The chloride antiporter CLCN7 is a modifier of lysosome dysfunction in FIG4 and VAC14 mutants

The phosphatase FIG4 and the scaffold protein VAC14 function in the biosynthesis of PI(3,5)P2, a signaling lipid that inhibits the lysosomal chloride transporter ClC-7. Loss-of-function mutations of FIG4 and VAC14 reduce PI(3,5)P2 and result in lysosomal disorders characterized by accumulation of enlarged lysosomes and neurodegeneration. Similarly, a gain of function mutation of CLCN7 encoding ClC-7 also results in enlarged lysosomes. We therefore tested the ability of reduced CLCN7 expression to compensate for loss of FIG4 or VAC14. Knock-out of CLCN7 corrected lysosomal swelling and partially corrected lysosomal hyperacidification in FIG4 null cell cultures. Knockout of the related transporter CLCN6 (ClC-6) in FIG4 null cells did not affect the lysosome phenotype. In the Fig4 null mouse, reduction of ClC-7 by expression of the dominant negative CLCN7 variant p.Gly215Arg improved growth and neurological function and increased lifespan by 20%. These observations demonstrate a role for the CLCN7 chloride transporter in pathogenesis of FIG4 and VAC14 disorders. Reduction of CLCN7 provides a new target for treatment of FIG4 and VAC14 deficiencies that lack specific therapies, such as Charcot-Marie-Tooth Type 4J and Yunis-Varón syndrome.


Introduction
The signaling lipid PI(3,5)P 2 is generated at the surface of endosomes and lysosomes by a protein complex that includes the kinase PIKFYVE, which synthesizes the lipid, the phosphatase FIG4, which degrades it, and the scaffold protein VAC14 [1][2][3]. Cells deficient in FIG4 or VAC14 have enlarged lysosomes bounded by membranes containing LAMP1 and LAMP2, and elevated levels of the autophagosome markers LC3-II and p62 [1,4,5]. Surprisingly, since FIG4 is a phosphatase that degrades phosphoinositides, loss of FIG4 leads to decreased levels of PI(3,5)P 2 in cells [5]. Mutant mice and patients with genetic disorders affecting FIG4 and VAC14 develop vacuolization of the CNS [5][6][7][8]. The abnormal lysosomes in FIG4 deficient cells are phase-lucent and distended and appear to be fluid-filled, raising the possibility that they contain elevated solute levels leading to osmotic swelling. If this is the case, elevated hydrostatic pressure and lysosomal membrane tension within the lysosome could impede the recruitment of membrane remodeling components and prevent regeneration of normal lysosomes via tubulation and membrane recycling [9][10][11].
Mutations of FIG4 and VAC14 have been identified in patients with neurodegenerative disorders. The lethal multisystem disorder Yunis-Varón Syndrome results from complete loss-offunction of FIG4 or VAC14 [7,12]. Mutations causing partial loss-of-function of FIG4 are associated with Charcot-Marie-Tooth disease type 4J and familial epilepsy with polymicrogyria [5,13]. Partial loss-of-function mutations of VAC14 are seen in pediatric-onset striatonigral degeneration and neurodegeneration with impaired CNS myelination [6]. Mice carrying a loss-of-function mutation of Fig4 display neurological phenotypes and reduced pigmentation [5]. Similar phenotypes, including hypopigmentation and neurological disorders, result from the gain-of-function mutation p.Tyr715Cys in the chloride transporter gene CLCN7 [14]. The analogous syndromes resulting from disruption of FIG4 and CLCN7 raises the possibility of a functional relationship between the two lysosomal proteins they encode. In contrast to the gain-of-function variant p.Tyr715Cys, the impaired trafficking of CLCN7 in the dominant negative variant p.Gly215Arg (mouse p.Gly213Arg) results excess bone deposition and osteopetrosis [15]. PI(3,5)P 2 directly regulates several lysosomal ion channels and transporters [16][17][18][19][20]. Recently, PI(3,5)P 2 was demonstrated to influence lysosomal pH via inhibition of the transporter ClC-7 [21]. The CLCN gene family encodes nine CLC transport proteins that function either as Cl − channels or as electrogenic Cl -/H + transporters [22][23][24]. ClC-6 and ClC-7, the proteins encoded by CLCN6 and CLCN7, share 45% amino acid sequence identity and constitute a distinct branch of the CLCN family [25]. The ClC-6 transporter is mainly located in late endosomes, while ClC-7 is located in both lysosomes and late endosomes [24,26,27]. ClC-6 and ClC-7 are Cl -/H + exchangers with exchange stoichiometry of 2 Clfor 1 H + that actively pump Clinto the endosome and lysosome in exchange for efflux of protons [23,24,28]. ClC-7 has been proposed to be a component of the lysosomal 'counterion pathway' which is required for acidification of the organelle, but its role has not yet been fully defined and it may serve to regulate lysosomal [Cl -] for other purposes [24,29,30].
The cryo-EM structure of human ClC-7 reveals a positively charged phosphoinositide binding pocket that is conserved in ClC-6 and in the plant anion/proton exchanger atCLC-a [31]. PI(3,5)P 2 tonically inhibits atCLC-a and human ClC-7 [21,32]. Deficiency of FIG4 or VAC14 leads to reduction of PI(3,5)P 2 [5] that could in turn result in reduced inhibition of ClC-7 and accumulation of Cl -. One possible mechanism for the lysosomal swelling could be increased osmotic pressure due to the accumulated Cl -, which would lead to osmotic flux into the organelle. This model would connect the dysregulation of H + /Clantiport to the vacuole formation observed in FIG4 and VAC14 deficiency or ClC-7 gain-of-function.
To probe the connections between alterations in the PI(3,5)P 2 pathway and lysosomal chloride/proton exchange, we examined gene interaction between FIG4, VAC14, CLCN6 and CLCN7 by pairwise combination of mutant genes in cultured cells and mutant mice. The results support a role for CLCN7 in the lysosome dysfunction of FIG4 and VAC14 mutants, represented in the final figure, and suggest that inhibitors of this chloride transporter may be therapeutic for FIG4 and VAC14 deficiency disorders.

