Defects in TRPM7 channel function deregulate thrombopoiesis through altered cellular Mg2+ homeostasis and cytoskeletal architecture

Mg2+ plays a vital role in platelet function, but despite implications for life-threatening conditions such as stroke or myocardial infarction, the mechanisms controlling [Mg2+]i in megakaryocytes (MKs) and platelets are largely unknown. Transient receptor potential melastatin-like 7 channel (TRPM7) is a ubiquitous, constitutively active cation channel with a cytosolic α-kinase domain that is critical for embryonic development and cell survival. Here we report that impaired channel function of TRPM7 in MKs causes macrothrombocytopenia in mice (Trpm7fl/fl-Pf4Cre) and likely in several members of a human pedigree that, in addition, suffer from atrial fibrillation. The defect in platelet biogenesis is mainly caused by cytoskeletal alterations resulting in impaired proplatelet formation by Trpm7fl/fl-Pf4Cre MKs, which is rescued by Mg2+ supplementation or chemical inhibition of non-muscle myosin IIA heavy chain activity. Collectively, our findings reveal that TRPM7 dysfunction may cause macrothrombocytopenia in humans and mice.

P latelets are continuously produced from megakaryocytes (MKs) in the bone marrow by a cytoskeleton-driven process of which the molecular regulation is not fully understood. MKs extend long cytoplasmic protrusions into bone marrow sinusoids, where larger fragments, so-called preplatelets, are shed and further divide within the circulation to give rise to platelets (Supplementary Movie 1) [1][2][3] .
Transient receptor potential melastatin-like 7 (TRPM7) channel and kinase domain, but not its kinase activity, are critical for embryonic development [4][5][6] and knockdown or cell-specific TRPM7 knockout approaches give rise to impaired cytoskeletal organization, cell migration, proliferation, polarization and survival. These defects could partially be explained by increased non-muscle myosin IIA heavy chain (NMMIIA)-mediated contractility of the actin cytoskeleton 5,[7][8][9][10][11][12][13] . Of note, among other substrates, the kinase domain of TRPM7 phosphorylates annexin I and NMMIIA, thus interfering with cell survival and cytoskeletal rearrangements 14,15 . Interestingly, several variants of NMMIIA similarly altered the contractility of the actomyosin complex in MKs, thereby interfering with proplatelet formation in humans and mice 16 . During megakaryopoiesis, NMMIIA activity is suppressed by phosphorylation of its C-terminus, enabling MK polyploidisation and ultimately proplatelet formation 17 . However, for proper platelet fission and sizing, NMMIIA needs to be re-activated under shear stress in the circulation 16,18 . Although both kinase and channel activity of TRPM7 have been proposed to regulate cytoskeletal dynamics, channel activity alone was sufficient to restore cell polarization, morphology and migration 10,13,19 , suggesting a critical role of cations therein. Consequently, the differential role of TRPM7 channel versus kinase activity in the regulation of the cytoskeleton still remains unclear. Moreover, TRPM7 has been implicated as a key regulator of signal conductance in the murine heart by regulating the expression of different pacemaker channels, such as HCN4 (ref. 20). Although TRPM7-mediated cation influx has been detected in MKs (ref. 21), its role in thrombopoiesis has not been investigated to date.
Here we report that impaired channel function but not kinase activity of TRPM7 in MKs causes macrothrombocytopenia in Trpm7 fl/fl-Pf4Cre mice and likely in several members of a human pedigree, which, in addition, feature atrial fibrillation. The impaired proplatelet formation is associated with cytoskeletal alterations due to increased actomyosin contractility and can be rescued by either Mg 2 þ supplementation or chemical inhibition of NMMIIA activity. Collectively, our findings reveal TRPM7 dysfunction as a novel cause of macrothrombocytopenia in mice and potentially in humans too.

Results
Defects in TRPM7 cause macrothrombocytopenia. We identified TRPM7 as the key Mg 2 þ channel and magnesium transporter 1 (MagT1) to be expressed in murine platelets ( Supplementary Fig. 1a) and generated MK-and platelet-specific TRPM7 knockout mice ( Supplementary  Fig. 1b,c). The absence of TRPM7 currents in patch clamp measurements confirmed the efficacy of the targeting strategy in primary bone marrow MKs ( Supplementary Fig. 1d). Unexpectedly, these mice displayed a macrothrombocytopenia (Fig. 1a,b) with enlarged and spherical platelets, often containing large vacuoles as revealed by electron microscopy (Fig. 1c). In contrast, mice carrying a kinase-dead K1646R mutation in Trpm7 (ref. 6; Trpm7 KI ) showed normal platelet counts, size and morphology, thus suggesting that the lack of TRPM7 channel function rather than its kinase activity accounts for the macrothrombocytopenia in the mutant mice ( Supplementary  Fig. 2a-e). In line with this notion, intracellular Mg 2 þ concentrations in Trpm7 fl/fl-Pf4Cre platelets, but not in Trpm7 KI platelets, were decreased ( Fig. 1d; Supplementary Fig. 2f) 6 .

