Human pluripotent stem cell-based models suggest preadipocyte senescence as a possible cause of metabolic complications of Werner and Bloom Syndromes

Werner Syndrome (WS) and Bloom Syndrome (BS) are disorders of DNA damage repair caused by biallelic disruption of the WRN or BLM DNA helicases respectively. Both are commonly associated with insulin resistant diabetes, usually accompanied by dyslipidemia and fatty liver, as seen in lipodystrophies. In keeping with this, progressive reduction of subcutaneous adipose tissue is commonly observed. To interrogate the underlying cause of adipose tissue dysfunction in these syndromes, CRISPR/Cas9 genome editing was used to generate human pluripotent stem cell (hPSC) lacking either functional WRN or BLM helicase. No deleterious effects were observed in WRN−/− or BLM−/− embryonic stem cells, however upon their differentiation into adipocyte precursors (AP), premature senescence emerged, impairing later stages of adipogenesis. The resulting adipocytes were also found to be senescent, with increased levels of senescent markers and senescence-associated secretory phenotype (SASP) components. SASP components initiate and reinforce senescence in adjacent cells, which is likely to create a positive feedback loop of cellular senescence within the adipocyte precursor compartment, as demonstrated in normal ageing. Such a scenario could progressively attenuate adipose mass and function, giving rise to “lipodystrophy-like” insulin resistance. Further assessment of pharmacological senolytic strategies are warranted to mitigate this component of Werner and Bloom syndromes.


Generation of WRN-and BLM-deficient human embryonic stem cell (ESC) lines.
To generate a WS cellular model using the well characterised H9 pluripotent human stem cell line, an sgRNA targeting exon 3 of the WRN gene was used (Fig. 1a). 24 colonies were picked for screening after targeting, and all but 2 wild-type clones were found to have biallelic gene disruption. No heterozygous clones were observed. Targeting efficiency determined by the percentage of mutated alleles was thus 92%. One wild-type (WRN +/+ ) and one mutant clone (WRN −/− ) were selected for further study. The WRN −/− clone selected harboured a homozygous 1 bp insertion (c.163insT), creating a premature stop codon (p.Y56Vfs*2) (Fig. 1a,b). The frameshift mutation disrupted WRN protein expression as determined by immunoblotting (Fig. 1c).
The BS cellular model was generated using an sgRNA targeting exon 3 of the BLM gene (Fig. 2a). Targeting efficiency was 52.1% with only one clone (BLM −/− ) showing biallelic gene disruption, which was due to a homozygous 11 bp deletion (c.381_392del; p.V127Pfs*11) (Fig. 2a,b). Nine clones out of 24 were found to be wild-type. Multiple attempts at immunoblotting failed to detect BLM protein in wild-type or targeted cells, so a functional assay was used instead to validate successful BLM disruption. A hallmark of BS is an increased rates of sister chromatid exchange (SCE), giving a "harlequin-like" chromosomal appearance on karyotyping after labelling of replicating DNA with BrdU 30,31 , which is the basis of diagnostic cytogenetic testing. The chromosomes of the BLM −/− clone had a classical harlequin-like appearance consistent with functional deficiency of BLM (Fig. 2c,d). Loss of WRN or BLM also did not affect proliferation rates of ESCs (Fig. 3a). As both WRN and BLM play important roles in telomere maintenance, telomere lengths were determined using a qPCR-based technique 32 . No significant differences in telomere lengths were found between WRN −/− and BLM −/− ESCs and their wild-type counterparts (Fig. 3b). Expression of telomerase components telomerase RNA component (TERC) and telomerase reverse transcriptase (TERT) showed no significant differences between WRN +/+ and WRN −/− ESCs, while dyskerin (DKC1) was slightly increased at the mRNA level (Fig. 3c). mRNA expression of TERC and DKC1 did not differ between BLM +/+ and BLM −/− ESCs, but expression of TERT was mildly increased. Taking findings together, we conclude that the loss of WRN or BLM in ESCs does not impair proliferation nor significantly perturb telomere maintenance in ESCs. Proliferation of WRN +/+, WRN −/− , BLM +/+ and BLM −/− AP cells was monitored over 10 days and WRN −/− and BLM −/− AP cells were both found to proliferate at a slower rate than wild-type counterparts (Fig. 4a). WRN −/− and BLM −/− AP cells were also found to have shorter telomeres (Fig. 4b). In line with prior studies, TERT expression was no longer detectable in AP cells (Data not shown). Expression of DKC1 was not affected by knockout of BLM or WRN, while a slight increase in TERC expression in both cases is presumed to be insignificant for telomere maintenance given absence of TERT (Fig. 4c).