Ethics statement
All experiments described involving animals were in accordance with institutional animal care and use guidelines, and experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Michigan under animal protocol number PRO00009797.

Cell culture and imaging
Human HAP1 cells (Horizon Discovery, #C631) were maintained in Iscove's Modified Dulbecco's Medium (IMDM) supplemented with 10% FBS, 100 units/mL penicillin, and 100 μg/ mL streptomycin (Invitrogen). Cells were cultured in a humidified incubator at 37˚C with 5% CO2. Transfection of plasmids into HAP1 cells was carried out with Lipofectamine 3000 (Life Technologies) following manufacturer's instructions. Cell morphology was evaluated with phase contrast images taken at 20x with the EVOS FLc system (Life Technologies). Visible vacuoles (>0.4 um) were measured using ImageJ software [33]. Ratiometric measurement of lysosomal pH was carried with the dye Oregon Green 488-dextran as previously described [21]. LAMP2 antibody # H4B4 was obtained from the Developmental Studies Hybridoma Bank, University of Iowa, and used at 3 ug/ml. The secondary antibody was anti-Mouse IgG Alexa fluor plus 488 from Invitrogen, A32723, used at 1/2000 dilution.

Generation of CRISPR edited HAP1 cell lines
Guide sequences targeting exon 2 of the FIG4 gene were designed with E-CRISP software [34] and obtained from IDT (Integrated DNA Technologies). The guide oligos were cloned into plasmid pSpCas9 (BB)-2A-Puro (PX459) V2.0 (Addgene # 62988) for expression of the sgRNA and Streptomyces pyrogenes Cas9. After 48 hours of selection for puromycin resistance, individual HAP1 cells were isolated by flow sorting into 96 well plates. FIG4 null clones were visually identified by their content of large vacuoles. Exon 2 of FIG4 was amplified from vacuolated clones and indels were identified by Sanger sequencing. The VAC14 null HAP1 cell line G9 was generated previously using the same method [35]. Guide RNAs targeting CLCN6 and CLCN7 were selected from the GeCKO library [36]. sgRNAs were obtained from IDT and cloned into pSpCas9 (BB)-2A-Puro (PX459) V2.0 as above. After transfection and puromycin selection, single cells were isolated by flow-sorting, plated in 96 well plates, and grown to confluence. Targeted exons were isolated with the TOPO TA cloning kit from Invitrogen (Cat. # 45-0071) and examined by Sanger sequencing to identify indels and in-frame termination codons. Wildtype transcripts were not detected in the selected null clones for FIG4, VAC14, CLCN6 and CLCN7.

