T lymphopenia was identified in the male proband by routine newborn SCID screening for T cell receptor excision circles (TREC) using blood from a Guthrie blood spot filter card, as described (1, 21, 22). Research activities were performed with parental informed consent under protocols approved by the institutional review board (IRB) at the University of California, San Francisco.

Research-based whole exome sequencing (WES) was performed on cells from the proband and parents with bioinformatic analysis and variant calling, as described (9, 23). Briefly, libraries were prepared by appending TruSeq adaptors (Illumina: San Diego, CA) to genomic DNA fragments of 200–300 bp that were enriched with 10 cycles of PCR, pooled and submitted to exon capture using a Roche Nimblegen kit (V3.0). After 10 additional amplification cycles, 100 bp paired end sequence reads were generated (HiSeq2000, llumina) and aligned against GRCh37 (Aug 2009 release) using BWA (v0.6.2). The results were converted to BAM, sorted by coordinate, indexed, and marked for PCR duplicate reads using the Picard toolkit (v 1.81) (http://picard.sourceforge.net). Local realignment was performed around known indel locations, and base quality scores were re-calibrated using GATK (v 2.6.5). Variants were called using GATK UnifiedGenotyper and freebayes (vesion 0.9.10) and variant quality scores were re-calibrated by VQSR (Variant Quality Score Recalibration) using the trio of exomes in this report, as well as 65 others sequenced at our site. HapMap v3.3, 1000 genomes high confidence SNPs (phase1 v3 2010-11-23) and Omni chip array sets were used as training data, and to provided truth sites for SNPs, while the Mills dataset from the 1000 genomes was used for indels, using a truth sensitivity cutoff of 99% (24). Variant annotation (including region, effect, allele frequency, disease phenotype annotation, and conservation) was performed using our custom tool Varant (http://compbio.berkeley.edu/proj/varant/). Particular attention was given to high-confidence, rare, likely-damaging, protein-altering variants in genes associated with primary cellular immunodeficiency (25). WES data from the JPSCID-15 trio were submitted to dbGaP, accession number phs002968.v1.p1.

Tubingen long fin zebrafish were maintained at 28.5°C under standard aquaculture conditions. Animal housing and handling were all performed in accordance with the approved protocol from the Institutional Animal Care and Use Committee (IACUC).

The zebrafish CHTF18 ortholog chtf18 (NM_001110102.3) was previously identified. CTBP2 paralogues ctbp2a (NM_001195491.1 and NM_131715.1), and ctbp2l (NM_001015064.1) were identified by homology and synteny, as described (10, 26). Antisense morpholino oligonucleotides (MO; Gene Tools) were used to block splicing of zebrafish chtf18 (5’ ACTGGAACAAAACACACCTCTTCAT 3’), ctbp2a (5’ TTCGGTCAGAAACAATCATACACAC 3’) and ctbp2l (5’ CCGGAAACAAGTAATGATGCAACTG 3’) upon injection into one-cell embryos using a PM1000 microinjector (Microdata Instrument, Inc.). The efficacy of the MO in disrupting splicing was assessed by reverse-transcriptase (RT)–PCR using the following oligonucleotides ( Table 1 ): 1) chtf18: F-AGGAGGCAAGATGTGGCG and R-CATTTCAGCAGACAGCGATTAG; 2) ctbp2a: F-CGGTGGGTGCGATGATG and R-TGTACCAGGCTGTGTGAGGG; 3) ctbp2l: F-CAGGAGCGGCTCCGAAGACGTTTTCCGGC and R-ACAGAAGGCAACGGTTGCCAGATC; and 4) actb: F-TGGCATCACACCTTCTAC and R-AGACCATCACCAGAGTCC (27). Tp53 rescue experiments were conducted as described in (28). To assess the capacity of wildtype and variant human CHTF18 orthologs to rescue the loss of zebrafish, wildtype and patient variant CHTF18 were cloned into heat shock-inducible expression vector pSGH2 using standard approaches. Sanger sequencing confirmed the R751W, V776L and E851Q variant alleles. Ectopic expression of wild-type and mutant human CHTF18 was achieved by injection of the heat-inducible pSGH2 vector (29) into one-cell stage embryos, which were heated to 37°C for 1 hour at 30 hours post fertilization. GFP+ embryos were selected at 5 dpf for analysis by WISH using a probe for lck as described (9). The impact of chtf18 knockdown on HSPC localization was assessed by injection of chtf18e6 splice blocking MO followed by the performance of WISH at 36 hours post fertilization, as described (28).

