Thioredoxin-interacting protein regulates haematopoietic stem cell ageing and rejuvenation by inhibiting p38 kinase activity

Ageing is a natural process in living organisms throughout their lifetime, and most elderly people suffer from ageing-associated diseases. One suggested way to tackle such diseases is to rejuvenate stem cells, which also undergo ageing. Here we report that the thioredoxin-interacting protein (TXNIP)-p38 mitogen-activated protein kinase (p38) axis regulates the ageing of haematopoietic stem cells (HSCs), by causing a higher frequency of long-term HSCs, lineage skewing, a decrease in engraftment, an increase in reactive oxygen species and loss of Cdc42 polarity. TXNIP inhibits p38 activity via direct interaction in HSCs. Furthermore, cell-penetrating peptide (CPP)-conjugated peptide derived from the TXNIP-p38 interaction motif inhibits p38 activity via this docking interaction. This peptide dramatically rejuvenates aged HSCs in vitro and in vivo. Our findings suggest that the TXNIP-p38 axis acts as a regulatory mechanism in HSC ageing and indicate the potent therapeutic potential of using CPP-conjugated peptide to rejuvenate aged HSCs.

T he ageing of stem cells underlies the ageing of tissues and represents a progressive decline in functional activities that maintain the homoeostasis and regeneration of the specific tissues in which stem cells reside 1,2 . In the haematopoietic system, haematopoietic stem cells (HSCs) continuously replenish blood cells, exhibiting a high turnover rate throughout life. The ageing of HSCs is likely the key process of decline in immune function with age or ageing-associated diseases and is driven by both extrinsic and intrinsic factors [3][4][5] . A series of studies have reported that aged mice exhibit remarkable changes in their haematopoietic systems, such as the expansion of CD34 À Flk2 À LSK (lineage À c-kit þ Sca-1 þ ) cells (LT-HSCs), lineage skewing, an increase in reactive oxygen species (ROS) and a decreased number of leucocytes in the peripheral blood (PB) [4][5][6][7] . HSC research groups have proposed several factors involved in HSC ageing, including mitochondrial DNA damage 8 , ROS and p38 (refs 9,10), DNA damage 3 , telomere shortening 11 , epigenetic alteration 12 , loss of Cdc42 polarity 4,13 , Wnt5a 13 , replication stress 14 and others 1 . Recent reports have also suggested possible mechanisms to rejuvenate aged HSCs via the reduction of Cdc42 activity using its inhibitor 4 , SIRT3 overexpression 15 and prolonged fasting 16 .
TXNIP is a known inhibitor of thioredoxin and is a tumour suppressor that blocks cell-cycle progression 17,18 . In our previous results, TXNIP was highly expressed in HSCs and its expression decreased as HSCs differentiated into lineage cells. TXNIP deficiency exhibited higher levels of ROS in HSCs and decreased HSC repopulation capacity. TXNIP acted as an antioxidant protein under oxidative stress by regulating p53 activity via direct interaction 19-21 . p38 is a Ser/Thr kinase that regulates the growth, proliferation, death and differentiation of cells in response to multiple stimuli 22,23 . Many researchers have observed p38 activation in various pathological conditions or during cellular ageing via elevated ROS, resulting in HSC defects. These researchers have also suggested that the pharmacological inhibition of p38 activity might restore the defects of HSCs in vitro and in vivo. For example, administration of SB203580, a p38 inhibitor, restored repopulation capacity, maintained the quiescence of HSCs and promoted the expansion of mouse or human HSCs ex vivo 1,9,11,22,24,25 .
On the basis of our previous data, we inferred the regulatory function of TXNIP in HSC ageing 20,21 . In this study, we show that the loss of TXNIP induces the premature ageing of HSCs by elevating ROS production and inducing ageing-associated genes via upregulating p38 activity. We also show that TXNIP interacts with p38 via docking interaction and inhibits p38 activity in HSCs. Furthermore, we examine the potential of TXNIP-derived peptide to inhibit p38 activity to rejuvenate aged HSCs in vitro and in vivo. Altogether, we propose the novel functions of TXNIP in HSC ageing via regulating p38 activity and the possibility of the rejuvenation of aged HSCs via inhibiting p38 activity with TXNIP-derived peptide.