FACS analysis of lysosome enlargement
Fluorescence-activated cell sorting (FACS) was carried out as previously described [35]. HAP1 cells were cultured in 6-well plates for 18 hours and labeled for 15 min with 5 μM LysoSensor Yellow/Blue DND-160 dye (PDMPO) (Invitrogen #L7545). After washing three times with PBS, cells were removed from plates by treatment with TrypLE Express Enzyme (Thermo Fisher Cat #12604013), suspended in PBS containing 2% FBS, and transferred to 5 mL flow vials on ice. The fluorescence of LysoSensor DND-160 was recorded in the yellow spectrum to detect variation in lysosome content, with excitation at 329 nm and emission at 546 nm. Cells were sorted in a MoFlo Astrios Cell Sorter (Beckman Coulter, Inc) in the University of Michigan Flow Cytometry Core.

Animals
The Fig4 null mouse mutant, also designated pale tremor (plt), arose spontaneously by transposon insertion on a mixed strain background. Two congenic mouse Fig4 null lines were subsequently generated, by >30 generations of backcrossing to strain C57BL/6J and strain C3H/ HeJ [37]. Clcn7-G213R mice were generated by targeting 129Sv ES cells [15] and kindly provided by Dr. Michael J Econs, Indiana University School of Medicine. The current Clcn7-G213R line was confirmed to be congenic on strain C57BL/6J by SNP genotyping with the miniMUGA panel (Neogen). To measure hindlimb clasping, mice were suspended for 1 min and the % time with hindlimbs retracted was measured [38]. For the hanging wire test, mice were placed on the wire mesh cage lid, the lid was inverted, and the time to falling was measured. The human CLCN7 variant p.Gly215Arg corresponds to p.Gly213Arg in the mouse gene. Histology on H&E stained sections of brain and dorsal root ganglia was carried out at Histoserv (Bethesda, MD). The percent of the deep cerebellar nucleus occupied by vacuoles was determined using ImageJ and Adobe Photoshop. Mice in this study were housed and cared for in accordance with NIH guidelines in a 12/12h light/dark cycle with standard mouse chow and water available ad libitum.

Generation of a FIG4 null HAP1 cell line
The FIG4 null cell line was generated by transfection of HAP1 cells with Cas9 and an sgRNA complementary to exon 2 (Fig 1). Individual mutated cells were sorted into 96 well plates and subjected to clonal expansion. During culture, the haploid HAP1 cells reverted to the more stable diploid state [39]. Sequencing of vacuolated clones identified the FIG4 null cell line (FIG4-F). The FIG4 gene contains a 283 bp insertion downstream of c.78A that results in the protein truncation allele p.Leu26 * 8 (Figs 1A and S1). The FIG4 null cells are filled with large spherical vacuoles (Fig 1B), bounded by a membrane that is positive for the lysosomal membrane marker LAMP2 (Fig 1C), consistent with the previously described FIG4 null phenotype [4,5]. Transfection with wildtype FIG4 cDNA corrected vacuolization ( Fig 1D). To confirm the lack of wildtype FIG4 expression, we amplified exon 2 by RTPCR. Two products were obtained, with lengths of 200 and 500 bp ( Fig 1E). TA cloning and Sanger sequencing demonstrated that the 500 bp RTPCR product contains the entire 283 bp insert with an in-frame stop codon encoding the protein Leu26*8 (S1 Fig). The 200 bp RTPCR product is spliced from exon 1 to a splice acceptor site near the 3' end of the 283 bp insert, and encode the protein product Arg21*1 (S1 Fig). There was no expression of wildtype FIG4 transcript in the HAP1 null line.
The expression of CLCN6 and CLCN7 in wildtype HAP1 cells and FIG4-F null cells was also demonstrated by RT-PCR ( Fig 1F).