Impaired proplatelet formation causes thrombocytopenia.
A mildly reduced platelet lifespan in Trpm7 fl/fl-Pf4Cre mice (t½ ¼ 43.6 h for wild type (WT) versus t½ ¼ 35.7 h for Trpm7 fl/fl-Pf4Cre mice), however, was insufficient to explain the reduced platelet count and was not associated with altered platelet terminal galactose levels ( Supplementary Fig. 3a,b). Immunostaining of whole-femora bone marrow sections (Fig. 2a) revealed an increased number of MKs in the mutant mice (6.3±0.3 for WT versus 13.3±1.6 for Trpm7 fl/fl-Pf4Cre mice; Fig. 2b). The MKs in Trpm7 fl/fl-Pf4Cre mice were also located further from bone marrow sinusoids than in controls (Fig. 2c) suggesting impaired migration of MK-precursors to bone marrow sinusoids. Although splenomegaly was not observed in Trpm7 fl/fl-Pf4Cre mice, we found an increased number of MKs in   Fig. 4a-d).
In contrast to a previous report on a neuroblastoma cell line 22 , the formation of podosomes, F-actin rich structures that are thought to serve cell migration and proplatelet protrusion 23 , was unaltered in Trpm7 fl/fl-Pf4Cre MKs ( Supplementary Fig. 5a) suggesting that other defects must account for their more distant localization from bone marrow sinusoids (Fig. 2c). Interestingly, mutant MKs displayed an increased mean ploidy as compared with controls (16.9 N ± 1.6 N in controls versus 22.4 N±3.4 N for Trpm7 fl/fl-Pf4Cre mice; Fig. 2d) thus excluding impaired MK maturation as the cause of thrombocytopenia. Despite the increased ploidy in vivo, we found a decreased proplatelet formation for both foetal liver- (Fig. 2e) and bone marrow-derived ( Supplementary Fig. 5b) Trpm7 fl/fl-Pf4Cre MKs in vitro. This was further confirmed in vivo by intravital  ARTICLE two-photon microscopy of the bone marrow (0.88% min À 1 ± 0.16% min À 1 of WT MKs formed proplatelets versus 0.23% min À 1 ±0.12% min À 1 of Trpm7 fl/fl-Pf4Cre MKs), which revealed that Trpm7 fl/fl-Pf4Cre MKs preferentially formed short and bulky proplatelet protrusions (0.85% min À 1 ± 0.13% min À 1 of observed MKs in Trpm7 fl/fl-Pf4Cre mice versus 0.05% min À 1 ± 0.04% min À 1 in WT mice) that remained attached to the cell body during the observation period ( Fig. 2f,g; Supplementary Movies 1-3). The actin and microtubule cytoskeleton are critical for proper proplatelet formation 2 . We therefore analysed the cytoskeletal architecture of Trpm7 fl/fl-Pf4Cre MKs and found an increased content and aberrant organization of microtubules in proplatelet-forming, resting and spread MKs ( Fig. 2h; Supplementary Fig. 6a-c). In support of this, extraction of the microtubule cytoskeleton of resting bone marrow-derived MKs by ultracentrifugation revealed an increased microtubule content in the Triton X100 insoluble pellet fraction ( Supplementary  Fig. 6d,e). Of note, increased microtubule stability is associated with post-translational modifications of a-tubulin that include detyrosination (Glu-tub) and acetylation (ac-tub) which could account for the increased number of microtubules in Trpm7 fl/fl-Pf4Cre MKs (ref. 24). Nonetheless, we found an increased prevalence of highly dynamic tyrosinated (Tyr) microtubules as evidenced by lower Glu-/Tyr-(2.5 ± 0.3 for control versus 1.5±0.2 for Trpm7 fl/fl-Pf4Cre MKs) and ac-/Tyr-tubulin ratios (2.2 ± 0.2 for control versus 1.6 ± 0.2 for samples of Trpm7 fl/fl-Pf4Cre mice), suggesting that accelerated microtubule assembly/dynamics accounted for the observed alterations ( Supplementary Fig. 6d-g). Moreover, in line with the critical role of microtubules in trafficking of intracellular cargo, electron microscopy revealed a non-homogeneous distribution of granules, tortuous membrane complexes and aberrantly sized proplatelets in mutant MKs ( Fig. 2i; Supplementary Fig. 7). Furthermore, we found thick and densely packed proplatelets in bone marrow sinusoids with signs of apoptosis reflecting impaired proplatelet fragmentation and release of preplatelets from Trpm7 fl/fl-Pf4Cre MKs (  Supplementary Fig. 8). The pronounced increase above initial platelet counts might be attributed to the increased number of MKs in Trpm7 fl/fl-Pf4Cre mice and alternative platelet release mechanisms as recently shown by Nishimura et al. 25 .