Results
Premature ageing of TXNIP À / À HSCs. To identify the function of TXNIP in HSCs, we confirmed the expression of TXNIP in various subpopulations of mouse bone marrow (BM) cells. In agreement with our previous data 20,21 , mRNA level of TXNIP was increased in LT-HSCs ( Supplementary Fig. 1a). Next, to determine the effect of TXNIP on HSC ageing, we analysed white blood cells (WBCs) in the PB of TXNIP þ / þ and TXNIP À / À mice at 2 (young), 6, 12 and 24 (old) months of age 3,6,26 . TXNIP À / À mice showed dramatically skewed differentiation to myeloid at the age of 12 months, even more than old TXNIP þ / þ mice ( Supplementary Fig. 1b). Ageing-associated phenotypes of 12-month-old TXNIP À / À mice in haematopoiesis were also observed in their LT-HSC frequency of LSKs in BM cells, which were comparable to those of old TXNIP þ / þ mice (Fig. 1a,b). Next, we analysed the frequency of BM cells and absolute number of LT-HSCs, ST-HSCs and MPPs in each mouse. TXNIP À / À mice showed the exhaustion of HSCs at 12-and 22-month age ( Supplementary Fig. 1c,d). Additionally, we investigated the ratio of WBCs in the PB of 12-month-old TXNIP þ / þ and TXNIP À / À female mice to confirm the ageing phenotype of female mice. Twelve-month-old TXNIP À / À female mice also showed markedly skewed differentiation to myeloid ( Supplementary Fig. 1e). These data strongly implied the possibility of HSC ageing in 12-month-old TXNIP À / À mice, similar to that in old TXNIP þ / þ mice. Next, to obtain direct evidence of HSC ageing in 12-month-old TXNIP À / À mice, we investigated levels of ROS, which are elevated with age, and the expression of ageing-associated genes in HSCs. We selected four representative genes from previous reports, including p16, p19, p21 and Wnt5a, which play critical roles in HSC ageing 1,13,27 . As expected, 12-month-old TXNIP À / À HSCs exhibited higher ROS (Fig. 1c) and an induction of p16, p19, p21 and Wnt5a comparable to that of old TXNIP þ / þ HSCs ( Fig. 1d-g).
To determine the roles of TXNIP on haematopoiesis, we administered 5-fluorouracil (5-FU), which targets cycling cells, leading to a transient leucopenia in the blood. In addition, we administered NAC (N-acetyl-L-cysteine), an antioxidant agent, to examine whether the increased ROS production contributes to the higher sensitivity of TXNIP À / À HSCs to 5-FU treatment 28,29 . TXNIP þ / þ mice recovered to normal status after 14 days in WBC counts ( Supplementary Fig. 1f), but all TXNIP À / À mice gradually died. Interestingly, NAC treatment fully rescued TXNIP À / À mice from 5-FU-induced leucopenia ( Supplementary Fig. 1g). 5-FU treatment highly induced the production of ROS in TXNIP À / À HSCs than TXNIP þ / þ HSCs and the induction of ROS was reversed by NAC treatment (Supplementary Fig. 1h). These results suggest that the increased production of ROS in TXNIP À / À HSCs may result in the defects in the repopulation capacity of HSCs under haematopoietic stress.
These data indicate that TXNIP may play a role in HSC ageing and that the loss of TXNIP may induce the premature ageing of HSCs by elevating ROS production and inducing ageingassociated genes. oxidative stress is p38. As previously noted, p38 plays crucial roles in regulating stress-induced signalling cascades and ageingassociated gene expression in HSCs 9,31 . To identify the role of ROS on p38 activation in TXNIP À / À HSCs, we administered NAC to 12-month-old TXNIP À / À mice. ROS level and p38 activation were decreased in HSCs by NAC treatment ( Supplementary Fig. 2a,b). To understand the relationship between TXNIP and p38 in HSCs, we first confirmed the expression of p38 isoforms (a, b, g and d) 22,23 . p38a was predominantly expressed and increased in LT-HSCs ( Supplementary Fig. 2c). Next, to confirm the relationship between TXNIP and p38 in HSCs ageing, we examined the levels of TXNIP and p38 activity in HSCs with age. TXNIP and p38 activity were increased in lin À cells and HSCs with age and a loss of TXNIP resulted in p38 activation in HSCs (Fig. 2a,b and Supplementary Fig. 2d-f).