Correction of enlarged lysosomes in FIG4 null cells by knockout of CLCN7
To evaluate the role of CLCN7 in FIG4-related lysosome dysfunction, we generated double knockout cells. The FIG4 null cells were transfected with a plasmid expressing Cas9 and three different sgRNAs targeting CLCN7. The sgRNA sequences are shown in Fig 2A (highlighted in grey) and S2 Fig. Three subclones designated FC7-1, FC7-2 and FC7-3 were isolated, one from each sgRNA. The FIG4 null cells reverted to the stable diploid state [39] and Sanger sequencing demonstrated that each CLCN7 null cell line contains two allelic indels (Figs 2A  and S2). These six indels in CLCN7 all generate null alleles with in-frame stop codons leading to premature protein truncation.  In contrast to the parental FIG4 null line, all three subclones that are also null for CLCN7 exhibit normal morphology and lack enlarged lysosomes (Figs 2B and S2). The effect of inactivation of CLCN7 on correcting lysosome morphology was quantitated by FACs sorting.
Line FC7-1 was incubated with fluorescent Lysosensor DND160 and the distribution of cell fluorescence was measured. The proportion of cells with elevated lysosomal fluorescence in the Fig4 null cells was 94 ± 4% (mean ± SD, n = 3) and was dramatically reduced in the double mutant lines (e.g FC7-1, Fig 2C, p< 0.0001, two tailed t-test). The percent of cells without vacuoles was restored to wildtype level in all 3 double mutant lines (Fig 2E). The diameter of the enlarged vacuoles in FIG4 null cells, which varied up to 30 uM, was restored to normal in the double mutant cells (Fig 2F). The striking correction of vacuolization by knockout of CLCN7 suggests a key role for ClC-7 in the enlarged lysosomes of FIG4 null cells.

Lysosomal pH in FIG4 -/-,CLCN7 -/double knockout cells
We previously reported that the pH of the enlarged lysosomes in FIG4 null HAP1 cells is significantly more acidic than in wildtype cells [35]. Knockout of CLCN7 in wildtype cells does not affect lysosomal pH [21,40,41]. To determine the impact of knockout of CLCN7 in the FIG4 null cells, we loaded lysosomes with Oregon Green 488-dextran and quantified lysosomal pH by ratiometric analysis. Calibration curves are shown in S3 Fig. In contrast to cells from patients with gain of function variants of CLCN7 [14], in the FIG4 null cells the extensive overlap between compartments staining with Oregon Green 488-dextran (OG488) and Lysotracker Blue indicates that the dextran-coupled dye is efficiently delivered to the large vacuoles (S4 Fig). The average lysosomal pH of the FIG4 null line was pH 4.10, compared with pH 4.65 in wildtype cells (Fig 3). Knockout of CLCN7 in the FIG4 null cells resulted in an increase of 0.15 pH units (Fig 3), equivalent to a~29% decrease in proton concentration in the lysosomal lumen. This partial correction in the double mutant line indicates that CLCN7 contributes to hyperacidity in FIG4 null cells, in combination with additional mechanisms that may play a role.

Vacuolization of VAC14 null cells is also rescued by knockout of CLCN7
VAC14 is a scaffold protein with repeated HEAT domains that is co-localized with FIG4 in the PI(3,5)P 2 biosynthetic complex. Like FIG4, loss of VAC14 results in reduced PI(3,5)P 2 and enlarged lysosomes [1]. To determine whether inactivation of CLCN7 is protective against loss of VAC14, the HAP1 null cell line G9 [35] was transfected with Cas9 and two different sgRNAs targeting exon 11. Two clonal lines containing biallelic null mutations in exon 11 of CLCN7 were isolated and designated VC7-1 and VC7-2 ( Fig 4A). The enlarged lysosomes of the VAC14 null cells were corrected in both double knockout lines (Fig 4B). The 43% of cells with enlarged lysosomes in the VAC14 null line was reduced to 9% and 10% in double knockout lines VC7-1 and VC7-2 (Fig 4C and 4D). LAMP2 staining of the vacuoles in VAC14 null cells is shown in Fig 4E. The number of vacuoles per cell and the size of vacuoles were also returned to normal by inactivation of CLCN7 (Fig 4F and 4G). This result is consistent with the proposed role of CLCN7 in the response to the low PI(3,5)P 2 levels in VAC14 mutant cells [1].

