Generation of inducible SMARCAL1 knock-down iPSC to model severe Schimke immune-osseous dysplasia reveals a link between replication stress and altered expression of master differentiation genes

The Schimke immuno-osseous dysplasia is an autosomal recessive genetic osteochondrodysplasia characterized by dysmorphism, spondyloepiphyseal dysplasia, nephrotic syndrome and frequently T cell immunodeficiency. Several hypotheses have been proposed to explain pathophysiology of the disease, however, the mechanism by which SMARCAL1 mutations cause the syndrome is elusive. Indeed, animal models of the disease are absent or useless to provide insight into the disease mechanism, since they do not recapitulate the phenotype. We generated a conditional knockdown model of SMARCAL1 in iPSCs to mimic conditions of cells with severe form the disease. Here, we characterize this model for the presence of phenotype linked to the replication caretaker role of SMARCAL1 using multiple cellular endpoints. Our data show that conditional knockdown of SMARCAL1 in human iPSCs induces replication-dependent and chronic accumulation of DNA damage triggering the DNA damage response. Furthermore, they indicate that accumulation of DNA damage and activation of the DNA damage response correlates with increased levels of R-loops and replication-transcription interference. Finally, we provide data showing that, in SMARCAL1-deficient iPSCs, DNA damage response can be maintained active also after differentiation, possibly contributing to the observed altered expression of a subset of germ layer-specific master genes. In conclusion, our conditional SMARCAL1 iPSCs may represent a powerful model where studying pathogenetic mechanisms of severe Schimke immuno-osseous dysplasia, thus overcoming the reported inability of different model systems to recapitulate the disease.


INTRODUCTION
The Schimke immuno-osseous dysplasia (SIOD) is an autosomal recessive genetic osteochondrodysplasia characterized by dysmorphism, spondyloepiphyseal dysplasia, nephrotic syndrome and frequently T cell immunodeficiency [1][2][3] . Patients usually suffer of other less penetrant features and, depending on the severity of the disease, they can undergo to premature death in the childhood or early adolescence 3 . The disease is caused by bi-allelic mutations in the SMARCAL1 gene 4 . SMARCAL1 encodes for a protein homologous to the SNF2 family of chromatin remodeling factors, however, recent works firmly demonstrated that SMARCAL1 is not involved in chromatin remodeling and transcriptional regulation, but rather in the processing of DNA structures at replication forks to promote formation of replication intermediates through its ATP-driven strandannealing activity 5,6 . Many mutations in the SMARCAL1 gene have been identified, ranging from frameshift and deletions that generally lead to protein loss, to missense mutations that differently affect expression, activity, stability and localization of the protein 1,7 .
Basing on the pathophysiology of the disease several hypotheses have been proposed 1,8 , however, the mechanism by which SMARCAL1 mutations cause SIOD are completely unknown. The recent demonstration that SMARCAL1 is critical to the response to perturbed replication and that its loss or impaired activity hampers recovery from replication stress and determines DNA damage formation, challenged the canon for SIOD hypothetical molecular pathology from transcriptional regulation to DNA damage prevention. Thus, it is tempting to speculate that, similarly to Seckel syndrome, Werner's syndrome and other genetic conditions caused by loss of genome caretaker proteins, SIOD may be generated by an accumulation of DNA damage and impaired proliferation or development that could follow.
Interestingly, SIOD patients bearing distinct SMARCAL1 mutations show a different degree of disease severity 7 . Thus, a phenotype-genotype correlation might exist although difficult to ascertain. Indeed, mutations resulting in the almost complete loss of protein are associated to severe SIOD. By contrast, mutations that similarly affect SMARCAL1 ATPase activity give raise to both severe or mild SIOD, arguing for the existence of genetic factors that can modulate disease phenotypes or of additional ATPase-independent SMARCAL1 functions that are affected by missense mutations [7][8][9] .
Unfortunately, models of SIOD are absent or were useless to provide insight into the disease mechanism. Indeed, deletion of SMARCAL1 in mice or fruit flies fails to recapitulate the disease phenotype 9 . Only a study from zebrafish evidenced cell proliferation and developmental defects 4 upon deletion of the smarcal1 orthologue 10 , suggesting that loss of SMARCAL1 could affect proliferation and development in humans too. Thus, although likely to exist, the correlation between SMARCAL1 mutations, replication stress, DNA damage formation, defects in proliferation and impaired development in SIOD pathogenesis is yet completely unexplored, largely because of the absence of useful models of the disease.
Induced Pluripotent Stem Cells (iPSCs) are very informative on the very first stages of development. Such model system is very useful for the identification of early events associated to disease pathophysiology. Moreover, it is genetically amenable and provides cell types for drug screening.
Here, we generated a model to study severe SIOD generating iPSCs in which expression of SMARCAL1 could be downregulated through a Tet-ON-regulated RNAi system. Using this cell model, we demonstrated that depletion of SMARCAL1 resulted in reduced proliferation, accumulation of DNA damage, replication defects and DNA damage response overactivation.
Moreover, our data show that the most striking phenotypes are correlated with increased R-loop accumulation and can be reversed preventing replication-transcription interference. Most importantly, using our iPSC cell model of severe SIOD, we show that replication-related DNA damage persists also in differentiated cells and that loss of SMARCAL1 affects expression of a subset of germ layer-specific marker genes. 5

