Fig. 4). Western blotting with anti-myc antibodies confirmed expression of the various pyruvate kinase fusion proteins, with pyruvate kinase-Sec13 (aa 1–322) had a molecular mass of 75 kDa while those containing various sequences of Foxn1 had sizes ranging from ~48–55 kDa (Supplemental Fig. 4). Anti-Gapdh Western blotting confirmed similar protein loading (Supplemental Fig. 4). These findings confirm the existence of a NLS within the DNA binding domain of Foxn1. A NLS typically contains a cluster of basic amino acids (Arg, Lys) that enable nuclear entry via importins⁴⁶. Sequence alignments with human versus mouse FOXN1, which are 98% sequence identical in the DNA binding domain, and the amino acid stretch between 314–344 in human FOXN1 and its paralog, FOXN4, revealed several candidate basic amino acids (Fig. 3F–G). We substituted numerous charged amino acids with Ala residues at varying locations in human FOXN1 (Fig. 3H). Notably, these variants eliminated the functional activity of the protein in luciferase assays (Supplemental Fig. 4D). The impact of these substitutions on nuclear localization was determined by IHC. Combinatorial substitutions of what we had considered the strongest candidates for mediating nuclear entry had a minimal impact on the FOXN1’s location (p.K331A/N334A/K335A) (Fig. 3H). In contrast, human FOXN1 p.K316A/R320W/K327A protein had a higher percentage of cells with just a cytosolic distribution, confirmed by quantitation (Fig. 3H–I). Taken together, these findings reveal that the human FOXN1 NLS resides within the DNA binding domain, with at least 3 basic amino acids involved in nuclear localization.

Figure 3.

Identification of human FOXN1 variants that impact nuclear localization. A. Wildtype Foxn1 and the Foxn1 protein variants, p.D313Tfs, p.R320W, p.L325P and p.C328R, were expressed in HeLa cells. Forty-eight hrs. post-transfection, the cells were processed for immunohistochemistry with antibodies recognizing the terminal region of Foxn1 (revealed as red). DAPI staining was used to define the nucleus (blue) of the cell, while Phalloidin reveals the cytosol (green). Data are representative of 3 independent experiments. B. The percentage of cells with Foxn1 localized in the cytosol was compared among the indicated mutants relative to wildtype Foxn1, which is nuclear in all cells. C. The crystal structure of the Forkhead box domain of Foxn1 is shown interacting with DNA. The locations of R320, L325, C328R, and W363 are indicated. D. Cartoon diagram illustrating the various pyruvate kinase fusion proteins formed with Sec13, which contains a defined NLS, along with different segments of Foxn1 (aa 271–344, aa 314–344, aa 331–395, aa 414–497). E. Plasmids encoding different pyruvate kinase fusions were transfected into HeLa cells and processed for imaging using anti-myc antibodies along with DAPI and phalloidin. Data are representative of 3 independent experiments. F. The amino acid sequence homology between the DNA binding domain of human and murine FOXN1 is shown, with the boxed area indicating the amino acid region 314–344. G. Homology comparison between human FOXN1 and its paralog, FOXN4. H. Human FOXN1 expression vectors with either the wildtype cDNA or the indicated variants, formed by site directed mutagenesis, were transfected into HeLa cells, and processed as in (A). I. The percentage of cells with human FOXN1 localized in the cytosol was compared among the indicated mutants relative to wildtype human FOXN1, which is nuclear in all cells.

Selected compound heterozygous Foxn1 mutations reduce the efficiency of thymopoiesis during embryogenesis

We reported previously on two patients with compound heterozygous FOXN1 mutations who each presented with a T^−/loB⁺NK⁺ phenotype with normal hair extrusion (Table 1)³¹. Pt.1 had two allelic variants, forming distinct proteins, FOXN1 p.D313fs169 and p.W363Cdel5aa. Pt.2 had c.1288C>T and c.1465delC on different alleles, resulting in the formation of two proteins, p.P430S and p.Q489Rfs60, respectively (Fig. 4A). We compared the impact of these various compound heterozygous mutations on the transcriptional activities of the human and murine wildtype proteins. In the case of Pt.1, the 35% normal activity of human FOXN1 p.W363Cdel5aa was reduced ~50% when co-expressed at a 1:1 ratio with p.D313fs169 (Fig. 4B)³¹. For Pt.2, human FOXN1 p.Q489Rfs functioned as a dominant negative against p.P430S (Fig. 4B). Introducing the same mutations into the mouse cDNA revealed that the murine Foxn1 p.D313Tfs12 had a minimal impact on the p.W363Cdel5aa protein (Fig. 4C). Matching the variants in Pt.2, the murine Foxn1 p.Q489Rfs only acted as a competitive inhibitor against p.P430S (Fig. 4C).

Figure 4.