WISH was performed as described (30) using anti-sense probes for lck, ikaros, tcrd, runx1, and foxn1 (28, 31). The stained embryos were photographed using the Nikon SMZ1500 stereomicroscope equipped with DS-Fi1 digital camera and Nikon Ar imaging software. Image J software was used to measure integrated density of the lck WISH staining of zebrafish thymi.

Genomic sequences were obtained using the NCBI and ENSEMBL databases. Multiple alignments of human, mouse and zebrafish CTF18 and CTBP2 amino acid sequences were obtained using Clustal X. Structural modeling was performed as described for human CTF18 using UCSF Chimera software and AlphaFold predicted structures for CTF18 wildtype, the arginine 751 to tryptophan (R751W) variant and glutamic acid 851 to glutamine (E851Q) variant, as well as the RFC1 clamp loading structure (PDB code 6VVO) (10, 32). CTF18 has limited homology to the clamp loader paralog RFC1 found in the CryoEM structure of human clamp loader bound to PCNA (32). AlphaFold-Multimer v2.3 in the ColabFold implementation (33) was used to predict the structure of a conserved portion of CTF18 (residues 595 to 865) that contained the 3 missense mutations identified in the proband and their interactions with the nearest neighbor subunit in the clamp loader complex, RFC3. The resulting heterodimeric models were subjected to a coordinate constrained method of protein relaxation using the Rosetta (34) molecular modeling suite to optimize side chain packing (35–37). These relaxed models were aligned by the RFC3 subunit using the UCSF Chimera 1.15 software package (38) and analyzed for contacts between CTF18 variants and RFC3 with its bound ADP. As a test to validate our modeling procedure, we used the same AlphaFold methods to recapitulate the RFC1-RFC3 interaction, and the resulting model superposed with RFC1 subunit of the 6VVO structure with an RMSD < 0.76 Ų for 200 alpha carbons in this region. Final refined modeling methods utilized AlphaFold3 (AF3) (39), which can predict the combined structure of multicomponent complexes including proteins, nucleic acids, small molecules, ions and modified residues. We retained a focus on the conserved fragment of CTF18 (583-863) bound to RFC3, and added an ADP molecule and magnesium ion that is often bound in the active site. Multiple AF3 runs were used to generate 20 models each of the WT R751, mutant R751W, and the mutant E851Q. In all cases the models scored with high confidence (ipTM range 0.89 to 0.91) and backbone superposition of models yielded RMSD values in the range of 0.4 to 0.8 Ų. The WT R751 rotamer chosen by the AF3 modeling was the same as that found in the PDB structure 8UN0 (state7) for 17 of the top 20 models, albeit with very minor differences in position due to slight backbone variation. In all cases the top ranked models were used to analyze the interface interactions and prepare figures with UCSF Chimerax version 1.7.1. To obtain a general overall interface assessment, we also analyzed the CTF18(583-863)-RFC3 interface with PISA analysis (40).

