Neutral sphingomyelinase (SMPD3) deficiency disrupts the Golgi secretory pathway and causes growth inhibition

Systemic loss of neutral sphingomyelinase (SMPD3) in mice leads to a novel form of systemic, juvenile hypoplasia (dwarfism). SMPD3 deficiency in mainly two growth regulating cell types contributes to the phenotype, in chondrocytes of skeletal growth zones to skeletal malformation and chondrodysplasia, and in hypothalamic neurosecretory neurons to systemic hypothalamus–pituitary–somatotropic hypoplasia. The unbiased smpd3−/− mouse mutant and derived smpd3−/− primary chondrocytes were instrumental in defining the enigmatic role underlying the systemic and cell autonomous role of SMPD3 in the Golgi compartment. Here we describe the unprecedented role of SMPD3. SMPD3 deficiency disrupts homeostasis of sphingomyelin (SM), ceramide (Cer) and diacylglycerol (DAG) in the Golgi SMPD3-SMS1 (SM-synthase1) cycle. Cer and DAG, two fusogenic intermediates, modify the membrane lipid bilayer for the initiation of vesicle formation and transport. Dysproteostasis, unfolded protein response, endoplasmic reticulum stress and apoptosis perturb the Golgi secretory pathway in the smpd3−/− mouse. Secretion of extracellular matrix proteins is arrested in chondrocytes and causes skeletal malformation and chondrodysplasia. Similarly, retarded secretion of proteo-hormones in hypothalamic neurosecretory neurons leads to hypothalamus induced combined pituitary hormone deficiency. SMPD3 in the regulation of the protein vesicular secretory pathway may become a diagnostic target in the etiology of unknown forms of juvenile growth and developmental inhibition.

The null-allelic smpd1-(Niemann-Pick, type A), 14,15 smpd2-11 and smpd3-13 mouse-mutants have served as unbiased genetic tools in studies on the systemic role of mammalian SMPDs in cellular SM metabolism and its pathophysiology. The current study elaborated the pivotal role of SMPD3 utilizing the smpd3 − / − mouse, which is characterized by the retardation of systemic and skeletal growth and development. In wild-type mice, the smpd3 mRNA is ubiquitously expressed. 2,16, The absence of SMPD3 in hypothalamic secretory neurons inhibited the secretion of proteo-hormones, slowed down the hypothalamus-pituitary-growth axis, and triggered systemic growth retardation resulting in a novel juvenile dwarf phenotype. 13 The autonomous expression of SMPD3 in chondrocytes was shown by functional reconstitution of SMPD3 in smpd3 − / − chondrocytes, expressing smpd3 as transgene, driven by the chondrocyte-specific Col2a1 promoter in the smpd3 − / − mutant. 16 Here, we describe a novel molecular mechanism underlying the bifurcated systemic and cell autonomous SMPD3 deficiency. We first documented the dominant role of SMPD3 and defined the subcellular localization in the Golgi compartment (GC), imperative for unraveling the molecular role of SMPD3 in mammalian cells. Our finding is ad variance with the proposed plasma membrane (PLM) topology of SMPD3. 18,19 We next focused our study on the cell-specific growth regulation of SMPD3 in primary chondrocytes of skeletal growth plates of p16 control and smpd3 − / − mice, corresponding to approximately age 4 years in human development. 20,21 Chondrocytes are competent secretory cells during the growth phase with an abundant secretion of extracellular matrix proteins (ECMs) for enchondral ossification in longitudinal growth. 20 Chondrocytes in culture have proven most powerful in exploring molecular features of growth and development. 22 We discovered the pivotal role of SMPD3 in Golgi vesicular protein transport. Inactivation of smpd3 stalled Golgi protein transport, disrupted proteostasis, induced ER stress and compromised chondrocyte function, leading to apoptosis and ultimately to skeletal malformation and severe chondrodysplasia.
We determined the lipidome of primary chondrocytes of control and smpd3 − / − mice in the growth phase as the structural platform for the functional analysis of SMPD3 in the Golgi secretory pathway (GSP). Our studies suggest a concerted action of SMPD3 and SMS1 in the Golgi complex, which maintains SM/phosphatidylcholine and ceramide (Cer) diacylglycerol (DAG) homeostasis during remodeling of the Golgi membrane lipid bilayer for vesicular transport. This homeostasis is perturbed in SMPD3-deficient GC.
Our studies delineate a novel function of SMPD3 in the lipiddriven formation of vesicle carriers in the GSP during growth and development, and provide insight into the molecular pathology of SMPD3 deficiency leading to an unprecedented mechanism of growth inhibition and to retarded development, manifested juvenile dwarfism and osteochondrodysplasia.