The phenotypic impact of distinct human FOXN1 compound heterozygous mutations on immune cells is phenocopied in mouse models. A. Pt.2 had compound heterozygous FOXN1 mutations, causing p.P430S and p.Q489Rfs60 amino acid sequence changes. B. The two distinct compound heterozygous FOXN1 mutations identified in Pt.1 and Pt.2 were compared in transcriptional reporter assays. Increasing amounts of one of the allelic variants was co-transfected against the second at 1:2, 1:1 and 2:1 ratios, along with the beta5t-luciferase and beta-galactosidase constructs. Luciferase activities were measured 48 hrs. post-transfection and results normalized with beta-galactosidase levels. C. The FOXN1 variants identified in Pt.1 and Pt.2 were introduced into the murine cDNA, and the consequences of these variants on murine Foxn1 transcriptional activity assayed as in B. For B and C, the transcriptional reporter assays were performed in triplicate. P-values were determined using one way ANOVA. Experiments shown are representative of 2 or more independent experiments. D. The compound heterozygous FOXN1 mutations identified in Pt.2 were independently introduced into the murine genome with CRISPR/Cas9 technologies. The presence of the c.1288C>T (murine p.P430S) and c.1465delC (murine p.Q489Rfs16) variants in the murine genome was confirmed by Sanger DNA sequencing. E. Representative pictures of 6–8-week-old littermates from wild-type Foxn1 (Foxn1^Wt/Wt), autosomal recessive Nude/SCID (Foxn1^{Q489Rfs/Q489Rfs}), single allelic (Foxn1^Wt/Q489Rfs), double allelic (Foxn1^P430S/P430S) and compound heterozygous mice (Foxn1^{P430S/Q489Rfs}), the latter genocopying Pt.2, are shown. F. The weights of the indicated Foxn1 variant mouse lines were determined using 6–8-week-old mice and compared with littermate controls. One way ANOVA was applied with at least 6 mice/genotype, revealing no statistical significance among the various lines.

We next assessed the impact of the Pt.2 FOXN1 mutations on thymopoiesis. For this, two new mouse lines were developed using CRISPR/Cas9 technologies (Fig. 4D). The mice (Foxn1^Wt/P430S and c.Foxn1^Wt/Q489Rfs) were intercrossed to obtain various single allelic and compound heterozygous Foxn1 lines along with littermate controls. The Foxn1^{P430S/Q489Rfs} mice that genocopied Pt.2 phenotypically resembled wildtype and heterozygous littermate controls, as evidenced by their normal body size along with the presence of fur and whiskers (Fig. 4E). Homozygous mutant mice with a complete loss of function variant (Foxn1^{Q489Rfs/Q489Rfs}) had a nude and small size phenotype (Fig. 44D–E)^{18, 31}. Foxn1^Wt/P430S and Foxn1^Wt/Q489Rfs mice were interbred to determine the impact of the compound heterozygous Foxn1 variants on thymopoiesis. In adult mice, thymuses from Foxn1^{P430S/Q489Rfs} compound heterozygous mice closely resembled littermate controls and those with single allelic mutations (Foxn1^Wt/P430S, Foxn1^Wt/Q489Rfs ), evidenced by similar thymus weights, cellularity, and comparable percentages of DN, DP, SP and DN1-DN4 thymocyte subsets (Supplemental Fig. 5A–E). Only the Foxn1^{Q489Rfs/Q489Rfs} nude mice had significantly reduced numbers and percentages of thymocyte subsets at the adult stages.