Generation and characterization of inducible SMARCAL1 knockdown iPSCs
To obtain an inducible model of severe SIOD, we expressed an shSMARCAL1 cassette under the control of a Tet-ON promoter through lentiviral transduction in the well-characterized normal iPS cell line WT I 11 (Fig. 1A). Low-passage iPSC were infected with the Tet-ON-shSMARCAL1 virus at 0.5 of MOI by spinfection, selected and tested for the knockdown efficiency by Western blotting.
As shown in Figure 1B, culture of inducible SMARCAL1 knockdown (iSML1) iPSC with doxycycline (DOX) for 48h resulted in less than 13% of total SMARCAL1. Western blotting analysis of the SMARCAL1 level after 7 or 14 days of continuing growth in DOX revealed that the high knockdown efficiency was stable over time in the iSML iPSC (Fig. 1C).
Since the ultimate goal of the use of an iPSC model is differentiation in multiple cell types, we next analysed whether SMARCAL1 knockdown altered the expression of pluripotency genes. To this end, cells grown for 7 days in the presence or not of DOX were analysed for the expression levels of two key pluripotency genes (NANOG and OCT4) by real-time PCR. The analysis of gene expression showed that SMARCAL1 knockdown does not reduce the expression of the main pluripotency marker genes (Fig. 1D).
Having shown that continuous culturing in DOX-containing medium is effective in maintaining SMARCAL1 downregulated, we analysed whether depletion of SMARCAL1 affected proliferation in iSML iPSCs. To this end, iSML iPSCs were grown in the presence or not of DOX for 7 or 14 days and the number of live cells recorded over time. SMARCAL1 downregulation was able to greatly affect proliferation of iSML iPSCs (Fig. 1E). Strikingly, the effect on proliferation was particularly evident starting from 7 days of growth, as shown by the steady cell number and the reduced size of colonies (Fig. 1E). Consistently, iSML iPSCs cultured in the presence of DOX also showed a significant reduction in the number of replicating cells, as evidenced by the decreased number of EdU-positive cells, although no differences were observed between cells grown in DOX for 7 or 14 days (Fig. 1F). Notably, inducible depletion of SMARCAL1 induced reduced proliferation and EdU-incorporation also in normal human primary fibroblasts ( Fig. S1A-C), suggesting that the phenotype is independent on the cell cycle type and not specific of iPSCs.
Collectively, these results indicate that inducible, long-term, depletion of SMARCAL1 in iPSCs is achievable. They also demonstrate that depletion of SMARCAL1, a condition mimicking the severe phenotype of SIOD cells, is sufficient to induce a time-dependent reduction in cell proliferation.

Depletion of SMARCAL1 induces DNA damage and checkpoint activation in iSML iPSCs
Transformed or cancer-derived SMARCAL1-depleted cells are characterized by elevated levels of DNA damage ( 5,12,13 ). Since inducible SMARCAL1 downregulation hampers proliferation in iPSCs ( Fig. 1), we analysed if this phenotype could correlate with enhanced DNA damage. To this end, we These results indicate that continuous cell proliferation with reduced levels of SMARCAL1 leads to DNA damage accumulation. It also results in activation of proteins involved in the DNA-damage response, which is more striking in iPSCs than observed in primary fibroblasts.

Depletion of SMARCAL1 in iPSCs resulted in reduced fork speed and defective replication
Having demonstrated that depletion of SMARCAL1 reduces proliferation of iPSCs cells and increases DNA damage, we tested if it also affected DNA replication dynamics. To this end, we performed single-molecule replication assays using dual-labelling with halogenated thymidine analogues and DNA fibres 14 (Fig. 4A). Analysis of IdU track length in dual-labelled fibres showed that loss of SMARCAL1 did not significantly reduce fork speed in iPSCs, although a small number of IdU tracks with reduced length was noticed in cells depleted of SMARCAL1 (Fig. 4B). Thus, we 7 analysed the fork symmetry, another parameter linked to the presence of stalled forks 15 , increasing labelling time to 30min (Fig. 4C). Notably, downregulation of SMARCAL1 resulted in an increasing number of asymmetric bi-directional forks, as evidenced by a Left/Right fork ratio higher than 1, where the left fork, per definition, is that showing the reduced length (Fig, 4C).
Collectively, these results indicate that downregulation of SMARCAL1 in iPSCs minimally affects DNA replication but induces a significant delay or stalling of a subset of replication forks.