All mouse strains were housed in the Laboratory Animal Facility at Fox Chase, which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, and handled in accordance with an IACUC approved protocol. Ctf18 R745W and E845Q knockin mutant mice were generated by CRISPR genome editing (10). Chf18 R745W mice were created using a 150 bp oligo donor to mutate R745 to W and silent cytosine (C) to thymine (T) to inactivate the PAM and prevent re-cutting of the repaired allele. Similarly, the Ctf18 E845Q mice were created using a 150 bp oligo to mutate E845 to Q and inactivate the PAM to prevent re-cutting of the repaired allele. The knockin mutation created BseYI and HpyCH4V restriction enzyme sites, respectively, to genotype the mutated alleles. All oligonucleotides used in study are listed in Table 1 .

Single-cell suspensions prepared from mouse thymus, spleen, bone marrow, or peripheral blood were stained with optimal amounts of the following fluorochrome-conjugated antibodies: anti-CD4 (GK1.5), anti-CD8 (53-6.7), anti-CD24 (M1/69), anti-CD25 (PC61), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-CD69 (H1.2F3), anti-CD73 (TY/11.8), anti-CD90.2 (30-H12) anti-B220 (RA3-6B2), anti-NK1.1 (PK136), anti-TCRγδ (GL3), anti-TCRβ (H57-597), anti-CD19 (eBio1D3), anti-CD3 (17A2), anti-Gr1 (RB6-8C5), anti-Sca1 (D7), and anti-cKit (2B8). The antibodies were purchased from BD Biosciences, eBioscience, BioLegend, or Tonbo Biosciences. Dead cells were excluded from analyses using propidium iodide (PI). Data were analyzed on either a BD LSRII or Symphony A5 flow cytometer (BD Pharmingen) using Flowjo 9.96 software (Treestar, Inc).

30,000 lineage depleted bone marrow cells from wildtype or mutant mice were loaded with the objective of collecting a single library of 10,000 cells by Chromium controller (V3 Chemistry version, 10X Genomics Inc, San Francisco, USA). Barcoding and DNA purification were performed according to the manufacturer’s protocol. Sequencing was performed using 2 x 150 pair-end configuration on an NovaSeqX Plus platform at a sequencing depth of 375G/lane. Cell-barcode and unique molecular identifiers were extracted and aligned through Cell Ranger and expression analysis was performed using Seurat V4, as described (41). Uniform Manifold Approximation and Projection (UMAP) plots were used to visualize the data based on canonical component analysis of the top 1000 genes with highest dispersion between all sample groups. Gene ontology and Gene Set Enrichment analysis was performed using EnrichR.

CD45.1 mice were lethally irradiated with a total dose of 11 Gy, split in two doses of 6.5Gy and 4.5Gy, separated by three hours. After 24 hours, mice were injected i.v. in the tail vein with either 100,000 lineage depleted adult bone marrow (BM) CD45.2 hematopoietic progenitors alone or together with 100,000 adult BM CD45.1+ competitor progenitors delivered in 200μl Hank’s Balanced Salt Solution (HBSS), as described (10). Engraftment was monitored by retro-orbital bleeding at the indicated intervals. Mice were sacrificed after 6 weeks of engraftment and bone marrow, thymus and spleen were analyzed by flow cytometry. For fetal liver (FL) transplants, 100,000 embryonic day 14.5 (E14.5) FL cells from CD45.2 mice were transferred as above with 100,000 CD45.1 FL competitor cells as described above. Seeding assays were performed by transplanting adult BM progenitors as above and determining the seeding of the BM of recipient mice 4 days later by performing flow cytometry on explanted bone marrow cells for the allelic marker (CD45.2/1).

To assess the capacity of WT and compound Chtf18 mutant HSPC to respond to repeated entry into cell cycle, mice were treated twice weekly for 60 days with two intraperitoneal doses of pIpC (15mg/kg) as described (42), with weekly monitoring of peripheral blood lymphocyte content by flow cytometry.

E14.5 FL progenitors from WT and compound Chtf18 mutant mice were cultured on OP9-DL1 monolayers, as described (43). Developmental progression was monitored by quantitating cell growth of control triplicate cultures or those treated with agents including Interferon-β, Mitomycin-C, Ganetespib, and Aramycin-C.