Pt.2 had a T^−/loB⁺NK⁺ phenotype within the first year of life. Moreover, many patients with single allelic FOXN1 mutations have low T cell numbers at birth that normalize over time²³. Given these observations, we considered the possibility that a thymus hypoplasia might be more evident during embryonic stages. This was previously revealed in the Foxn1^Wt/nu heterozygous mice relative to controls²³. To assess this, Foxn1^Wt/P430S and Foxn1^Wt/Q489Rfs mice were intercrossed, and thymuses isolated from embryonic day 16–16.5 (e16–16.5). Thymus cellularity along with DP thymocyte and TEC percentages were analyzed. Five different genotypes were compared with the diverse intercrosses, Foxn1^Wt/Wt, Foxn1^Wt/P430S, Foxn1^Wt/Q489Rfs, Foxn1^{Q489Rfs/Q489Rfs} and Foxn1^{P430S/Q489Rfs}. Thymuses from the compound heterozygous mice matching Pt.2 (Foxn1^{P430S/Q489Rfs}) had a slightly lower cell number compared to the single allelic variants, including a reduced DP percentage, and a slightly higher percentage of TECs (EpCam⁺) (Fig. 5A–C). The mild phenotypes in the mice genocopying Pt.2 were in sharp contrast to the severe block in T cell development noted with the autosomal recessive Foxn1^{Q489Rfs/Q489Rfs} line (Fig. 5A–C). For example, the Foxn1^{Q489Rfs/Q489Rfs} autosomal recessive mice had a 100-fold reduction in cellularity, with no DP cells evident (Fig. 5A–C). In the Foxn1^{P430S/Q489Rfs} embryonic thymuses, the higher percentage of TECs was consistent with a delayed or reduced expansion of the DP thymocytes. Some of the statistically significant different suggested that the mice genocopying Pt.2 could be more severe at earlier developmental stages. Consequently, we next examined thymopoiesis at e13–13.5. The thymuses from the Foxn1^{P430S/Q489Rfs} embryos were significantly smaller at this early developmental stage (Fig. 5D). Two representative examples are shown (Fig. 5D, e.g., 1 and 2). To identify the differences among the embryonic thymuses, the percentage of early thymus progenitors (ETPs) and mesenchymal and thymic epithelial cells was determined following flow cytometry using antibodies specific for the cell surface proteins CD117, CD45 (ETPs), Pdgfra (Mesenchymal cells) and EpCam (TECs) (Fig. 5E–F). Comparing the numbers of cells and the percentages of the cell subsets revealed that the Foxn1^{P430S/Q489Rfs} had statistically significant reductions in the total cell number and percentage of ETPs (Fig. 5G). The percentage of TECs and mesenchymal cells was not significantly different, implying a functional defect in the TECs (Fig. 5G). Taken together, these findings suggest the compound heterozygous Foxn1^{P430S/Q489Rfs} mutations cause a transient hypoplasia that corrects post-natally, consistent with the T^loB⁺NK⁺ presentation revealed in many patients with single allelic FOXN1 variants, including those with dominant negative activities²³.

Figure 5.

Mice with compound heterozygous Foxn1^{P320S/Q489Rfs} mutations that genocopy Pt.2 have a transient thymic hypoplasia during embryogenesis. A-C. Embryonic day 16.5 (E16.5) thymuses were obtained from the indicated Foxn1 wildtype (Wt) or variant embryos, with genotypes determined by PCR. The number and percent of the different cell subsets in the tissues was determined by cell counting and flow cytometric analyses following cell surface staining with mAbs detecting CD4, CD8, EpCam (TECs) and CD45 (hematopoietic cells). Of note, the Foxn1^{Q489Rfs/Q489Rfs} embryos have a Nude/SCID phenotype, evident 4–5 days post-natally. A. The percentage of double position (CD4⁺CD8⁺) thymocytes was calculated following electronic gating of the boxed area. B. Epithelial (EpCam^+CD45⁻) and hematopoietic cell percentages cells (CD45⁺) were calculated by gating the selected subsets of cells. C. The total cell number and percentages of DP thymocytes and thymic epithelial cells (TECs) were determined using a minimum of 4 mice/group. Wildtype controls were always derived from littermates. Statistically significant differences were established by one-way ANOVA (Brown-Forsythe and Welch tests). D. The cardiothoracic regions of embryonic day 13–13.5 (e13.5) embryos were accessed to visualize the pharyngeal region, with images taken. The thymic regions were demarcated with dotted lines. Two representative images of embryos from the Foxn1^{P430S/Q489Rfs} (e.g., 1 and 2) were provided along with Foxn1^Wt/Wt wildtype and Foxn1^Wt/P320S controls. Solid yellow line = 1 mm. E-F. Thymic tissues from e13–13.5 embryos were processed into single cell suspensions. The cell populations were compared by flow cytometric following mAb staining for CD117 and CD45 or Pdgfra and EpCam to identify (E) early thymic progenitors (CD117⁺CD45⁺) and (F) mesenchymal cells (Pdgfra⁺) and TECs (EpCam⁺). G. The total number of cells present in the indicated e13.5 thymuses was enumerated. In addition, the percentage of ETPs, mesenchymal cells (Mes) and TECs was determined using a minimum of 5 thymuses per genotype. Statistically significant differences were established by one-way ANOVA (Brown-Forsythe and Welch tests).

Thymopoiesis in autosomal recessive FOXN1 mutant embryonic thymuses assays is restored with a TAT-Foxn1 wildtype fusion protein