WRN −/− and BLM −/− AP cells exhibit attenuated adipocyte differentiation capacity. To investigate later stages of adipocyte development, WRN +/+ , WRN −/− , BLM +/+ and BLM −/− AP cells were next subjected to a previously described adipocyte differentiation protocol 33 (Fig. 6a). WRN −/− and BLM −/− AP cells differentiated less efficiently than wild-type counterparts as assessed by intensity of Oil Red O staining (Fig. 6b), and also adiponectin secretion in the case of the WRN knockout experiment, where differentiation of control cells was better (Fig. 6c). mRNA expression levels of adipocyte markers FABP4, CEBPA, GLUT4, ADIPOQ and PPARG2 assessed by qPCR were all found to be significantly downregulated in WRN −/− cells relative to WRN +/+ cells. BLM −/− cells were also found to express lower levels of FABP4, C/EBPα, GLUT4, ADIPOQ and PPARG2 compared to BLM +/+ cells (Fig. 7a). WRN −/− and BLM −/− cells still showed increased mRNA expression of p16 www.nature.com/scientificreports www.nature.com/scientificreports/ (Fig. 7b). Expression of SASP component Activin A [34][35][36] , was also increased in both WRN −/− and BLM −/− cells relative to the wild-type cells (Fig. 7b), although whether this signal arose from postmitotic differentiated adipocytes or residual senescent APs was not determined. Collectively our data show evidence of increased senescence from early stages of adipocyte development, with attenuated adipocyte differentiation in both WRN and BLM null human cells. This is consistent with the hypothesis that the lipodystrophy-like metabolic complications of WS and BS could arise from premature senescence in the adipocyte precursor compartment. www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
In recent years there has been growing interest in the role played by senescence in adipose tissue in obesity and/or ageing, with an increase in senescence of preadipocytes shown to decrease adipogenesis and yield lipodystrophy, lipotoxicity and inflammation 19 . Accumulation of DNA damage and/or telomere attrition trigger cellular senescence, and this is a common feature of monogenic disorders of DNA damage repair. Only a subset of this large number of disorders features a "lipodystrophic" pattern of insulin resistance, however, including Werner and Bloom Syndrome. This led us to hypothesise a particular vulnerability to telomere attrition and senescence in the mesenchymal lineage leading ultimately to mature adipocytes.
Faithful in vivo models of WS and BS would be the ideal tool for dissection of this phenomenon, however although Wrn knockout mice have been generated, to date there are no BS mouse models, as loss of Blm caused embryonic lethality 37,38 . Furthermore only when the telomerase component Terc was also knocked in Wrn knockout mice, and when several generations were allowed to pass, did they display phenotypes resembling WS 39 . This implicated telomere attrition in WS, given the much longer telomeres found in mice than humans. This makes mice challenging from a practical, financial and ethical point of view as models of WS and emphasizes the need for human cells to model human disease.
Here we report successful generation of WRN −/− and BLM −/− ESCs and isogenic wild-type counterparts using CRISPR/Cas9. WRN-null hPSCs have previously been developed either by reprogramming WS dermal fibroblasts into induced pluripotent stem cells 15,40 or by deleting exons 15 and 16 of the WRN gene in H9 and H1 cells 16 . , indicating that the gradients of the lines were not significantly different from one another. (b) qPCR was performed using primers specific for telomeric ends and single copy reference gene 36B4. The relative telomere lengths were then determined by calculating the ratio between the telomeric DNA product and 36B4. Data are represented as means ± SD, n = 3. ns, not statistically significant. (c) mRNA from WRN +/+ , WRN −/− , BLM +/+ and BLM −/− ESCs were extracted for qPCR analysis to determine the expression levels of telomerase complex genes TERT, TERC and DKC1. The housekeeping gene HPRT was used as a loading control. Data are represented as means ± SD, n = 3. **p < 0.01. ***p < 0.001, ns, not statistically significant. t test. www.nature.com/scientificreports www.nature.com/scientificreports/ Consistent with our findings, all reports agree on no overt differences between wild-type and WRN-null cells in maintenance of pluripotency 15,16,40 . To date, no BLM-null hPSCs have been reported although Blm-null mouse ESCs were first reported in 2004 41 . This study is therefore the first to successfully model BS in hPSCs. Similar to the WRN −/− ESCs, the BLM −/− ESCs were capable of maintaining pluripotency and could be propagated over several passages without loss of pluripotency or proliferative capabilities. Although care was taken to utilise control cells that had gone through the same targeting procedure with the same guide RNA, albeit without editing of the target locus, and that had been subject to the same bottleneck of selection, a limitation of our study is that only single clones for each genotype were studied.