Altered NMMIIA activity impairs proplatelet formation. NMMIIA has been described as a downstream effector of TRPM7 kinase 14,22 and importantly, abnormal function of NMMIIA has been associated with impaired formation and fragmentation of proplatelets in humans and mice 16 . Moreover, Mg 2 þ was recently shown to modify NMMIIA activity by regulating ADP release and its affinity to actin filaments, thus further linking TPRM7 channel to NMMIIA function 26 . Strikingly, analysis of NMMIIA localization in MKs on whole-bonemarrow sections in situ revealed a homogeneous distribution in the cell body of control (92.1 ± 2.0% in WT versus 14.7 ± 1.1% in Trpm7 fl/fl-Pf4Cre mice), while it predominated in the cell cortex of Trpm7 fl/fl-Pf4Cre MKs (0.4 ± 0.8% in WT versus 48.1 ± 3.6% in Trpm7 fl/fl-Pf4Cre mice; Fig. 3a,b). A similar distribution pattern was found in vitro for foetal liver-derived MKs where in addition a significant accumulation of NMMIIA in proplatelets of mature WT MKs was observed (Fig. 3c). Moreover, NMMIIA localization was also altered in Trpm7 fl/fl-Pf4Cre platelets. However, in contrast to MKs, NMMIIA predominated in the cell cortex of WT platelets reminiscent of the marginal band, whereas it was homogeneously distributed in platelets of Trpm7 fl/fl-Pf4Cre mice ( Supplementary Fig. 9a-c). This observation is in line with a report on resting foreskin fibroblasts in which NMMIIA and a-tubulin staining overlapped extensively at the cell cortex under resting, non-contractile conditions. However, on induction of cell migration and activation of NMMIIA, the overlap of NMMIIA and a-tubulin staining decreased, allowing NMMIIA to exert its contractile effects on the actin cytoskeleton 27 . To further analyse the cause of the absent NMMIIA staining and the altered localization to the cell cortex in Trpm7 fl/fl-Pf4Cre MKs in situ, we next allowed bone marrow-derived MKs to spread on a collagen type I-coated surface. Surprisingly, on spreading of Trpm7 fl/fl-Pf4Cre MKs (Fig. 3d,e) or platelets (Supplementary Fig. 10a-d) we found a rapid degradation of NMMIIA that could be rescued by pretreatment with the NMMIIA inhibitor blebbistatin or by Mg 2 þ supplementation ( Fig. 3e-g; Supplementary Fig. 10e-h). Based on these findings we speculated that deregulated [Mg 2 þ ] i in Trpm7 fl/fl-Pf4Cre cells may cause an increased activity of NMMIIA resulting in its rapid degradation on cell stimulation. In agreement with the increased co-localization of NMMIIA to actin filaments in resting Trpm7 fl/fl-Pf4Cre platelets (Supplementary Fig. 9), sedimentation of cross-linked actin filaments of resting platelets revealed an increased amount of NMMIIA in the pellet fraction ( Supplementary Fig. 11a-c). On activation, NMMIIA efficiently cross-linked actin filaments and accumulated in the pellet fraction of WT platelets ( Supplementary Fig. 11d-f). In sharp contrast, we found a marked reduction in NMMIIA on stimulation of Trpm7 fl/fl-Pf4Cre platelets (Supplementary Fig. 11d-f), similar to the findings in spread MKs (Fig. 3c) or platelets ( Supplementary Fig. 10a-d). To exclude that the observed effects were due to the short observation period, bone marrow-derived MKs were cultured for 48 h in the presence of collagen types I or IV, two major components of the extracellular matrix in the bone marrow. Strikingly, we found a reduced content of NMMIIA in Trpm7 fl/fl-Pf4Cre MKs (reduction of 39.9 ± 10.3% for collagen I and 76.9 ± 14.5% for collagen IV) cultured in the presence of different collagens as compared with control cells (Fig. 3h; Supplementary Fig. 12a,b). Together these results suggested an increased NMMIIA activity under resting conditions and that NMMIIA undergoes a rapid degradation on stimulation of platelets and MKs from Trpm7 fl/fl-Pf4Cre mice thus allowing proplatelet formation and cell spreading.