Assessing the impact of FOXN1 mutations on thymopoiesis currently necessitates the development of mouse knock-in lines, which can take 1–2 years to characterize (Fig. 4). To develop a more rapid screening system to define the impact of the FOXN1 variants on thymopoiesis, we capitalized on our success with fetal thymus organ cultures (FTOC) and reaggregated thymus organ cultures (RTOC)⁴². At e11.5 of embryogenesis (week 7 in humans), the thymus anlage is formed and Foxn1 expression is initiated⁷. Foxn1 expression is not needed to reach this stage of development⁸. However, as the thymus remains too under-developed to isolate, we used e12–12.5 fetal thymuses. To eliminate endogenous Foxn1 functions, Foxn1^Wt/T313fs mice were intercrossed to obtain embryos with varying genotypes, including those with autosomal recessive Foxn1 mutations (Foxn1^{D313Tfs/D313Tfs}). At e12–12.5, thymuses from the Foxn1 mutant embryos were much smaller that littermate controls (Fig. 6A, dotted area). To develop RTOC procedures that could be used with the hypoplastic tissues, we selected normal thymuses as a source of different cell subsets. Control thymuses from e12–12.5 were processed into single cell suspensions and stained with mAbs detecting TECs (EpCam⁺) and mesenchymal cells (Pdgfra⁺). This enabled a flow sorting of cells into 3 distinct cell subgroups, mesenchymal (Group I), TECs (Group 2) and a remaining pool of early thymus progenitors (ETPs), endothelial cells, and macrophages (Group III) (Fig. 6B). Reaggregate thymus organ cultures were prepared using these cell subsets along with cells prepared from pooled hypoplastic thymuses (Fig. 6C)⁴⁷. Cells from normal and hypoplastic thymuses were reaggregated, overlayed onto membranes, and cultured for 10-days to monitor thymocyte growth and T cell differentiation (Fig. 6C). Reaggregates of control embryonic thymuses expanded in size and supported CD4⁺CD8⁺ T cell development (Fig. 6D). Equivalent numbers of reaggregated cells using Foxn1^{D313Tfs/D313Tfs} embryonic thymuses failed to expand and CD4⁺CD8⁺ thymocytes development did not occur (Fig. 6D). To correct for the functional deficiency of Foxn1 in Foxn1^{D313Tfs/D313Tfs} reaggregates, a TAT-FOXN1 fusion protein was purified from bacteria and added to the cells (Supplemental Fig. 6A–B). The TAT domain enables Foxn1 entry into TECs, first revealed by intrathymic injections of the protein⁴⁰. We confirmed that pure TAT-Foxn1 wildtype protein entered primary embryonic e12–12.5 TECs by Western immunoblotting with anti-Foxn1 mAbs using lysates from ~20,000 transduced cells (Supplemental Fig. 6C–D). Since our prior experiments suggested that the Foxn1-deficient thymuses had limited numbers of ETPs, reaggregates of the Foxn1^{T313fs/T313fs} embryonic thymuses were supplemented with Group III cells (with ETPs) from normal tissues. In such reaggregates, the TAT-Foxn1 wildtype protein supported thymus growth and CD4⁺CD8⁺ T cell development (Fig. 6D). In one RTOC culture, the TAT-Foxn1 wildtype protein enabled equivalent cell expansion as the normal controls (Fig. 6E). This was not always seen, an issue we suggest is due to the technical challenges of aggregating very small numbers of cells. Also noteworthy, the addition of Group III cells alone supported some growth for the Foxn1^{D313Tfs/D313Tfs} reaggregates (Fig. 6D). We hypothesize this is facilitated by the presence of endothelial cells that retain TEC progenitor potential. In future directions, we will use primary TECs from Foxn1^{D313Tfs/D313Tfs} embryonic thymuses that are expanded in culture in the presence/absence of TAT-Foxn1 followed by reaggregate cultures with normal ETPs.

Figure 6.

Thymopoiesis restored in Foxn1-mutant embryonic tissues supplemented with purified TAT-Foxn1 fusion proteins. A. Foxn1^Wt/T313fs heterozygous mice were intercrossed, and embryos isolated from e12-e12.5 timed pregnancies. The cardiothoracic region was accessed to visualize the thymic lobes, demarcated with dotted lines. Solid yellow line = 1 mm. B. Single cell preparations from e12–12.5 embryonic thymuses (Foxn1^Wt/Wt) were stained with mAbs versus Pdgfra (mesenchymal cells) and EpCam (TECs). The cells were sorted into 3 categories: Group I (mesenchymal), Group II (ETPs, endothelial, others) and Group III (TECs) for us in reaggregate thymus organ cultures (RTOCs). C. Cartoon depiction of RTOC procedures. Flow sorted Group 2 cells were obtained from normal (Foxn1^Wt/Wt) embryonic thymuses to acquire ETPs, which were deficient in the small thymuses from Foxn1^{T313fs/T313fs} embryos. In addition, single cell suspensions of pooled Foxn1^{T313fs/T313fs} embryonic thymuses were prepared and incubated with either media alone or a purified TAT-Foxn1 wildtype protein. These were re-aggregated with control ETPs to obtain a minimum of ~13,000 total cells containing sufficient ETPs to develop into thymocytes. Reaggregates were placed onto membranes layered on top of gel foam pads. The RTOCs were cultured for 10-days. D. RTOC assays were undertaken with embryos from Foxn1^Wt/Wt or Foxn1^Wt/T313fs intercrosses. The first column corresponds to control RTOCs from normal sized thymus lobes. The 2^nd column corresponds to reaggregates using equivalent cell numbers from Foxn1^{D313Tfs/D313Tfs} embryonic thymuses. The 3^rd and 4^th columns represent reaggregates of cells from Foxn1^{D313Tfs/D313Tfs} embryonic thymuses supplemented with ETPs without and with TAT-Foxn1 Wt protein. Yellow bar = 1 mm. E. The total cell number and percent of DP cells was determined from the indicated RTOCs after 10-days of culture by cell counting and flow cytometric analyses following cell surface staining with mAbs detecting CD4 and CD8. Statistically significant differences were established by one-way ANOVA (Brown-Forsythe and Welch tests).