TXNIP interacts with p38 directly in HSCs. Our results imply that the interaction between TXNIP and the p38 pathway may regulate p38 activity in HSCs. We first tested the direct interaction between TXNIP and p38 using immunoprecipitation and an in situ proximity ligation assay (PLA). These two proteins directly interacted in BM cells and HSCs (Fig. 2c,d).
Next, to investigate the effect of ROS on their interaction, we administered H 2 O 2 . TXNIP was quickly induced and then decreased, but p38 activity increased continuously up to 60 min in BM cells ( Supplementary Fig. 2g). The interaction between TXNIP and p38 was increased by H 2 O 2 treatment and ageing in HSCs (Fig. 2d). Glutathione S-transferase (GST) pull-down assay confirmed these results in TXNIP-and p38-overexpressed 293T cells ( Supplementary Fig. 2h).
To examine the importance of p38 kinase activity on their interaction, we constructed a kinase-dead dominant-negative mutant for p38a (p38AF) by replacing the Thr-Gly-Tyr motif (activating phosphorylation sites) with Ala-Gly-Phe 32 . p38AF and treatment with SB203580 did not affect p38 interaction ability with TXNIP ( Fig. 2e and Supplementary Fig. 2i). Previous studies have proposed that p38 complexed with its activator or substrate via a docking domain and recognized short docking motifs on the interaction partners. The docking motif contains basic residues and a hydrophobic-X-hydrophobic sub-motif (K/R 2-3 -X 1-6 -f-X-f) 33,34 . To determine the residues necessary for TXNIP-p38 interaction, we mutated four potential docking motifs in TXNIP by site-directed mutagenesis of hydrophobic  residues in sub-motifs. L290 and L292 residues of TXNIP were important for their interaction (Fig. 2f). p38 docking domain mutants decreased the interaction between TXNIP and p38 (Supplementary Fig. 2j and Fig. 2g) 34 . Furthermore, to confirm the unique interaction between TXNIP and p38 via a docking site, we mutated four residues of the TXNIP docking site, including basic residues, and mutated the Q120 (glutamine) of p38 to A (alanine). Their interaction was reduced markedly by these mutations (Supplementary Fig. 2k). These data demonstrate that TXNIP interacts with p38 directly via docking interaction and that its interaction may inhibit p38 activity in HSCs.

Rejuvenation of HSCs by p38 inhibition in vivo.
To understand the roles of p38 in TXNIP À / À HSCs in vivo, we crossed TXNIP À / À mice with p38 AF/ þ mice, which contained a dominant-negative allele and homozygous p38 AF/AF embryos that died on approximately day 11.5, to generate TXNIP À / À /p38 AF/ þ mice 32,35 . To investigate whether p38 inhibition is critical to rejuvenate the defects of aged HSCs, we isolated CD45.2 þ LT-HSCs from indicated mice and then transplanted with competitor BM cells (CD45.1 þ ) into lethally irradiated congenic recipients (CD45.1 þ ). Surprisingly, the engraftment and skewing of WBCs were dramatically restored to a level comparable to those of young TXNIP þ / þ HSCs in TXNIP À / À / p38 AF/ þ HSCs ( Fig. 3a,b). Also, a higher frequency of LT-HSCs returned to TXNIP þ / þ HSC levels in TXNIP À / À /p38 AF/ þ HSCs (Fig. 3c). SB203580 administration also had similar effects on TXNIP À / À and old HSCs. TXNIP À / À /p38 AF/ þ HSCs-received or long-term SB203580-administered recipients maintained low levels of p38 activity and ROS in their donorderived HSCs (Fig. 3d,e). The skewing of WBCs in TXNIP À / À mice at the age of 12 months was also restored by crossing with the p38 dominant-negative allele to a level comparable to that of young mice (Fig. 3f). Furthermore, TXNIP À / À /p38 AF/ þ mice survived with 5-FU treatment ( Supplementary Fig. 3). Collectively, our data identified the function of p38 with respect to TXNIP À / À HSC ageing and indicated the possible application of p38 inhibitors as a rejuvenating drug for aged HSCs in vivo.