Deregulated [Mg 2 þ ] i perturbs NMMIIA activity. We hypothesized that deregulated Mg 2 þ homeostasis 26 and increased NMMIIA activity may account for the impaired proplatelet formation in Trpm7 fl/fl-Pf4Cre MKs (refs 16,26). In support of this, either inhibition of NMMIIA activity or Mg 2 þ supplementation could almost fully restore proplatelet formation of Trpm7 fl/fl-Pf4Cre MKs in vitro (Fig. 4a,b). Of note, while pretreatment of foetal liver- (Fig. 4c) or bone marrow-derived MKs (Supplementary Fig. 13) with the Ca 2 þ chelators 1,2-bis (2-aminophenoxy) ethane-N,N,N 0 ,N 0 -tetraacetic acid tetrakis (acetoxymethyl ester) (BAPTA-AM) or ethylene glycol tetraacetic acid (EGTA) did not exert gross effects; non-specific chelation of Ca 2 þ and Mg 2 þ with ethylenediaminetetraacetic acid (EDTA) significantly reduced proplatelet formation by MKs, suggesting that Mg 2 þ is critical for this process 28,29 . According to a previous report 8 , Mg 2 þ supplementation should restore [Mg 2 þ ] i in mutant MKs, which cannot be fully achieved through an upregulation of MagT1 expression under normal culture conditions (Fig. 4d,e). were efficiently disassembled on cold challenge, thus excluding enhanced stability as cause for the increased microtubule content (Fig. 5e-h). Interestingly, pretreatment of WT platelets with EDTA, but not EGTA, BAPTA-AM or MgCl 2 , mimicked these cytoskeletal alterations ( Fig. 6a-c), changes that could also be reverted by blebbistatin ( Fig. 6d-f), thus further supporting the notion that reduced [Mg 2 þ ] i alters the subcellular localization and activity of NMMIIA resulting in cytoskeletal disorganization.
In agreement with this, we observed an increased content of filamentous actin in resting Trpm7 fl/fl-Pf4Cre platelets ( Supplementary Fig. 15a,b) and a decreased polymerization of filamentous actin on platelet activation ( Fig. 6g; Supplementary  Fig. 15b,c). Moreover, spread Trpm7 fl/fl-Pf4Cre platelets displayed an increased surface area ( Fig. 6h; Supplementary Fig. 16) most likely reflecting the activation-dependent rapid degradation of NMMIIA and consequently the loss of coherent cytoplasmic contractile forces normally generated by activated NMMIIA (refs 31,32).
Variants in TRPM7 may cause macrothrombocytopenia. We next hypothesized that some DNA variants of extreme low-frequency affecting TRPM7 channel function might also cause macrothrombocytopenia in humans.  Supplementary  Fig. 2). Unfortunately, the family with the p.R902C variant was unavailable for further studies.
Further studies were focussed on the index case UCN 0012 with a p.C721G (c.2161T4G) variant and pedigree members. Sanger sequencing showed that the p.C721G variant was present in two further pedigree members with macrothrombocytopenia, but was absent in one asymptomatic pedigree member, indicating segregation with the TRPM7 genotype (Fig. 7a, Supplementary Table 2 and Supplementary Fig. 17). For the fourth patient (pedigree member 2), now deceased, macrothrombocytopenia was detected during her life (Fig. 7a). All other blood cell parameters and platelet function were normal for all p.C721G patients. Strikingly, paroxysmal atrial fibrillation was also present for the index case (pedigree member 5) and her mother (pedigree member 2).
Similarly to Trpm7 fl/fl-Pf4Cre mouse platelets, we found a reduced content of Mg 2 þ and an increased concentration of Ca 2 þ in platelets from all tested patients with the p.C721G substitution as compared with healthy controls (Fig. 7b). Patch clamp studies on HEK293 cells confirmed that the p.C721G variant reduced TRPM7 channel activity by 85 ± 4% as compared with WT controls (Fig. 7c) despite being localized to the cell membrane ( Supplementary Fig. 18). Likewise, in vitro studies on the p.R902C variant revealed, although less pronounced, a reduced TRPM7 channel activity by 39 ± 6% ( Supplementary  Fig. 19). Platelets from all tested patients displayed an increased size, with a spherical shape and contained numerous vacuoles. Moreover, while platelets from controls displayed the typical discoid shape and microtubules organized into coils, the marginal band, platelets from the p.C721G patients showed an aberrant distribution of granules and an increased number and anarchic organization of microtubules ( Fig. 7d; Supplementary  Fig. 20). The abnormal cytoskeletal organization could be confirmed by immunostaining on resting p.C721G platelets (Fig. 7e,f; Supplementary Figs 21 and 22) and similarly to the Trpm7 fl/fl-Pf4Cre platelets, which was not associated with increased microtubule stability ( Fig. 7g; Supplementary Fig. 21).