Upon differentiation of WRN −/− or BLM −/− ESCs into AP cells, premature senescence quickly emerged, as assessed by decreased proliferation, increased SA-β-gal staining, and transcriptional upregulation of senescence markers and SASP components. These findings in WRN −/− cells are consistent with previous reports 15,16,40,42 , which suggest that cells derived from the mesenchymal lineage are particularly susceptible to senescence in the absence of WRN 15,43,44 . Whether BS, like WS, also exhibits a segmental, or lineage restricted pattern of premature senescence remains to be determined. More generally, the determinants of this lineage specificity are unclear given that both WRN and BLM are ubiquitously expressed 45,46 .
WRN −/− and BLM −/− AP cells showed impaired differentiation into adipocytes compared to wild-type counterparts, in keeping with senescence in the APs, which prior work suggests cells can autonomously induce senescence and dysfunction of otherwise healthy cells through secretion of SASP components such as IL-6 [47][48][49]   www.nature.com/scientificreports www.nature.com/scientificreports/ BLM −/− ESC lines generated using CRISPR/Cas9, we also attempted shRNA-mediated WRN or BLM knockdown in primary adipose-derived stem cells and the Simpson-Golabi-Behmel Syndrome (SGBS) human preadipocyte line. However, we were unable to sustain these cells in culture long enough for study due to rapidly increased senescence, consistent with observations in the WRN-and BLM-null ESC-derived APs. This adds to evidence that the lipodystrophy-like metabolic phenotype observed in WS and BS patients may be attributable to senescence in the adipose lineage.
Importantly, proof of the concept that clearance of senescent cells from ageing adipose tissue enhances metabolic state has been established using both genetic and pharmacological approaches, and a growing raft of senolytic agents is now available and being assessed in a variety of organ-specific disease models and in clinical trials. Activin A has been reported to impair adipogenesis through the activation of JAK/STAT 35 and Smad2 signalling pathways 36 in an autocrine/paracrine manner. Inhibition of Activin A has been shown to boost adipogenesis 36 , reducing lipotoxicity and improving insulin sensitivity 34 . As Activin A was upregulated in WRN −/− and BLM −/− AP cells, inhibitors targeting this pathway warrant assessment in BS and WS and their in vivo models where available. It will also be of great interest to extend findings from this study to other forms of monogenic severe insulin resistance caused by defects in genes involved in DNA damage repair and/or cell control (e.g. NSMCE2 50 , POLD1 51 , PCNT 52 , POC1A 53 ) to establish if these are unified by a high propensity for senescence in the adipose lineage. This has the potential to define a rare disease target population where trials of senolytics may be of particular value. www.nature.com/scientificreports www.nature.com/scientificreports/

Maintenance of H9 human embryonic stem cell (hESC) line. The human embryonic stem cell line H9
(WiCell) was maintained on Matrigel (Corning)-coated plates at 37°C and 5% CO 2 , in mTeSR1 medium (Stem Cell Technologies). Fresh mTeSR1 medium was applied to the H9 cells every day. Cells were split at a 1:10 ratio approximately every 5 to 7 days when enlarging colonies began to merge. CRISPR/Cas9 gene editing. The CRISPR plasmid construct pSpCas9(BB)-2A-Puro (pX459, Plasmid #48139) was obtained from Addgene. Single guide RNA (sgRNA) sequences were designed using a CRISPR design tool (http://crispr.mit.edu) and cloned into the pX459 plasmid. Sequences of sgRNAs are listed in Table S1. H9 cells were dissociated into single cells with Accutase (Stem Cell Technologies). Two million cells were electroporated with 10 µg pX458-sgRNA using the program CA137 on the Lonza Amaxa 4D Nucleofector (Lonza). One day post electroporation, 1 µg/ml puromycin