TXNIP-derived peptide inhibits p38 activity. As noted in a previous report, TXNIP recombinant was insoluble 36 . Therefore, we prepared a GST-fused truncation mutant for TXNIP (150a.a-317a.a) (GST-TXNIP-T), and p38 was prepared with a His-tagged protein. GST-TXNIP-T interacted with His-p38 and inhibited p38 kinase activity in vitro ( Supplementary Fig. 4a,b). Next, to exclude pitfalls from the overexpression of fully cloned genes in cells, we designed a short peptide from the docking motif of TXNIP to target p38. We first generated green fluorescent protein (GFP) plus helix-forming peptide linkers, (EAAAK) 5 , and TXNIP-derived peptide clones ( Supplementary Fig. 4c). The lentiviral GFP-fused peptides were overexpressed in 293T cells, and the TN13 peptide with the highest affinity toward p38 was selected via immunoprecipitation assay ( Supplementary Fig. 4d).
To understand the complex between TN13 and p38, we designed a putative complex model from the peptide in the structure of p38 MAPK (PDB ID: 1LEW) 34 . In our putative model, TN13 interacted with the docking region of p38, and this interaction was very similar to those of other peptides ( Supplementary  Fig. 4e,f) 34 . Next, to investigate the physiological consequences of the interaction between TN13 and p38 in cells, we designed a fluorescein isothiocyanate (FITC)-linked peptide bearing a CPP.
To deliver the synthesized peptide into cells, an HIV TAT transduction domain sequence (YGRKKRRQRRR) was linked to the N-terminus of the TN13 peptide, and to monitor the cell-penetrating efficiency of the peptide, FITC was linked to the N-terminus of the TAT sequence (TAT-TN13) 37,38 . The interaction between TAT-TN13 and His-p38 was determined by isothermal titration calorimetry (ITC), which measures the binding equilibrium directly by determining the heat evolved on association of a ligand with its binding protein 39 , and their interaction was compared with TAT control ( Supplementary  Fig. 4g,h). To determine the specificity of TN13, we performed kinase assays for p38 isoforms in vitro. SB203580 completely inhibited kinase activity of two isoforms, p38a and p38b, but TN13 showed a specificity on p38a ( Supplementary Fig. 4i). It suggests that TN13 may have high specificity on p38a isoform than SB203580 and this specificity may reduce the side effects in vitro and in vivo.
TAT-TN13 rejuvenates aged HSCs in vitro. We confirmed the cell-penetrating efficiency of TAT-TN13 into HSCs using confocal images ( Supplementary Fig. 5a). Penetrated TAT-TN13 markedly inhibited p38 phosphorylation in old BM cells and old HSCs and was comparable to SB203580 (Fig. 4a,b). To identify the inhibitory mechanism of TAT-TN13, we investigated the interaction between p38 and its upstream kinases, MKK3 and MKK6, during TAT-TN13 treatment. TAT-TN13 treatment inhibited the interaction between p38 and MKK3 or MKK6 in 293T cells ( Supplementary Fig. 5b,c). These results imply that TAT-TN13 shares the docking region of p38 for MKK3 or MKK6 and inhibits p38 activity via blocking the interaction between p38 and its upstream kinases. Our peptide also competed with TXNIP for interaction with p38 in BM cells and HSCs (Fig. 4c,d). These data also supported the docking interaction between TXNIP and p38. As shown in Fig. 2, our data suggested that the increased p38 activity with age was a critical change that resulted in HSC defects. We hypothesized that p38 inhibition in old HSCs via TAT-TN13 treatment may return HSCs to young states or at least improve their functional defects, as shown via SB203580 administration. To confirm the rejuvenating potential of TAT-TN13, we treated old HSCs with TAT-TN13 (10 mM) for 16 h in vitro. As expected, elevated p38 activity in old HSCs was dramatically reduced by TAT-TN13 treatment to levels comparable to those of SB203580 treatment, and ROS levels were decreased in TAT-TN13-treated old HSCs (Fig. 4e,f). Recently, Geiger and his colleagues reported the role of Cdc42 activity in the cell-intrinsic ageing of HSCs. The authors proposed that the polarity of Cdc42 is a marker for differentiating young and old phenotypes of HSCs and that a pharmacological reduction of Cdc42 activity rejuvenates aged HSC function 4,13 . Therefore, we used the polarity of Cdc42 as an indicator of the rejuvenating phenotype of aged HSCs. After TAT-TN13 treatment for 16 h, HSCs were fixed and stained with Cdc42 antibody. Most depolarized old HSCs returned to polarized HSCs (Fig. 4g), and the expressions of ageing-associated genes were remarkably decreased in TAT-TN13-treated old HSCs (Fig. 4h-k). In addition, ageing-associated gene expression exhibited similar patterns after competitive transplantation assay in vivo ( Supplementary Fig. 6a-d). Furthermore, TAT-TN13 treatment increased the homing ability of old HSCs after short-term transplantation (Fig. 4l). Taken together, our results indicate that the regulation of p38 activity might be a critical tool to rejuvenate aged HSCs, and we confirmed the potential of TAT-TN13 to inhibit p38 activity to rejuvenate aged HSCs in vitro.