Altered regulation of NMMIIA in p.C721G platelets. Since the ultrastructure of human platelets from individuals carrying the p.C721G variant strongly resembled that of Trpm7 fl/fl-Pf4Cre mouse platelets, we next analysed the distribution and stability of NMMIIA. Whereas in control platelets NMMIIA localized to the marginal band, it was homogeneously distributed throughout the cytoplasm of platelets from the patients (Fig. 8a,b; Supplementary  Fig. 22). In line with the observations on Trpm7 fl/fl-Pf4Cre mouse platelets, spread platelets from the patients showed an increased surface area, a strong decrease in NMMIIA and an increased content of microtubules (Fig. 8c,d; Supplementary Figs 23  and 24). As for the mouse model, blebbistatin prevented loss of NMMIIA on spreading (Fig. 8d,e; Supplementary Fig. 25) and rescued the cytoskeletal organization of resting platelets after cold challenge (Fig. 8f,g; Supplementary Fig. 26). As mentioned above, patients 2 and 5 suffered from atrial fibrillation which might be associated with alterations in [Mg 2 þ ] i (ref. 33). Furthermore, lack of TRPM7 has been associated with altered expression of channels, such as HCN4 encoding the pacemaker current in the conduction system of the heart that has also previously been linked to atrial fibrillation 20,34 .

Discussion
In the present study, we provide evidence that defects in TRPM7 channel function cause macrothrombocytopenia in mice and likely in humans too. These results demonstrate a critical role of TRPM7-mediated Mg 2 þ influx in regulating NMMIIA activity and cytoskeletal rearrangements during thrombopoiesis. In support of this, it has been shown that Mg 2 þ inhibits NMMIIA activity by reducing the ADP release rate and its affinity for actin 26 . Furthermore, using Trpm7 KI mice ( Supplementary  Fig. 2), we convincingly show that these effects occur independently of TRPM7 a-kinase activity. However, it is too early to exclude that the kinase domain also acts as a molecular hub orchestrating yet unknown signalling events, and thereby contributes to the development of macrothrombocytopenia.
The observed loss of NMMIIA in Trpm7 fl/fl-Pf4Cre bone marrow MKs in situ and in vitro on spreading could represent a physiological process to overcome the inhibitory effects of increased NMMIIA activity enabling proplatelet formation (Fig. 3a,b,d-h; Supplementary Figs 10-12). In support of this hypothesis, NMMIIA degradation was most pronounced on culture of bone marrow-derived MKs on collagen type IV, which predominates in the vascular niche where proplatelet formation occurs (Fig. 3a,b,h; Supplementary Fig. 12). Furthermore, using different antibodies and experimental approaches, we provide several independent lines of evidence that the observed loss of NMMIIA on activation or spreading of TRPM7-deficient cells represents protein degradation. However, it is important to note It has recently been reported that plasma Thpo levels are also regulated via Ashwell-Morell receptor-dependent clearance of desialylated platelets, which in turn triggers expression of Thpo in hepatocytes and thus regulates platelet production 35 . However, it is important to note that in agreement with the moderately reduced platelet lifespan, we did not observe differences in platelet terminal galactose levels suggesting that the decreased plasma Thpo levels are reciprocally associated to the increased number of MKs in the bone marrow and spleen (Fig. 2a,b; Supplementary Figs 3 and 4) 36 . Furthermore, the marked increase in platelet counts after immuno-depletion might result from an alternative platelet release mechanism as recently described by Nishimura et al. 25 .