Graphic representations of data were generated using GraphPad prism V9, following statistical analysis using ANOVA and Student’s t tests. Significance values are indicated.

The male infant (JPSCID-15) was born at term after an uncomplicated pregnancy to nonconsanguineous, healthy parents with no known family history of immune deficiency. Newborn screening revealed low T cell receptor excision circles (TREC; normal >25), with T-cell lymphopenia subsequently documented by flow cytometry ( Table 2 ) (21). The proportion of naïve recent thymic emigrant (i.e., CD45RA+) CD4+ and CD8+ T cells was reduced, with CD8+ cells exhibiting a slightly greater reduction (available data shown in Table 2 ) (46, 47). B and NK cell numbers were normal. The patient had a normal physical exam with no syndromic features. Proliferation to phytohemagglutinin, immunoglobulin levels and titers to killed vaccines were obtained over the ensuing months and were normal (44, 45). During his first year the patient experienced one episode of pneumonia treated successfully with oral antibiotics, but no other infections. Growth and development remained normal until he was lost to follow-up after his second birthday. Genetic workup included normal fluorescent in situ hybridization for chromosome 22q11.2 deletion and normal alpha-fetoprotein that ruled out DiGeorge syndrome and ataxia telangiectasia, respectively. Sanger sequencing of a panel of SCID genes, including ADA, DCLRE1C, IL2RG, IL7R, JAK3, RAG1, RAG2, RMRP, and ZAP70, revealed no mutations.

While sequencing of known genes implicated in immunodeficiency revealed no mutations, whole exome sequencing (WES) revealed several potential variants, with the most likely variants being mutations in CTBP2 and CHTF18. CTBP2 is a transcriptional corepressor which binds Proline-Isoleucine-Aspartate-Leucine-Serine (PIDLS) to repress target genes, including IL-2 (48, 49). In zebrafish, ctbp2a/ribeye a and ctbp2l/ribeye b are the major proteins in synaptic ribbons and loss of both genes leads to decreased electron density, synaptic localization and recruitment of calcium channels (50). The patient had two variants in CTFBP2: 1) a de novo CTFBP2 D437N missense mutation affecting an aspartic acid (D) residue conserved in mice and zebrafish Ctbp2a isoform 2, but not in Ctbp2a isoform 1 or Ctbp2l; and 2) a 12aa insertion following P389 in a portion of the protein conserved in all isoforms ( Supplementary Figure 1A ). To investigate the role of CTBP2 in T cell development, we employed morpholino oligonucleotides (MO) to knock down ctbp2 expression in zebrafish embryos and assessed the impact on T cell development by performing whole mouse in situ hybridization (WISH) using a T cell specific lck probe at five days post fertilization (dpf; Supplementary Figure 1B ). Splice site MO targeting ctbp2a, ctbp2l, alone or in combination attenuated ctbp2a and ctbp2l expression, but did not arrest in T cell development, suggesting that the CTBP2 patient variants were not responsible for disease ( Supplementary Figure 1B ).