Rejuvenation of aged HSCs by GFP-TN13 and TAT-TN13 in vivo.
HSCs were isolated from mice and transduced three times with GFP-TN13-expressing lentiviral vector for 36 h. After incubation in HSC media, sorted GFP þ HSCs were competitively transplanted into recipients (Fig. 5a). GFP-TN13 expression in 12-month-old TXNIP À / À or old HSCs exhibited restoration of their ageing phenotypes (Fig. 5b-d). Elevated p38 activity in 12-month-old TXNIP À / À and old HSCs was dramatically reduced by GFP-TN13 expression to levels comparable to those of young HSCs, and ROS levels were decreased in GFP-TN13transduced 12-month-old TXNIP À / À and old HSCs (Fig. 5e,f). Next, to investigate the possible application of TAT-TN13 as a rejuvenating drug for aged HSCs in vivo, we performed a competitive transplantation of TAT-TN13-treated old HSCs or 12-month-old TXNIP À / À HSCs. Both of them exhibited restoration of aged phenotypes of HSCs comparable to SB203580 treatment (Fig. 6a-c) and maintained low levels of p38 activity and ROS (Fig. 6d,e). To assess the effect of TAT-TN13 on haematopoietic stress in old HSCs, 5-FU (100 mg kg À 1 ) was i.p. injected into young and old mice, and on the next day, TAT-TN13 (25 mg kg À 1 ) was administered via i.p. injection daily for 4 days. Surprisingly, TAT-TN13 administration restored the WBCs of old mice better than those of young mice at early time points and also saved their lives (Supplementary Fig. 7). Overall, these results demonstrated the potential of GFP-TN13 or TAT-TN13 as an applicable therapeutic tool for rejuvenating aged HSCs in vivo (Fig. 7).

Discussion
Although stem cell ageing is complex and context dependent, stem cells apart from their environment share certain conserved molecular mechanisms in their ageing process. Recent studies have shown that impaired functions of aged stem cells were restored by pharmacological treatments, and these findings indicated new therapeutic targets for rejuvenating stem cells 1,4 .
A potentially promising target molecule in stem cell ageing is p38, which is activated by ROS in aged stem cells. p38 activation ARTICLE is induced by various pathological conditions or cellular ageing and results in HSC defects 1,4,6,40 . SB203580 administration restored the repopulation capacity, maintained the quiescence of HSCs and promoted the expansion of mouse and human HSCs ex vivo 1,25 . p38 was also activated in muscle stem cells from aged mice, and p38 inhibition has been shown to restore defects of muscle stem cells 41,42 . In human tissue-derived mesenchymal stem cells and endometrium-derived mesenchymal stem cells, p38 plays an important role in cellular senescence, and the pharmacological inhibition of p38 abrogated ageing phenotypes 43,44 . As noted above, the regulation of p38 activity appears to be a promising target for stem cell rejuvenation.