Our goal was to define the impact of selected TAT-Foxn1 variants on thymopoiesis. To address this, we next assessed whether a Foxn1 variant could antagonize normal RTOC growth. For this purpose, the p.H321N mutation was engineered into the TAT-Foxn1 wildtype protein (Supplemental Fig. 6A–B). RTOCs of normal e12–12.5 thymuses cell subsets were cultured either in media alone, with the TAT-Foxn1 wildtype protein, or with TAT-Foxn1 H321N (Supplemental Fig. 7). In such RTOCs, TAT-Foxn1 H321N variants had variable suppressive effects on thymus growth (Supplemental Fig. 7C). This variability likely depends on the variant selected, as Foxn1 p.H321N does not bind to DNA as effectively as the wildtype protein¹⁶. This suggests that TAT-Foxn1 variants may be more effectively screened for partial or complete restoration of organoid growth with the Foxn1^{T313fs/T313fs} hypoplastic thymuses. Overall, our proof of concept experiments reveal a novel methodology to evaluate the impact of human FOXN1 variants on thymopoiesis. Such RTOC assays involve a 10-day culture period, which will enable a rapid screen for thymopoietic impacts of Foxn1. The development of primary TEC cultures and bio banked ETPs will likely make this technology more widely available for investigators.

DISCUSSION

Several human diseases result in a thymus hypoplasia due to genetic mutations that impact the stromal cell populations^{3, 42, 48}. Humans with autosomal recessive, single allelic and compound heterozygous FOXN1 mutations can have impaired TEC functions, distinct from the mesenchymal cell problems occurring in 22q11.2 deletion syndrome patients (DiGeorge)^{20, 24, 42}. There are ~400 human FOXN1 variants listed in the ClinVar database, with the clinical impacts of these variants designated as benign, likely benign, uncertain significance, pathogenic, or conflicting interpretations of pathogenicity. To improve on our understanding of how these varied mutations affect protein functions, we characterized 35 of the FOXN1 variants using transcriptional reporter assays, revealing those with benign, partial- and complete- loss and gain- of function activities (Tables 1, Supplemental Table 1).

Interpreting the consequences of the diverse FOXN1 variants on human thymopoiesis necessitates several considerations. First, knowledge of the allelic presentation is required. For autosomal recessive FOXN1 variants, the functional impact directly correlates with the level of activity revealed in transcriptional reporter assays (Fig. 1). For single allelic variants, the functional consequence depends on how a particular variant influences the wildtype allele. For example, human FOXN1 frameshift variants resulting from nucleotide deletions after the DNA binding domain and prior to the transactivation domain form dominant negatives (Fig. 2, Supplemental Fig. 2). Variants fitting this category include FOXN1 c.1201–1216del, c.1205delC, c.1206delT, c.1293del, c.1315delC, c.1364–1367del, c.1370delA, c.1392–1401del, c.1418delC, c.1465delC, along with the others listed in the ClinVar database containing single nucleotide deletions between c.1135-c.1538 (Table 1). Noteworthy, the dominant negative effect is human specific as similar frameshifts created in the murine cDNA only act as competitive inhibitors. A key feature of the human dominant negative activity is the elimination of the acidic amino acids needed for transactivation functions. The dominant negative activity of human FOXN1 is not completely understood but may relate to protein dimerization and/or protein aggregation changes that alter the nuclear distribution of the wildtype protein⁴⁴. Additionally, nucleotide deletions prior to the region encoding DNA binding domain are likely benign when affecting only one allele. This is because the protein that is formed would lack the DNA binding domain and lack the capacity to enter the nucleus.

Among the diverse hFOXN1 variants characterized were several that reduced the nuclear localization of the protein (Fig. 3; p.R320W, p.L325P and p.C328R). This enabled us to identify a NLS within the highly conserved DNA binding domain (aa 314–344). Using homology searches with diverse NLS sequences, a strong sequence homology was evident between amino acids 331–344 of FOXN1 with the corresponding region in FOXN4, a paralog. In addition, some sequence similarity with the NLS of Pho4 was found^{46, 49}. Our studies also revealed gain of function consequences for two unrelated FOXN1 variants (FOXN1 p.H321R, p.E359K) revealed. FOXN1 p.H321R had a 2-fold increase in activity. We suggest introduction of a positively charged Arg residue may have increased the DNA binding affinity.