The here-described human disorder ( Fig. 7; Supplementary Tables 1 and 2) is clearly distinguishable from MYH9-related platelet disease, which can be associated with hearing loss, cataracts or renal failure 37 . However, patients with variants in TRPM7 and MYH9 both display macrothrombocytopenia with more spherical platelets and partially, increased actomyosin contractility (Figs 7 and 8; Supplementary Figs 21-26) 16 . The different clinical outcomes and symptoms of patients with different genetic variants of TRPM7 are most likely due to the location and functional consequences of the mutations (Supplementary Table 1). While both the p.C721G and the p.R902C variant are located in the channel domain and reduce channel activity ( Fig. 7c and Supplementary Fig. 19), the p.G1353D variant is not located within the channel domain which may likely explain the distinct phenotype and absence of macrothrombocytopenia in the index patient UCN 0025. Similar genotype-phenotype-relationships have been observed for patients suffering from MYH9-related disorder 38,39 , Wiskott-Aldrich syndrome 2,40 or filaminopathy A 41 . Although we could not observe a dominant negative effect of the p.C721G variant on TRPM7 channel function in vitro, we speculate that the defect might be masked by the highexpression levels of TRPM7 in the used system or diminished channel activity may not represent the only mechanistic explanation for the observed phenotypes. Furthermore, Trpm7 þ / À mice did not display macrothrombocytopenia suggesting that gene haploinsufficiency does not account for the observed disease ( Supplementary Fig. 27). However, the impaired channel activity and the striking phenotypic similarities in platelets from patients and Trpm7 fl/fl-Pf4Cre mice strongly suggest that the variants in TRPM7 account for the observed macrothrombocytopenia. Our study provides for the first time several independent lines of evidence that proper regulation of Mg 2 þ homeostasis in MKs by TRPM7 plays a critical role in thrombopoiesis and platelet sizing, both in humans and mice. Collectively, our results highlight the clinical need to carefully control platelet count and size in patients with deregulated Mg 2 þ homeostasis. Ultimately, our findings suggest, after careful consideration and assessment of adverse events, that Mg 2 þ supplementation may be used as a potential treatment of patients with increased activity of NMMIIA to manage thrombocytopenia. Despite the possibility of severe potential side effects in patients with impaired kidney function, Mg 2 þ supplementation is considered as a relatively safe therapeutic intervention [42][43][44] . Nevertheless, further studies in animal models and patients are required to assess the efficacy and safety of Mg 2 þ supplementation in the management of certain cases of thrombocytopenia.
Finally, the fact that two of the patients in our study also suffered from paroxysmal atrial fibrillation promotes TRPM7 as a novel candidate interfering with conductance in cardiac cells.

Methods
Semi-quantitative reverse transcription PCR. Total platelet RNA was isolated after lysis using TRIzol reagent (15596018, Invitrogen) and tissue RNA was gained using the Qiagen RNeasy kit. To generate cDNA 1 mg RNA was reverse transcribed with the SuperScript II reverse transcriptase (18064014, Invitrogen) according to the manufacturer's instructions. b-actin transcripts were determined as control. Primers used for the amplification of Mg 2 þ transporters and channels are listed in Supplementary Table 3.
Mice. Conditional Trpm7-deficient mice were generated by intercrossing Trpm7 fl/fl mice (exon 17 flanked by loxP sites) with mice carrying the Cre-recombinase under the platelet factor 4 (Pf4) promoter 45 (Supplementary  Fig. 1b). Trpm7 fl/fl mice were described earlier 4 , Pf4-Cre mice were kindly provided by Dr Radek Skoda and Trpm7 KI mice were generously provided by Dr Masayuki Matsushita 6 . All mice used in experiments were 12-to 16-week-old and sex-matched, if not stated otherwise. For experiments on MKs, only male animals were used. All animal experiments were approved by the district government of Lower Franconia (Bezirksregierung Unterfranken). Trpm7 fl/fl-Pf4Cre mice were genotyped via PCR on genomic DNA extracted from ear biopsies using the following primer pair: Trpm7KO-F: 5 0 -gaggtactggcaattgtgagc-3 0 and Trpm7KO-R: 5 0 -accacaaaatctctgccctct-3 0 yielding a 1,200-bp product for the floxed, 1000 bp product for the WT and a 400-bp fragment for the recombined allele ( Supplementary Fig. 1c). For all experiments, the respective Trpm7 fl/fl (for Trpm7 fl/fl-Pf4Cre mice) or Trpm7 WT/WT (for Trpm7 KI/KI mice) littermate controls were used. All mice were derived from the following breeding strategies for Trpm7 fl/fl X Trpm7 fl/fl-Pf4Cre , yielding 50% Trpm7 fl/fl WT and 50% Trpm7 fl/fl-Pf4Cre mice or Trpm7 WT/KI X Trpm7 WT/KI resulting in 25% Trpm7 WT/WT , 50% Trpm7 WT/KI and 25% Trpm7 KI/KI mice, respectively.