Inactivation of CLCN6 does not rescue FIG4 null cells
The amino acid sequences of the ClC-6 and ClC-7 transporters are 55% divergent, and the two proteins were reported to differ in subcellular localization, with CLCN6 restricted to the endosome [26]. To determine whether knockout of CLCN6 also alters the enlarged vacuole phenotype of FIG4 null cells, we transfected the FIG4 null cell line with Cas9 and sgRNAs targeting exon 2 and exon 4 of CLCN6. Two double mutant cell lines with biallelic frame-shifting indels in CLCN6 were isolated (Fig 5A). Both double mutant cell lines were as highly vacuolated as the FIG4 null parent line (Fig 5B and 5C). The size of the vacuoles in these double mutant lines (Fig 5D) did not differ from the FIG4 null lines (Fig 2F). The failure of CLCN6 disruption to correct vacuolation of FIG4 null cells demonstrates divergence in the function of these two members of the CLCN gene family.

PLOS GENETICS
Genetic interaction between FIG4 and CLCN7 modifies lysosome function There was also improved performance on two tests of neuromuscular function. On postnatal day 30, Fig4 null mice could support their weight in the hanging wire test for only 18 ± 2 seconds before falling, while Fig4 -/-,Clcn7 G213R/+ mice remained for 55 ± 8 seconds (Fig 7B). Abnormal hindlimb and forelimb clasping persisted for > 50% of test time in Fig4 null mice (Fig 7C). In Fig4 -/-,Clcn7 G213R/+ mice, limb clasping was less extreme and persisted for only 17% of test time (Fig 7C).
Histological examination of tissues from the Fig4 -/-,Clcn7 G213R/+ double mutant mice revealed partial rescue of tissues affected in the Fig4 null plt mice. The earliest vacuolization in the plt mice is seen in the deep cerebellar nucleus at E16 [5]. In the double mutant mice, the deep cerebellar nucleus at 30 days of age is partially corrected, compared with Fig4 null mice (Fig 8A and 8B). Dorsal root ganglia are extensively vacuolated in the plt mutant, and partially protected in the double mutant mice (Fig 8A and 8B). Spleen is also highly vacuolated in Fig4 null mice, and partially restored in the double mutant mice (Fig 8A and 8B).
Consistent with the improved behavior and pathology, maximal survival of mutant mice was increased from 6 weeks for Fig4 null mice to 8 weeks in Fig4 -/-,Clcn7 G213R/+ mice (p<0.01, Log Rank Test) (Fig 9).
In summary, the improved lysosome morphology brought about by knockout of CLCN7 in cultured FIG4 null HAP1cells was confirmed in vivo in the double knockout mice by improved histopathology and extended survival. The data suggest that reduction of CLCN7 activity could positively impact human genetic deficiencies human FIG4 and VAC14.