For patients with single allelic variant or compound heterozygous mutations, co-titration experiments with luciferase reporter assays provides an excellent experimental strategy to define how one allelic variant impacts either the wildtype FOXN1 or second allelic variant. For example, any FOXN1 variants that have frameshifts or stop codons eliminating the DNA binding domain (e.g., FOXN1 p.D313fs) or missense mutations impacting amino acids required for DNA binding (e.g., FOXN1 p.R320W) have weak impacts on the transcriptional activity of the normal allele¹⁶. In vivo, the impact of such single allelic FOXN1 variants may be even less severe if the mRNA undergoes non-sense mediated decay. This is difficult to assess in transfected cells as the expression vectors enforce mRNA over-expression, which may override normal decay processes. It is also important to note that most TEC cell lines lose expression of the endogenous FOXN1, suggesting that co-transfection experiments still remain more informative. The best approach for testing the RNA stability is to generate knock-in mice and/or create cell lines wherein the mutation is introduced into the genome. We have used this technology to characterize two distinct Foxn1 compound heterozygous mice genocopying Pt.1 and Pt.2, respectively. In our previous report, the Foxn1 mutant mice genocopying Pt. 1 matched the T⁻B⁺NK⁺ phenotype with normal hair³¹. In the case of Pt.2, the resulting compound heterozygous Foxn1 mutations introduced in mice (Foxn1^{P320S/Q489Rfs}) resulted in a relatively mild and transient hypoplasia of the thymus that was very pronounced during embryonic stages. Per clinical information, Pt.2 had low TRECs at birth, and consequently, a peripheral T cell lymphopenia (~100 CD3⁺ T cells/μl)³¹. The species selective differences we described between human and mouse FOXN1 may partly explain the less severe impact of the Pt.2 mutations genocopied in mice. First, murine Foxn1 is more active transcriptionally than human FOXN1, even when both cDNAs have been engineered with the same position of the FLAG epitope and their expression driven by the same expression vector (Supplemental Fig. 2G). Second, the human FOXN1 Q489Rfs in Pt.2 is a dominant negative, attenuating the function of the wildtype allele. It does not function as a dominant negative in mice nor in transcriptional reporter assays. Third, human FOXN1 can form dimers to mediate transcriptional functions⁴⁴. It remains unclear if murine Foxn1 forms such dimers. Finally, we speculate that epigenetic modifications in some TEC cells in the expanding post-natal thymus in human could undergo clonal selection since higher expression of the wildtype FOXN1 allele may provide a growth advantage. This could explain the improved T cell numbers that occurs over time for some patients with single allelic FOXN1 variants²³. Such increased T cell output should preclude an early decision for a thymus transplant in a patient, since a delay permits more careful monitoring of T cell recovery^{23, 48, 50}.

In the current manuscript, a new experimental strategy was developed to more rapidly define the impact of selected FOXN1 variants on thymopoiesis. Reaggregate thymus organ cultures were modified by using embryonic thymuses lacking a functional Foxn1 gene (Foxn1^{D313Tfs/D313Tfs}) as a source of Foxn1-deficient TECs. Supplementing these RTOCs with purified TAT-Foxn1 fusion proteins revealed restoration of thymopoiesis within a 10-day culture. This strategy also required the addition of normal ETPs, as these cell were limiting in the hypoplastic thymuses, even when the latter were pooled to achieve cell numbers equivalent to controls. This need is consistent with the Foxn1^{D313Tfs/D313Tfs} TECs expressing reduced chemokines, particularly Ccl25, which recruits ETPs³¹. Some additional modifications we are testing includes biobanking the various cell subsets needed for RTOC assays. Expanding the TECs in defined culture media prior to the biobanking would override the extreme cell number limitation currently faced with the Foxn1 mutant TECs. Such changes would enable high-thru put screening strategies to define the impact of the many FOXN1 variants not yet studied.

In summary, a combination of in vitro functional assays, imaging analyses, mouse lines genocopying patients, and reaggregate thymus organ cultures establish the impact of human FOXN1 variants on thymopoiesis. These approaches have yielded key insights into the impact of single allelic, compound heterozygous and autosomal recessive FOXN1 mutations on T cell development in humans. Such information will help select appropriate clinical interventions for patients with diverse thymic stromal cell deficiencies leading to in-born errors of immunity.