The only other highly ranked set of variants impacted the CHTF18 gene. The CTF18 protein, encoded by the CHTF18 gene, is a component of a secondary clamp loader complex, which loads PCNA onto DNA during replication (16, 17). CTF18 forms a pentamer with RFC2, RFC3, RFC4 and RFC5 (51). The proband exhibited two maternal (R751W and V776L) and one paternal (E851Q) missense variant in the human CHTF18 gene, in residues that are conserved from human to mouse and zebrafish ( Figures 1 , 2A ; Supplementary Figure 2 ). To assess the extent to which these variants might disrupt the structure of the CTF18-containing secondary clamp loader, we examined a recently solved cryo-EM structure of the complex (51) and compared it to that of the RFC1-containing primary clamp loader complex using AlphaFold 2 (AF2) and AlphaFold 3 (AF3), with a particular focus on the impact of the CTF18 variants ( Figures 2B, C ) (32, 33, 54) (39). In the RFC1 complex, the RFC3-R9 residue is a key component of the ADP binding pocket and the ε amino group hydrogen bonds to the 3´ hydroxyl of ADP ( Figure 2B ). The analogous CTF18 loop, which is in close contact with RFC-R9, revealed a slightly different conformation containing a short alpha helix, as there are two extra amino acids in this CTF18 loop compared to the RFC1 loop ( Figure 2C ). Nevertheless, the analogous CTF18 residue to RFC1-R986, CTF18-R751, also points toward RFC3-R9 and the ADP binding pocket ( Figures 2B, C ; arrow), suggesting that it may also contribute to the dynamics of ADP binding and the ATP hydrolysis cycle. Moreover, the CryoEM structure led to speculation that the alternative CTF18 clamp loader complex is a much more mobile multi-subunit assembly compared to the canonical RFC1-based clamp loader. This may lower loading activity, thus limiting the alternative clamp loader complex to leading strand loading (51). Interestingly, the AF3 models of R751W variant forces a downward rotamer change for RFC3-R9 resulting in RFC3-R9 stably making four hydrogen bonds with either RFC3-D6 or the 3’hydroxyl of ADP (see blue dashed lines in Figure 2D , arrows indicate R751W and RFC3 R9). The R751W variant could cause a perturbation of the ATP hydrolysis cycle (i.e. slower ADP release) by rigidly fixing the rotamer of RFC3-R9. In contrast, the CTF18-E851Q variant is located away from the interface with RFC3 in a solvent exposed location and not near any other subunit in the CryoEM structures ( Figure 2E , arrow indicates E851Q). This is a mild side chain substitution and it is unclear how this variant could alter function.

To investigate the role of CTF18 protein in supporting T-cell development in vivo, we employed zebrafish, a tractable genetic model in which hematopoietic development is highly conserved (26, 55). Expression of the zebrafish chtf18 ortholog was knocked down using a MO that interferes with chtf18 splicing and induces nonsense mediated decay ( Figure 3A ). The chtf18 MO attenuated chtf18 expression and markedly impaired T cell development at 5dpf, as indicated by decreased whole mount in situ hybridization (WISH) staining with an RNA probe for the T cell specific kinase, lck ( Figure 3A ). The blockade in T cell development caused by knockdown of chtf18 indicated that Ctf18 performs an important function in supporting T cell development; however, it did not provide insight into whether the CTF18 patient variants damage CTF18 function. To determine whether the patient variants (R751W, V776L and E851Q) abrogated the function of CTF18 protein, we performed rescue experiments in zebrafish embryos. Endogenous chtf18 was knocked down by MO as in Figure 3A and heat shock induction was employed to re-express the wild type and variant human CHTF18 orthologs to determine if they rescue the block in T cell development caused by loss of zebrafish Ctf18 ( Figure 3B ) (9, 10). Re-expression of the wild type (WT) human ortholog compensated for the loss of Ctf18 and rescued T cell development, indicating that CTF18 function was conserved from human to zebrafish ( Figure 3B ). Moreover, the maternally inherited V776L variant also rescued development, indicating that this mutation is not damaging ( Figure 3B ); however, re-expression of either the R751W (maternal) or the E851Q (paternal) variant failed to rescue T cell development, indicating that each of these variants damage CTF18 function ( Figure 3B ). Taken together, these data suggest that T cell insufficiency in patient JPSCID-15 may result from compound heterozygosity of two damaging CHTF18 mutations (R751W and E851Q).