selection was applied to the cells for 48 hours. Cells were then fed fresh mTESR1 media every day until colonies large enough for manual picking and genotyping via Sanger sequencing. Sequences of genotyping primers are listed in Table S4. After targeting, ability of WRN +/+ , WRN −/− , BLM +/+ and BLM −/− ESCs to differentiated into all 3 germ layers was verified as described by Vallier et al. 2009 54 . Sister chromatid exchange (Sce) assay. BLM +/+ and BLM −/− ESCs were grown to 80% confluence before they were treated with 10 μM BrdU (Thermofisher) for 48 hours, after which mitotic arrest was induced with 150 ng/ml Colcemid (Gibco) for 30 minutes. Cells were then incubated in 75 mM KCl for 15 minutes at 37°C before fixation in a solution of 3 parts methanol (Sigma-Aldrich) to 1 part glacial acetic acid (Sigma-Aldrich). The cell suspension was dropped onto pre-chilled slides, counterstained with 0.1 mg/ml acridine orange (Molecular Probes), and mounted in Sorenson Buffer, pH 6.8 (0.1 M Na 2 HPO 4 (Sigma-Aldrich), 0.1 M NaH 2 PO 4 (Sigma-Aldrich)). Chromatids were visualized using the Zeiss LSM 510 Meta Laser Scanning Microscope (Carl Zeiss) under the FITC filter. SCE events in a metaphase spread was counted and normalized to the total number of chromosomes.  33 . Briefly, H9 colonies were detached from wells and broken into clumps of 5-10 cells in EB formation medium (15% Knockout serum replacement (Thermo Fisher Scientific), 1% GlutaMAX (Thermo Fisher Scientific) in DMEM (Sigma-Aldrich)) supplemented with 4 μM Y-27632 (Sigma-Aldrich). Cell clumps were seeded onto Ultralow attachment plates (Corning) and fed every other day for 5 days after which the EBs were collected and plated onto 0.1% gelatin-coated plates in EB plating medium (10% KnockOut serum replacement (Thermo Fisher Scientific), 1% Glutamax in DMEM (Passage number, P0). Upon reaching 90% confluency, the cellular outgrowths from the EBs were trypsinized and replated onto 0.1% gelatin-coated plates and fed AP medium (15% Knockout serum replacement, 1% GlutaMAX, 2.5 ng/ml bFGF (R&D Systems) in DMEM) every other day (P1). All experiments were performed on P1 AP cells.

Adipocyte differentiation of Ap cells. AP cells were differentiated into adipocytes as described by
Ahfeldt et al., 2012 33 . Briefly, AP cells were transduced with lenti-PPARγ2 viruses (described in Chen et al., 2017 55 ) in the presence of 8 µg/ml polybrene. Adipocyte differentiation medium (15% knockout serum replacement, 0.5% non-essential amino acids (Invitrogen), 1% Glutamax, 1 μM dexamethasone (Sigma-Aldrich), 10 μg/ ml insulin (Actrapid, Novo Nordisk), 0.5 μM rosiglitazone (Sigma-Aldrich) in DMEM) supplemented with 1 μg/ ml doxycycline (Sigma-Aldrich) was applied to confluent transduced AP cells to induce adipocyte differentiation. Adipocyte differentiation medium containing doxycycline was applied to the AP cells every other day for 21 days after which the cells were maintained in doxycycline-free differentiation medium for a further 7 days. cell proliferation assay. Cells were dissociated to single cells and counted. Two thousand cells were seeded into each well of a 96-well plate in triplicate. Cells were harvested at 0, 2, 4, 6, 8, and 10 days and analyzed using the CyQuant Cell Proliferation assay (Invitrogen) per manufacturer's guidelines. Fluorescence was measured at 520 nm using the Tecan Infinite M1000 Pro Microplate Reader (Tecan). www.nature.com/scientificreports www.nature.com/scientificreports/ telomere length measurement. Genomic DNA was extracted using the Gentra PureGene cell kit (Qiagen) according to manufacturer's instructions. The protocol used for measuring telomere length was developed by Cawthon, 2002 32 . Briefly, qPCR was performed using primers specific for telomeres (T) and single copy gene 36B4 (S). Primer sequences can be found in Table S1. T values were then normalized against S to obtain an index of relative telomere length.