Supplementary Material

Supplemental Table 2

Mouse strains, biological reagents, chemical reagents, plasmids

Supplemental Table 1

Anticipated functional consequences of human FOXN1 mutations in patients

Supplemental Table 3

Oligonucleotides used for Foxn1 screening and site-directed mutagenesis

Supplemental Figure 1

Human and murine FOXN1 variants have distinct functional consequences. A. The functional activity of selected human FOXN1 variants, designated as benign or unknown consequence in the ClinVar database were compared using transcriptional reporter assays. FOXN1 wildtype or variant constructs were transfected into HeLa cells (300 ng plasmid) in combination with a beta5t-luciferase transcriptional reporter construct and beta-galactosidase. Transcriptional activity was compared following transfection efficiency normalizations with beta galactosidase. P values were determined using one-way ANOVA, with triplicate samples/group tested and verified in 3 independent experiments. B. The impact of the FOXN1 variants on protein size was assessed by Western immunoblotting with anti-FLAG antibodies. Blots were re-probed with antibodies specific for Gapdh, an endogenous protein to reveal protein loading levels. Blots are representative of 2 independent experiments. C. Human wildtype and c.1465delC (p.Q489Rfs60) variants were subcloned into the identical mammalian expression vector as that with the murine equivalents. Transcriptional reporter assays performed as in (A) using 300 ng of each construct, either alone, or in a 1:1 ratio of control and variant constructs, as indicated. P values were determined using one-way ANOVA, with triplicate samples/group tested and verified in 3 independent experiments. D. The constructs shown in C were transfected (500 ng) into HeLa cells, and lysates generated for Western immunoblotting with anti-FLAG followed by anti-gapdh mAbs. Blotting was replicated in 3 independent experiments.

Supplemental Figure 2

Selected human FOXN1 variants exhibit dominant negative functions. A. Cartoon diagram depicting the human FOXN1 variants with similar frameshifted stop codons at amino acid 550. The red line denotes de novo amino acid sequences formed after the nucleotide deletion in the cDNA. Only p.E303Sfs246 does not have DN activity. B. Selected human FOXN1 variants along with the wildtype construct were transfected into HeLa cells for Western blotting. Proteins in the cell lysates were processed for Western immunoblotting to detect FOXN1 followed by blotting for Gapdh. Indicated kDa markers are shown. C. Luciferase reporter assays were undertaken with empty vector, wildtype human FOXN1 and the indicated variants, alone or in combination at 1:2, 1:1 and 2:1 plasmid ratios. D. Various paired human FOXN1 and murine Foxn1 constructs were compared by Western immunoblotting with anti-Foxn1 followed by anti-gapdh mAbs. E. The human and murine FOXN1 p.D528fs20 and p.E379stop construct were compared in transcriptional reporter assays as in C. F. A dominant negative human FOXN1 variant c.1370delA (p.H457Pfs93) was transfected alone or in conjunction with murine Foxn1 Wt at 2:1, 1:1 and 1:2 plasmid ratios and tested with the transcriptional reporter assay as in C. In C., E., and F., luciferase data were representative of triplicate samples/group, with experiments repeated in 3 independent assays. P values were determined using One-way ANOVA.

Supplemental Figure 3

Characterization of Foxn1 variants to identify those that impact the nuclear distribution of the protein. A. Diagram of different hFOXN1 variants, cloned into the murine cDNA, is shown with the location of the amino acid change indicated. B. Wildtype Foxn1 and various Foxn1 variants were expressed in HeLa cells and processed for immunohistochemistry with antibodies recognizing the terminal region of Foxn1 (revealed as red). DAPI staining was used to define the nucleus (blue) of the cell. Cells were analyzed with an EVOS cell imaging platform. C. Human FOXN1 expression vectors with either the Wt cDNA or the variants formed by site directed mutagenesis were transfected into HeLa cells and processed as in B.

Supplemental Figure 4

Mapping of the NLS of FOXN1 using pyruvate kinase fusion protein. A. Cartoon diagram illustrating the various pyruvate kinase fusion proteins formed with Sec13, which contains a defined NLS, along with different segments of Foxn1 (aa 271–455, aa 271–378) and two where amino acids substitutions were engineered (aa 271–279 R320W; aa 271–378 RK341GW). B. Plasmids encoding the different pyruvate kinase fusions (from Figure 3E) and those depicted in A were transfected into HeLa cells. The samples subsequently processed for Western blotting using antibodies recognizing the myc-tag. The membranes were re-probed with antibodies specific for Gapdh, providing for internal loading controls. C. HeLa cells were transfected with the indicated constructs (shown in A) and processed for IHC imaging using anti-myc antibodies along with DAPI and phalloidin. Data are representative of 3 independent experiments. D. Wildtype human FOXN1 and the indicated variants were transfected into HeLa cells (500 ng plasmid) in combination with a beta5t-luciferase transcriptional reporter construct and beta-galactosidase, the latter for normalization purposes. The levels of transcriptional activity were compared among the variants relative to wildtype FOXN1. All variants were statistically significantly different than the control (p < 0.0001), determined using one-way ANOVA, with triplicate samples/variant tested.