Discussion
PI(3,5)P 2 is emerging as an important regulator of lysosomal physiology and ion homeostasis. Previous work demonstrated that depletion of PI(3,5)P 2 by PIKFYVE inhibition or by FIG4 mutation results in altered lysosomal membrane dynamics. PI(3,5)P 2 modulates a number of known lysosomal and endosomal ion channels and transporters, including TRPML1 [43] and TPC1 [16,17,19,37,44,45], though the physiological significance of this modulation is unclear. More recently, PI(3,5)P 2 was shown to inhibit ClC-7 function, thereby limiting the normal extent of acidification by the v-type ATPase [21]. Coordinated regulation of multiple channels and transporters may be important for simultaneously controlling lysosomal pH, ion composition, and osmotic balance.
Inactivation of FIG4 or VAC14 in HAP1 cells results in accumulation of enlarged hyperacidic vacuoles derived from lysosomes [35]. Because of the similarity between these and the vacuoles formed upon ClC-7 gain-of function, we examined the role of CLCN7 in this process by generating cell lines with mutations of the phosphoinositide biosynthetic genes FIG4 or VAC14 together with mutations of the lysosomal chloride transporter genes. Remarkably, knockout of CLCN7 reduced vacuole formation in FIG4 null and VAC14 null cells. Knockout of CLCN7 also increased vacuolar pH in Fig4 null cells (though not to normal levels), supporting the role of ClC-7 in vacuolar acidification of PI(3,5)P 2 deficient cells [21,35]. These observations suggest that the chloride ions carried by ClC-7 and/or the resulting pH changes are essential effectors of the lysosomal swelling in PI(3,5)P 2 deficiency. Interestingly, the lysosomal swelling is corrected by the lysosomotrophic agent chloroquine, which also alkalinizes lysosomes in cultured cells [14,46]. Gain-of-function mutations of CLCN6 and CLCN7 induce enlarged lysosomes in mammalian cells similar to those of PI(3,5)P 2 deficient cells, raising the possibility that accumulation of Cl − can induce lysosome enlargement [14,47]. While knockout of CLCN7 corrected the enlarged lysosomes in FIG4 null cells, knockout of the paralog CLCN6 did not, demonstrating a difference between the two evolutionarily related transporters. This may result from a difference in subcellular, recently described functional differences [48], or lack of CLCN6 protein expression in HAP1 cells. Although expression of the CLCN6 transcript was detected in HAP1 cells, and only mutant transcript was found in mutant CLCN6 mutant cells, we were unable to confirm translation of this transcript in wildtype HAP1 cells. Since expression of CLCN6 protein may be limited to neurons [26], the difference between CLCN7 and CLCN6 in our experiments may reflect a difference in expression of the protein rather than a difference in their role  in response to phosphoinositides. Other evidence suggests that lysosomal enlargement and hyperacidification due to PI(3,5)P 2 depletion might occur through divergent mechanisms, suggesting a more complex mechanism of lysosomal size change [21].
The dominant negative mutant CLCN7-G215R causes defective trafficking of the ClC-7 transporter [42]. Heterozyzous mice carrying the corresponding G213R mutation exhibit mild osteopetrosis [15]. Combining G213R with the FIG4 null mutation resulted in significant improvements in the mutant phenotype, with increased growth and muscle strength, and a 20% extension in lifespan from 38 days to 48 days. In recent work on the human CLCN7 gainof-function mutation Y715C, the enlarged vacuoles were corrected by overexpression of oculocutaneous albinism II (OCA2) Cl-ion channel, permitting exit of Cland possibly of osmotically coupled water from the lysosome [49]. These observations support a connection between chloride overload in enlarged lysosomes and vacuole formation (Fig 10).
Mutations of FIG4 and CLCN7 both impair bone remodeling. In the FIG4 null disorder Yunis-Varón Syndrome, osteoblast cultures accumulate large vacuoles and patients display multiple skeletal defects including reduced bone density, digital abnormalities and cleidocranial dysplasia [7], suggesting a defect in bone deposition. In contrast to bone loss in the FIG4 null mutant, there is excessive bone matrix in patients with dominant negative mutations and reduced expression of CLCN7 [27]. The G215R allele of CLCN7, used here to reduce CLCN7 function in the mouse, was originally identified in a patient with dominant osteopetrosis [50] which results from impaired osteoclast-mediated bone resorption and is thus physiologically quite different from the deposition defect in FIG4 mutants. Hypopigmentation is another lysosome-related defect shared by loss-of-function of FIG4 and gain-of-function of CLCN7, whose precise mechanism remains unclear [51].
Inhibitors of CLCN7 have been investigated for treatment of osteoporosis [52,53]. These include micromolar levels of several non-specific inhibitors [42], blocking antibodies [54], and genetic down-regulation by antisense oligonucleotides and siRNA [55]. These strategies for reducing CLCN7 function are potential therapies for disorders caused by loss of function of FIG4 and VAC14.
Our data implicate CLCN7 in the osmotic swelling and hyperacidification of lysosomes in PI(3,5)P 2 deficient cells caused by loss-of-function of FIG4 or VAC14. Direct interaction between VAC14 and ClC-7 proteins was reported in the BioPlex Interactome 3.0 database [56], suggesting that the PI(3,5)P 2 biosynthesis complex may be associated with ClC-7 in the lysosome membrane. Conversion of PI(3)P to PI(3,5)P 2 may be a rate-limiting step in the reformation of lysosomes from autolysosomes [4,[57][58][59][60]. The work described here identifies CLCN7 as a potential target for treating neurodegenerative diseases caused by deficiency of PI (3,5)P 2 . HAP1 cells with WT genotype (black), FIG4 null (pink), and double mutant FC7-1 (teal), were subjected to a ratiometric assay after setting the pH using pH-controlled bathing buffers and a combination of ionophores [21]. Each symbol represents the average lysosomal 490/435 fluorescence ratio from 14-15 cells, comprising 4-5 cells each from three independent experiments. The data are presented as mean ± standard deviation (SD).