SMARCAL1-depleted iPSCs
Embryonic stem cells are characterised by reduced G1-phase 16 , a condition reminiscent of cells with activated oncogenes and that is correlated with enhanced frequency of replication-transcription conflicts, which requires replication caretaker function 17 . Hence, we tested whether increased DNA damage observed in iSML iPSCs grown in DOX was correlated with unresolved replicationtranscription conflicts. To this end, we grew iSML iPSCs in DOX for 7 days and exposed cells to immunofluorescence. DRB is a transcription inhibitor and is widely used to prevent replicationtranscription conflicts without affecting, in the short-term, proliferation 18  Increased number of replication-transcription conflicts can be associated to enhanced accumulation of R-loops 20,21 . To test if R-loops accumulated in iPSCs in which SMARCAL1 was downregulated, we purified genomic DNA from cells treated or not with DOX for 7 days and assessed the presence of R-loops by dot blot using the S9.6 anti-RNA-DNA hybrids antibody 21,22 (Fig. 5C). As shown in Figure 5D, cells depleted of SMARCAL1 had a substantially-elevated amount of genomic R-loop, suggesting that SMARCAL1 contributes to their prevention or resolution.
Accumulation of R-loops and replication-transcription conflicts may underlie DNA replication defects. Thus, we evaluated replication fork rate in iSML iPSCs cultured in DOX for 14 days and 8 treated or not with DRB the last 4h (Fig. 5F). Interestingly, fork speed was unaffected by DRB treatment in SMARCAL1-depleted cells (+DOX).
Altogether, these results indicate that, in iPSCs growing in the absence of SMARCAL1, the increased DNA damage and depends on replication-transcription conflicts possibly deriving from accumulation of R-loops.

SMARCAL1-depleted iPSCs persist upon their differentiation
Since we demonstrated that sustained depletion of SMARCAL1 induced the accumulation of replication defects and DNA damage in undifferentiated S-phase iSML iPSCs, we assessed if the presence of such DNA damage would persist also after spontaneous plurilineage differentiation.
To this end, iSML iPSCs were grown for 7 days in the presence of DOX before switching from pluripotency maintenance medium to differentiation conditions 11 (Fig. 6A). As shown in Figure   6B, SMARCAL1 knockdown was stable even in differentiated cells. Of note, the relative amount of SMARCAL1 in iSML iPSCs cultured without DOX (i.e. wild-type) declined during differentiation, consistent with its main function in replicating cells. We next analysed the presence of DNA damage by anti-γ-H2AX immunofluorescence in the population of differentiated iSML iPSCs. The  Figure 6C). Interestingly, immunofluorescence analysis of the activation of ATM, which is a readout of both DNA damage and checkpoint activation, revealed an even much more difference between cells cultured in the absence or presence of DOX (Fig. 6D). Notably, the percentage of actively-replicating cells was very low at the end of the 15days differentiation protocol, in both iSML iPSCs ±DOX, as evaluated by EdU incorporation (Fig. 6E). Thus, differences in the number of cells staining positive to the DNA damage markers are unlikely related to an excess of undifferentiated, replicating, cells in the iPSC growing in DOX.
Our data indicate that depletion of SMARCAL1 in iPSCs stimulates the accumulation of persistent DNA damage and long-term DNA damage response (DDR) activation also in differentiated cells.
To determine whether such persistent DNA damage and active checkpoint signal would interfere with pluripotency, we induced formation of embryoid bodies and differentiation into the three germ layers and analysed expression of common marker genes by real-time PCR (Fig. 7A). Analyses of germ layer-specific genes showed that expression of Brachyury (mesoderm) Nestin (ectoderm) AFP and NR2F2 (an inhibitor of OCT4 expressed during early phases of human pluripotent stem cells 9 differentiation; 23 ) were altered in cells depleted of SMARCAL1 while other genes were not affected (Fig. 7B).
Collectively, these results indicate that accumulation of DNA damage and increased DDR caused by loss of SMARCAL1 in undifferentiated cells persist even after differentiation into the progenitors of the three germ-layers, suggesting that the effect of the loss of a replication caretaker maybe "inherited" by differentiating cells. Moreover, our data show that expression of germ layerspecific genes in iPSC-differentiated cells is affected by depletion of SMARCAL1. iPSCs. A telomeric function of SMARCAL1 has been also shown 29 , however, the persistence of these phenotypes in iPSCs, which re-express hTERT 30 , suggests that they are not specifically related with telomere erosion and supports the presence of a more genome-wide replication stress.
Interestingly, persistence of phenotypes in iPSCs also differentiate SMARCAL1 loss from that of WRN, another critical replication caretaker 31 . Indeed, proliferation potential of WS cells is rescued after reprogramming 32 . From this point of view, SMARCAL1-depleted iPSCs behaves more likely as those generated from FA-A cells, which retain all the key cellular defects of the syndrome 33 .
In pluripotent stem cells, the most likely source of replication stress is linked to short G1 phase and increased origin firing 26,34 . A similar mechanism for the generation of replication stress has been put forward after oncogene activation 17 . Of note, in this case, most of the replication stress would derive from interference between transcription and replication 17 . Conditional knockdown of SMARCAL1 in iPSCs does trigger a substantial accumulation of R-loops, which are linked to transcription-replication conflicts 20,21 . This observation would suggest that SMARCAL1 counteracts accumulation of R-loops and transcription-replication conflicts as supported by rescue of DNA damage and ATM activation by transcription inhibition (Fig 5).
The recent observation of a non-canonical ATM activation that is dependent on R-loop accumulation and alternative processing 35 , and increased upon defective replication or mild replication stress 19 , is consistent with our data. Notably, although Smarcal1 KO MEFs do not have any significant proliferation defect compared with wild-type cells, they show a slow-growth phenotype if treated with α -amanitin 9 . Since α -amanitin interferes with elongation phase of transcription and not with initiation like DRB, it is possible that slowed RNA polII increases the chance of replication-transcription conflicts in Smarcal1 KO MEFs, resulting in a proliferation defect as observed in our iPSC model.