Supplemental Figure 5

Thymopoiesis and peripheral lymphoid populations in adult mice with compound heterozygous mice are like controls. A. Single cell suspensions were prepared from the thymus of 6–8-week-old mice with the indicated Foxn1 mutations. Flow cytometric analyses were used to reveal the percentages of (A) double negative (CD4⁻CD8⁻), double positive (CD4⁺CD8⁺) and single positive (CD4⁺CD8^, CD4⁻CD8⁺) thymocytes, and B. CD4⁻CD8⁻ subset percentages, defined as DN1 (CD44⁺CD25⁻), DN2 (CD44⁺CD25⁺), DN3 (CD44⁻CD25⁻) and DN4 (CD44⁻CD25⁺). C. Mature SP thymocyte development was monitored by gating on TCRbeta⁺, CD69^hi or CD69^lo cell subsets. D. The thymus weight and cell numbers along with the percentage of DN4, DP and CD4 SP determined following flow cytometry and gating. E. Lymph node cellularity and the percentage of T cells, B cells, and NK cells was determined from a minimum of 4 mice/group using littermates. Statistically significant differences were established by one-way ANOVA (Brown-Forsythe and Welch tests).

Supplemental Figure 6

TAT-Foxn1 expression and purification in bacteria. A. TAT-Foxn1 wildtype and p.H321N constructs were expressed in BL21 codon plus bacteria in the absence or presence of IPTG, as indicated with minus or plus signs. Lysates were prepared from the bacteria and resolved on SDS-PAGE followed by Coomassie staining (lanes 1–5). In lanes 5–6, the lysates were subjected to high-speed centrifugation with the supernatants analyzed. Red arrow indicates TAT-Foxn1 wildtype and H321N. B. TAT-Foxn1 wild-type and H321N were purified on Ni²⁺ bead columns, eluted and concentrated with 50 kDa spin cut-off columns. After dialysis, the samples were resolved on SDS-PAGE gels and stained with CBB. Red arrow indicates purified TAT-Foxn1 with 3 and 10 μl of sample loaded/lane, as indicated. Lane 5 contains 10 ug of a bovine serum albumin loading control (BSA). C. Thymic epithelial cells were sorted from thymuses obtained from e13.5 wildtype embryos. These cells were cultured in EpiCult media for 2 weeks. A representative image of the cells is indicated after crystal violet staining. D. TEC cells prepared in C. were untreated or transduced with TAT-Foxn1 and cultured for 24 hrs. The cells were harvested, washed in PBS, and lysed in RIPA buffers. The lysates were resolved on 8% SDS-PAGE followed by anti-Foxn1 immunoblotting and then anti-gapdh. Total cell numbers were about 20,000 cells/lysates, limiting the detection of the endogenous Foxn1 protein.

Supplemental Figure 7

Thymopoiesis is partly attenuated in reaggregate thymus organ cultures in the presence of TAT-Foxn1 H321N fusion protein. A. Diagram depicting reaggregate thymus organ culture (RTOC) strategies with control thymuses. Single cell preparations were prepared from e12–12.5 embryonic thymuses (Foxn1^Wt/Wt) and the cells were stained with mAbs versus Pdgfra (mesenchymal cells) and EpCam (TECs). Cells were sorted into 3 categories: Group I (mesenchymal), Group II (ETPs, endothelial, others) and Group III (TECs). These groups were reaggregated, layered onto membranes and cultured for 10-days. Thymopoiesis was assessed after 10-days by cell enumeration and flow cytometry. B. RTOCs (13,000 cells/cluster) were prepared with wildtype e12–12.5 embryos and cultured in media alone, with TAT-Foxn1 wildtype, or with TAT-Foxn1 H321N. Live cell images revealed the growth of reaggregates. Yellow bar = 1 mm. The percent of thymocyte subsets was determined by flow cytometric analyses following cell surface staining with mAbs detecting CD4 and CD8. C. The total cell number and percent of DP cells was determined from the indicated RTOCs. Statistical significance was not achieved after assessments with one-way ANOVA.

Key Messages.

• Human FOXN1 variants impact T cell development by modulating transcriptional activity, nuclear localization, DNA binding, or form dominant negatives

• FOXN1 mutations have distinct consequences on thymopoiesis when present as autosomal recessive, compound heterozygous or single allelic variants

Abbreviations:

AR

Autosomal recessive

DN

Dominant negative

FTOC

Fetal thymic organ cultures

FOXN1

Forkhead Box N1

OMIM

Online Mendelian inheritance in man

NLS

nuclear localization signal

RTOC

Reaggregate thymus organ cultures

SCID

Severe combined immunodeficiency

TECs

thymic epithelial cells

TCR

T cell receptor

TRECs

T cell receptor excision circles