Platelet preparation. Mice were bled under isoflurane anaesthesia. Blood was collected in heparin (20 U ml À 1 , Ratiopharm) and centrifuged twice for 6 min at 300 g. Platelet-rich plasma (PRP) was supplemented with 2 ml ml À 1 of apyrase ARTICLE (0.02 U ml À 1 ; A6410, Sigma-Aldrich) and 5 ml ml À 1 prostacyclin I2 (PGI 2 ) (0.1 mg ml À 1 ; P6188, Sigma-Aldrich) and platelets were pelleted by centrifugation for 5 min at 800 g, washed twice with Tyrodes-N-2-hydroxyethyl-piperazine-N 0 2ethanesulfonic acid (HEPES) buffer (134 mM NaCl, 0.34 mM Na 2 HPO 4 , 2.9 mM KCl, 12 mM NaHCO 3 , 5 mM HEPES, 5 mM glucose, 0.35% bovine serum albumin (BSA), pH 7.4) containing 2 ml ml À 1 apyrase and 5 ml ml À 1 PGI 2 . Blood samples of patients were obtained after informed consent in accordance with the Declaration of Helsinki. Ethical approval was obtained from Inserm RBM 04-14 for the project 'Network on the inherited diseases of platelet function and platelet production' and by the Ethics Committee of the University Hospital Würzburg. Fresh blood samples of patients and healthy volunteers were collected in 1/10 volume of acid-citrate-dextrose and centrifuged for 10 min at 200 g. PRP was collected, supplemented with 2 ml of apyrase (0.02 U ml À 1 , Sigma-Aldrich) and 5 ml PGI 2 (0.1 mg ml À 1 , Sigma-Aldrich) per ml PRP. Before adherence to poly-L-lysine (PLL)-coated slides or spreading on different matrices, platelets were pelleted by centrifugation for 10 min at 800 g and washed twice with Tyrodes-HEPES buffer containing 2 ml ml À 1 apyrase and 5 ml ml À 1 PGI 2 . The samples were allowed to rest for 30 min at 37°C prior to being used in experiments.
Inductively coupled plasma mass spectrometry (ICP-MS). The cation content of 4 Â 10 7 platelets was analysed by ICP-MS on platelets from PRP. ICP-MS analysis was performed by ALS Scandinavia AB (Lulea, Sweden).
Determination of platelet lifespan. The clearance of platelets from the circulation was determined by the retro-orbital injection of 5 mg DyLight 488-labelled anti-GPIX antibody 2,47 in PBS into male mice. The percentage of labelled platelets was determined by daily blood withdrawal and subsequent analysis by flow cytometry using fluorophore-conjugated platelet-specific antibodies 2 .
MK spreading. In vitro cultivated bone marrow MKs of male mice were allowed to adhere and spread at 37°C and 5% CO 2 on coverslips coated with fibrillar collagen type I (50 mg ml À 1 ; Nycomed), fibrinogen (100 mg ml À 1 ; F4883, Sigma-Aldrich) or CRP (6 mg ml À 1 ) for the indicated time. MK spreading was stopped by fixation and permeabilization of the cells using PHEM buffer supplemented with 4% PFA and 1% IGEPAL CA-630.
Determination of MK ploidy. To determine bone marrow MK ploidy, both femora of male mice were isolated and the bone marrow was flushed and homogenized. Non-specific binding sites of the 5D7 antibody were blocked by incubation of the cell suspension with 0.02 mg ml À 1 anti-FcgR antibody (553142 (2.4G2), BD Pharmingen). Afterwards, MKs were stained using a fluorescein isothiocyanate-conjugated anti-CD41 antibody (10 mg ml À 1 , 5D7). Finally, cells were fixed, permeabilized and DNA was stained using 50 mg ml À 1 propidium iodide (P3561, Invitrogen) staining solution with 100 mg ml À 1 RNaseA (EN0202, Fermentas) in PBS. Analysis was performed by flow cytometry and FlowJo software (Tree Star Inc. Platelet depletion. Circulating platelets were depleted in mice by injection of 20 mg per 30 g body weight anti-GPIba-antibodies (Emfret, Eibelstadt, Germany) and platelet counts were monitored by flow cytometry for 10 days.
The BRIDGE bleeding and platelet disorders collection. We searched for cases with coding variants of extreme low frequency in the TPRM7 gene by analysing data from 702 index cases with BPDs of unknown genetic basis recruited to the BRIDGE study of the NIHR BioResource-Rare Diseases. Gene variants were identified in genome sequencing data by comparison with reference human genome build GRCh37 and consequences were predicted using the Ensembl version 75 TPRM7 transcript ENST00000313478. Variants predicted to alter the amino acid sequence of the protein were retained in the analysis set if they occurred at frequencies of o1 in 10,000 out of 67,000 individuals of the Exome Aggregation Consortium (ExAC) database, o1 in 1,000 in the 10,000 subjects of UK10K database and o1 in 100 in 3,453 subjects with unrelated rare disorders or unaffected pedigree members recruited to other branches of the NIHR BioResource. We selected cases with extreme low-frequency variants and on the basis of clinical and laboratory characteristics, which were coded using Human Phenotype Ontology (HPO) terms 50 retrieved from the BRIDGE-BPD study database. We then focused our further exploration on the cases with extremely rare variants that were unobserved in the nearly 81,000 control DNA samples. The first step was to contact the families to explain the necessity to control these results by performing co-segregation studies. One family met these requirements and additional family members were examined in Paris at a French Centre for inherited platelet diseases.