In this study, we analysed the functions of TXNIP on HSC ageing with TXNIP À / À mice that showed premature ageing phenotypes of HSCs. Ageing of TXNIP À / À HSCs was mostly due to the elevation of ROS and the induction of p38 activity. We identified the direct interaction between TXNIP and p38 via docking region of p38. We investigated the cellular function of p38 activity in TXNIP À / À HSCs and old HSCs using p38 chemical inhibitor and TXNIP À / À /p38 AF/ þ mice in vivo. From these data, we provided information that the activation of p38 in TXNIP À / À HSCs was major cause of HSC ageing and the inhibition of p38 activity in HSCs could rejuvenate the ageing phenotypes of aged HSCs. The HSC ageing is initiated by DNA damage such as deletions and mutation, epigenetic alterations and altered expression of certain key transcription factors. Deregulation of these intrinsic factors drive HSCs into physiological ageing with combinatorial effects of these alterations. This implies HSC ageing seems to be reversible and can be reprogrammed by modulating these key factors. The rejuvenation of aged HSCs is the reversal of these alterations and functional restoration of aged HSCs. Several approaches were tried and showed the evidences of at least partial HSC rejuvenation. Cdc42 inhibition restored the level of H4K16 acetylation of aged HSCs to that of young HSCs 4 . The overexpression of Sirt3 in aged HSCs decreased ROS level with partial recovery of all lineages 15 . mTOR inhibitor, rapamycinreduced HSC numbers and increased reconstitution potential with balanced haematopoietic precursors 45,46 . Besides intrinsic factors, prolonged fasting restored frequency of myeloid-biased HSC population through reducing IGF-1 and protein kinase A activity 16 . The rejuvenation of HSCs will eventually improve and prevent the risks of ageing-associated diseases, organ dysfunction, malignancy and cancer.
Recently, many researchers have discussed the potential role of CPPs in the intracellular delivery and confirmed their efficiency in vitro and in vivo 47 . Especially, HIV TAT protein transduction domain was prominent in delivering efficiency in human and mouse haematopoietic cells 38,[48][49][50] . And also some of groups proposed the TAT-conjugated peptide inhibitors for p38 and showed possible uses in vitro and in vivo 33,37 . Here, to develop our own therapeutic method to rejuvenate the aged HSCs, we designed CPP-conjugated peptide (TAT-TN13). TAT-TN13 inhibited the p38 activity efficiently via direct interaction and rejuvenated aged HSCs in vitro. Finally, we demonstrated the possibility of TAT-TN13 on rejuvenation of aged HSCs in vivo and TAT-TN13 had comparable effect to SB203580 and also rescued the old mice from 5-FU administration.
In conclusion, we demonstrated that TXNIP plays a crucial role in HSC ageing by inhibiting p38 activity via direct interaction. Overall, these data showed the possibility of TAT-TN13 derived from the docking motif of TXNIP as an applicable therapeutic drug for the rejuvenation of aged HSCs in vivo.
Male mice were used for overall study and female mice were used only in Supplementary Fig. 1e as a control  Competitive transplantation. For competitive repopulation assays, LT-HSCs were isolated from young or old mice (CD45.2 þ ). 400-500 LT-HSCs were i.v. injected with competitor BM cells (CD45.1 þ , 1.0 Â 10 6 or 1.5 Â 10 6 ) into lethally irradiated (9 Gy) CD45.1 þ congenic recipients (6-8 weeks old). The repopulation of donor-derived cells was monitored by staining PB from tail vein and BM cells with antibodies against indicated surface markers after 16 weeks.
Recombinant constructs. For GST pull-down assay, we constructed TXNIP and p38a mutants using site-directed mutagenesis. Human TXNIP or p38a clone was used as a template. For TXNIP mutants construction, we used the primers as follows: In vitro kinase assay. GST, GST-TXNIP (150-317) and His-p38a were purified using affinity chromatography. To examine the kinase activity of p38 in vitro, we used p38 MAP Kinase assay kit (9820, Cell Signaling). We added 0.5-1 mg of His-p38a for kinase assay and its activity was determined by phospho-ATF2 levels.
ITC assay. For ITC assay, all measurements were carried out at 25°C on a microcalorimetry system iTC200 (GE Healthcare). p38a and peptides were dialysed against a solution containing PBS. The samples were centrifuged to remove any residuals before the measurements. The experiments were carried out by titrating 1,050 mM TAT-TN13 peptide into 57.99 mM p38a protein. Dilution enthalpies were determined in separate experiments (titrant into buffer) and subtracted from the enthalpies of the binding between the proteins. Data were analysed using the Origin software (OriginLab).
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information files or from the corresponding author upon reasonable request.