To gain insight into the breadth and stage of developmental arrest caused by Ctf18 loss, we performed WISH using probes for other T cell subsets and thymic cell types. Chtf18 knockdown did not appear to impair thymic seeding as staining for ikaros, a marker of early thymic progenitors, was not altered ( Supplementary Figure 3A ). The arrest of T cell development appeared to be restricted to αβ lineage T cells, since the γδ T cell subset identified using WISH with a tcrd probe was not affected by Ctf18 loss ( Supplementary Figure 3A ). Moreover, the block in αβ T cell development was likely to be cell autonomous, since WISH using a foxn1 probe suggests that thymic stroma was not disrupted by chtf18 knockdown ( Supplementary Figure 3A ). Finally, because CTF18 functions as a secondary clamp loader during DNA replication (17, 53, 56), we reasoned that chtf18 knockdown might cause DNA damage resulting in a p53-dependent developmental arrest; however, trp53 knockdown failed to rescue development in chtf18 morphants, despite rescuing the p53-dependent arrest caused by knockdown of rpl22 ( Supplementary Figure 3B ) (28). Together, these findings suggest that ctf18 loss arrests the development of αβ lineage T cells in a cell-autonomous, but p53-independent, manner.

To understand the mechanistic basis for Ctf18 action, we generated knockin mice bearing patient variants in the orthologous residues of mouse Ctf18, R745W and E845Q ( Supplementary Figure 4 ). The knockin mice were generated using CRISPR-induced cutting and HDR oligos for repair. The resulting compound heterozygous Chtf18^R745W/E845Q (Chtf18 mut) mice were fertile and viable, with no outward signs of abnormalities (data not shown). We next examined T cell development by flow cytometry ( Figure 4A ). We observed no significant difference in any of the thymic subsets defined by CD4, CD8, CD44, and CD25. Moreover, splenic T cell numbers were not diminished; nor did they exhibit any signs of homeostatic proliferation marked by changes in expression of CD62L and CD44 ( Figures 4B, C ) (57). Developmental defects in some cases are evident only when hematopoietic stem and progenitor cells (HSPC) are stressed (10, 58). To place Chtf18 mut HSPC under stress, we performed mixed bone marrow chimera analysis using allotype marked (CD45.1) competitor cells ( Figure 5A ). While the wild type Chtf18+/+ (WT) HSPC effectively reconstituted the thymus and spleen of recipient mice, the Chtf18 mut HSPC were unable to reconstitute recipient mice ( Figures 5A, B ). The failure to reconstitute recipient mice did not result from failure to home to the bone marrow, as a seeding assay revealed no defect in Chtf18 mut HSPC reaching the bone marrow of the recipients ( Figure 5C ). Together, these data indicate that Chtf18 mut HSPC are able to support hematopoiesis at baseline, but fail to do so under stress during competitive bone marrow transplantation.

To elucidate the molecular basis for the impaired capacity of Chtf18 mut HSPC to reconstitute hematopoiesis in competitive transplant experiments, we performed single-cell RNA-seq (scRNA-seq) on lineage negative (Lin-) HSPC. Lin- HSPC were selected because of their functional impairment in competitive transfer experiments, and because publicly available scRNA-Seq analysis revealed that Chtf18 mRNA levels are highest in the stem and progenitor cells in the bone marrow ( Supplementary Figures 5A, B ). Chtf18 mRNA levels were also found to elevated in CD8 immature single positive cells (ISP), which are transitioning from the CD4-8- double negative (DN) to the CD4+8+ double positive (DP) stage of thymocyte development ( Supplementary Figure 5C, D ). The scRNA-seq analysis revealed eight clusters defined by lineage specific stem cell markers derived from the Tabula Muris dataset ( Supplementary Figure 6A ) (https://www.czbiohub.org/sf/tabula-muris/). The clusters showed minimal changes in representation in the Chtf18 mut HSPC, except for the myeloerythroid progenitors (MEP), which were substantially reduced ( Figure 6A ). Pathway analysis revealed that among the differentially expressed genes (DEG), there was no enrichment for DNA-damage response pathways, but there was marked enrichment in pathways related to interferon signaling and protein folding ( Figure 6B ; Supplementary Figures 6B, C ). Expression of the remaining clamp loader components (RFC2-5) was increased in Chtf18mut myeloid progenitors ( Figure 6C ). Interestingly, expression of Myc, which has been implicated in controlling HSC quiescence, was markedly elevated in Chtf18 mut HSC/MPP ( Figure 6C ) (59). Using transcriptomic analysis to infer cell cycle status, we found that long-term HSC (LT-HSC), short term HSC (ST-HSC) and CD34 expressing LT-HSC (LT-34F) all exhibited an increase in cells in cycle (G2M/S) in the Chtf18 mut mice, consistent with the potential loss of quiescence by HSC ( Figure 6D ).