Interestingly, DNA damage and ATM activation caused by transcription-replication interference in
iPSCs depleted of SMARCAL1 persist also after differentiation of embryonic body in cells of the three germ layers. Thus, a replication-dependent phenotype seems to be inherited in differentiated cells. Most importantly, loss of SMARCAL1 affects expression of a subset of germ layer-specific master genes. The pathogenetic mechanisms responsible for SIOD are still elusive, however, SMARCAL1 deficiency has been reported to pathologically modulate gene expression 9,25,36-38 . An intriguing possibility is that loss of SMARCAL1 function indirectly influences gene expression through increased levels of replication stress, as suggested for loss of WRN or FANCJ helicase [39][40][41] . Accumulation or persistence of R-loops could be involved in this mechanism making attractive to assess if R-loops preferentially accumulates at affected genes. 1 2 One of the germ layer master genes showing increased expression in SMARCAL1 knockdown iPSCs is BRACHYURY. Notably, expression of BRACHYURY has been found elevated in cordomas and correlates with increased cellular proliferation in the bone 42 , hence providing a possible link to osseous dysplasia, which is one of the clinical phenotypes of SIOD 3 .
Altogether, our work indicates that conditional downregulation of SMARCAL1 in iPSCs recapitulates phenotypes observed in specialized cells following SMARCAL1 depletion. Most importantly, our study demonstrates that loss of SMARCAL1 induces the accumulation of DNA damage and ATM activation in iPSCs through replication stress correlated with replicationtranscription conflicts. Since mutations in SMARCAL1 cause the multisystemic genetic disease  analysed data and wrote the paper. All authors approved the paper.

CONFLICT OF INTEREST
The authors declare that they do not have any conflict of interest. e1006107.

Human iPSC culture, infection and differentiation
Human iPSCs used in this study belong to the WT I line described in Lenzi

Growth curve
The cells were seeded at 1.8 × 104 cells per plate. After trypsinization, cells were counted through electronic counting cells (BioRad) for the following 6 days. The growth curve of the cell cultures was expressed as number of cells as a function of time.

DNA fibre analysis
Cells were pulse-labelled with 25 μ M CldU and then labeled with 250 μ M IdU with or without treatment as reported in the experimental schemes. DNA fibers were prepared and spread out as previously described 43 . Images were acquired randomly from fields with untangled fibres using 1 9 Eclipse 80i Nikon Fluorescence Microscope, equipped with a Video Confocal (ViCo) system. A minimum of 100 individual fibres were analyzed for each experiment and each experiment was repeated three times.

Western blot analysis
Western blots were performed using standard methods. The antibodies used are listed below. Blots were developed using Westernbright ECL (Advasta) according to the manufacturer's instructions.
Quantification was performed on scanned images of blots using Image Lab software, and values shown on the graphs represent a normalization of the protein content evaluated through LaminB1.

Antibodies
The          Multiple t-test comparison (unpaired) was used to assess significance.