Co-segregation studies. Genomic DNA from peripheral blood mononuclear cells was extracted using a commercial DNA purification kit (Qiagen) and exon 17 of TRPM7 was amplified by polymerase chain reaction using the respective primer pairs: TRPM7_Ex17_forw: 5 0 -ggagaatgtgctctggattc-3 0 and TRPM7_Ex17_rev: 5 0 -gccaatcatccatcttgctc-3 0 , expected product size 606 bp. PCR products were purified with the help of QIAquick PCR purification kit (Qiagen) and processed for Sanger sequencing.
Platelet function testing. Platelet function testing of human blood was performed by light transmission turbidometric aggregometry. To this end, PRP (2.5 Â 10 5 platelet per ml) was isolated and stimulated with different concentrations of agonists (ADP, 5 and 10 mM; collagen I, 1 and 5 mg ml À 1 ; epinephrine, 5 mM; arachidonic acid, 1 mM; ristocetin, 0.6 and 1.5 mg ml À 1 ; and TRAP (thrombin receptor activating peptide), 20 mM). Light transmission was followed over time with PPP set as 100% light transmission.
Transient expression of WT and p.C721G TRPM7 variants. The TRPM7 p.C721G and p.R902C variants were generated by site-directed mutagenesis on the pcDNA3.1-TRPM7 expression construct 51 using the QuickChange II XL sitedirected mutagenesis kit (200517, Agilent Technologies) with the following primer pairs: Trpm7_C721G_for: 5 0 -ctggagtaattcaaccggcctcaagttagcgtttc-3 0 and Trpm7_C721G_rev: 5 0 -gaaactgctaacttgaggccggttgaattactccag-3 0 . The mutation was confirmed by sequencing. For transient expression of WT TRPM7 and the p.C721G variant, human embryonic kidney (HEK) 293 cells were maintained at 37°C and 5% CO 2 in Earle's minimal essential medium supplemented with 10% foetal bovine serum, 100 mg ml À 1 streptomycin and 100 U ml À 1 penicillin (Invitrogen). Cells were transiently co-transfected with eukaryotic expression vectors encoding WT TRPM7 or the p.C721G and p.R902C variant with 100 ng of an enhanced green fluorescent protein (EGFP) reporter construct (pcDNA3.1) using Lipofectamine2000 (Invitrogen) according to the manufacturer's instructions and processed for patch clamp measurements or confocal microscopy.
Electrophysiology. Patch clamp experiments were performed at a whole-cell configuration. Currents were elicited by a ramp protocol from À 100 to þ 100 mV over 50 ms acquired at 0.5 Hz and a holding potential of 0 mV. Inward currents were extracted at À 80 mV, outward currents at þ 80 mV and plotted versus time.
Image analysis. All images of an experiment were acquired with the same laser power, scan speed, detector settings, processed equally and analysed blinded. Microtubule surface area was determined with the help of Fiji 52 by thresholding the a-tubulin staining of platelets and MKs using the same settings for both groups. NMMIIA distribution in resting platelets was analysed on three-dimensional surface plots and profile plots with the help of Fiji 52 . Means of the first and last maxima and the mean between the first and last minima of the histograms were determined and the NMMIIA max to NMMIIA min was calculated and depicted. NMMIIA distribution in spread MKs was analysed by grouping and counting the different distribution patterns. The surface of spread platelets was measured by thresholding the F-actin staining. To analyse the distribution/content of NMMIIA in spread platelets the area of the NMMIIA staining was correlated to the total cell size (F-actin) staining.
Data analysis. The presented results are mean±s.d. from at least three independent experiments per group, if not otherwise stated. Data distribution was analysed using the Shapiro-Wilk test and differences between control and knockout mice were statistically analysed using Student's t-test or Wilcoxon-Mann-Whitney test, respectively. P-valueso0.05 were considered as statistically significant *Po0.05; **Po0.01; ***Po0.001. Results with a P value40.05 were considered as not significant (NS).