DNase digestion removed contaminating DNA before first-strand cDNA synthesis with the ImProm-II Reverse Transcription System (Promega). qPCR was undertaken using the ABI PRISM 7900 Sequence Detection System (Applied Biosystems) with each well of a 384-well plate containing a 7 µl reaction volume made up of SYBR Green PCR Master Mix (Applied Biosystems), the appropriate primers (Table S1 and S2, Sigma-Aldrich) and 2 µl cDNA. The sample cycle threshold (Ct) values were then normalized against a standard curve generated with serial dilutions of a neat cDNA standard made up of pooled cDNA samples. An internal housekeeping gene (human HPRT) was used to normalize the expression of the genes of interests. protein expression analysis. Cells were washed in ice-cold DPBS and then lysed in 50-100 μl RIPA buffer (Sigma-Aldrich) supplemented with complete protease inhibitor cocktail (Roche) to extract total protein. Protein was resolved on NuPage 4-12% gradient Bis-Tris minigels (Invitrogen) with 1X MOPS buffer (Invitrogen) and then transferred onto a nitrocellulose membrane at 20 V for 7 to 10 minutes using the iBlot dry blotting system (Invitrogen). 5% powdered skimmed milk or 5% BSA, both diluted in Tris-buffered saline (TBST: 0.05 M Tris, 0.138 M NaCl, 0.0027 M KCl, pH 8.0) with 0.1% Tween-20 (Sigma-Aldrich) were used as blocking buffer. Membranes were incubated overnight at 4 °C with the primary antibodies. Membranes were washed 3 times in TBST and then incubated with the appropriate HRP-linked secondary antibodies. The list of antibodies used in this study can be found in (Table S3). The membrane was then washed 3 times in TBST before the protein was interest was visualized using the EMD Millipore Immobilon Western Chemiluminescent HRP Substrate (ECL) (Millipore) and the ChemiDoc (Biorad).
immunostaining. Cells were fixed in 4% neutral-buffered formaldehyde (Sigma-Aldrich) and permeabilized and blocked in 10% donkey serum (Sigma-Aldrich) in 0.1% Triton X-100 in DPBS (PBST) for 20 minutes at RT. Cells were incubated with primary antibodies diluted in 1% donkey serum in PBST for 1 hour at RT before they were washed 3 times in DPBS. Appropriate Alexa-fluor-conjugated secondary antibodies diluted 1:1000 in 1% donkey serum in DPBS were applied to the cells for 30 minutes in the dark at RT. Antibodies used in this study are listed in Table S3. Cells were then washed 3 more times in DPBS. DAPI was diluted 1:10,000 in DPBS, added to the cells and incubated for 2 minutes at RT in the dark. Cells were washed a further 3 times with DPBS and then imaged using the EVOS FL imaging system (Life Technologies). flow cytometry. AP cells were incubated in 0.125% trypsin-EDTA (Sigma-Aldrich) for 1 minute at 37 °C to detach them before resuspension in AP medium and transfer into a 15 ml falcon tube. 1 × 10 6 cells were transferred into each round-bottom polystyrene tube (BD Biosciences), washed once in fluorescence-activated cell sorting (FACS) buffer made up of 0.005 g/ml bovine serum albumin (BSA) (Sigma-Aldrich) in DPBS and then incubated with PE-labeled antibodies against CD73 (BD Biosciences) for 30 minutes in the dark at RT. Cells were pelleted by centrifuging at 1300 rpm for 3 minutes at RT and washed with FACS buffer. The cell pellet was resuspended in 100 μl 1X DPBS and analysed on a BD FACSCalibur flow cytometer with the percentage of CD73-expressing cells determined by acquiring 10,000 events per cell type. Data were analyzed using Flowing software developed by the Turku Center for Biotechnology (Finland).
oil Red o staining. Differentiated AP cells were fixed with 10% neutral-buffered formalin solution (Sigma-Aldrich) for 10 minutes at RT at day 28 then rinsed twice in 60% isopropanol. Freshly prepared Oil Red O working solution (6 parts 0.25% Oil Red O in isopropanol: 1 part 60% isopropanol: 3 parts water) was applied to cells for 30 minutes at RT. Excess Oil Red O stain was removed with 60% isopropanol washing before image acquisition.
Statistical analysis. The two-tailed Student's T test was used to test for statistical significance where only a single parameter was compared between two groups. A p value of <0.05 was considered statistically significant. All error bars indicate mean + /− standard deviation (SD). All statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software Inc., San Diego, CA).