The increased expression of Myc and proliferative transcriptomic signature of Chtf18 mut HSC raised the possibility that their failure to reconstitute hematopoiesis in the competitive setting resulted from exhaustion. To test this possibility, we performed competitive transfers using FL HSC, which are highly proliferative (60). Interestingly, Chtf18 mut FL HSC also failed to support hematopoiesis upon competitive transfer into irradiated hosts ( Supplementary Figure 7A ). The failure of Cht18 mut HSC to support hematopoiesis was not evident in non-competitive transfers ( Supplementary Figure 7B ), indicating that Chtf18 mut HSPC had the capacity to support hematopoiesis but were unable to do so when forced to compete with fully competent WT HSPC. The elevated transcriptomic signatures of IFN signaling and protein folding in Cht18 mut HSPC were reminiscent of what is observed in some mouse models of Fanconi Anemia (Fanca-/-) (61, 62), in which defects do not manifest until stressed (42). To test this possibility, Chtf18 mut mice were repeatedly stressed by biweekly pIpC treatments; however, unlike the lethality and hematopoietic defects observed following serial treatment of Fanca^-/- mice with pIpC, biweekly treatment of Chtf18 mut mice had no significant impact on their viability or peripheral lymphoid populations ( Supplementary Figure 7C ). Likewise, Cht18 mut FL HSPC grew normally in vitro and did not exhibit increased sensitivity to interferon-β (IFNβ), heat shock protein inhibitors (Ganetespib), or DNA damage induced by Mitomycin c (MMC) ( Supplementary Figures 7D–F ), unlike what has been observed for cells with mutations in FANC family proteins (42, 61, 62). Finally, contrary to reports in cell lines lacking CTF18, Chtf18 mutant HSPC did not exhibit increased sensitivity to the chain-terminating chemotherapeutic agent, Cytrabine (AraC) ( Supplementary Figure 7G ) (63).

Because the defect in Chtf18 mutant HSPC did not phenocopy that seen in Fanca-/- mice, we re-examined the expression signature alterations in Chtf18 mut HSPC. Interestingly, we found that Chtf18 mut common lymphoid progenitors (CLP) exhibited changes in the expression of genes regulating cell-cell interactions, migration, and chemotaxis ( Figures 7A ; Supplementary Figure 8A ). Specifically, the expression of several Rho family GTPases and their regulators (RhoA, Cdc42, Pak2, and Rock1) was diminished in Chtf18 mut CLP ( Figures 7A ; Supplementary Figure 8A ). Rho family GTPases have been shown to regulate HSC migration, and localization, in addition to playing a critical role in the asymmetric division that is required to balance self-renewal with differentiation (64–67). Indeed, loss of these effectors markedly perturbs these processes and so may explain the defective function of Chtf18 mut HSPC in the competitive transplant setting (64–66). To determine if Chtf18 mut HSPC caused the mis-localization of HSPC, we employed the zebrafish model, in which HSPC positioning can be readily evaluated in embryos by WISH using a runx1 probe to identify emerging HSPC ( Figure 7B ). Indeed, we observed that knockdown of chtf18 using MO resulted in a marked dispersal of HSPC away from the embryo midline, indicating that Ctf18 loss caused the mis-localization of HSPC ( Figure 7B ). Together, these data suggest that the competitive disadvantage displayed by Chtf18 mut HSPC resulted from alteration of their positioning during development.