Results
SMPD3 is the key neutral sphingomyelinase, localized in detergent-insoluble membrane domains of the Golgi complex. We first ascertained the subcellular topology, a prerequisite for exploring the mechanistic role of SMPD3. Biochemical and immunohistochemical analyses conclusively proved the absence of SMPD3 in smpd3 − / − chondrocytes (Figures 1a and d). SMPD3 topology was restricted to the Golgi complex in different control cell types: in primary control chondrocytes (Figure 1c), peritoneal macrophages ( Figure 1e) and C57Bl/6 EMFIs (Figure 1f). Colocalization with SMS1 and Golgi-specific marker K58 ascertained the allocation to the GC (Figures 1g and h). SMPD3 and SMS1 resided in the GC of control chondrocytes and SMPD3 overexpressing HEK293 cells (Figures 1i and j).
To further substantiate the Golgi topology of SMPD3 and of SMS1, we used mouse brain, which shows highest SMPD3 expression of all tissues. 2,16 The Golgi fraction of the premyelinating brain of (p14) control and smpd3 − / − mice was further separated into detergent (Triton X-100)-insoluble membrane domains (DIMs) and subnatant for WB (Figures 1k-n). More than two-thirds of total SMPD3 resided in DIMS of control GC (Figure 1k), SMPD3 was absent in smpd3 − / − GC (Figure 1m). Similarly, two-thirds of total SMS1 was concentrated in DIMs of control GC (Figure 1l), but was equally distributed in SMPD3deficient DIMs and subnatant (Figure 1n).
To explore whether SMPD3 is required selectively for the canonical Golgi-mediated or non-canonical secretory pathway, we analyzed the secretion of several cytokines into the medium of control and smpd3 − / − peritoneal macrophages, unstimulated ( Figure 1o) and stimulated ( Figure 1p) with lipopolysaccharide (LPS). The unchanged secretion of cytokines in SMPD3-deficient macrophages clearly indicated SMPD3 function to be restricted to the canonical GSP.
A valuable tool in dissecting Golgi localization, morphology and vesicle formation is exposure of chondrocytes to Brefeldin (BFA). We followed the time-resolved disintegration by IHC of the GC of SMPD3 and K58 (Supplementary Figure S2a), and Col2a and cartilage oligomeric matrix protein (COMP), in control and smpd3 − / − chondrocytes (Supplementary Figure S2b). Golgi membrane stacks disintegrated, vesiculated and fused with ER membranes within 5-30 min.
We then assessed the contribution of bona fide SMPD4 (ref. 3) to cellular nSMase activity in HEK293 cell clones, stably transfected with full-length smpd4-egfp, with threefold to an eightfold overexpression, documented by qRT-PCR (Supplementary Figure S1a). SMPD4-EGFP colocalized largely with K58 in the GC (Supplementary Figure S1c). Surprisingly, the sensitive radioactive nSmase assay revealed an unchanged basal nSMase activity in the postnuclear fraction of all smpd4-egfp overexpressing cell clones (Supplementary Figure S1b). Our biochemical experiments and IHC were unable to substantiate nSMase activity of SMPD4.
The SMPD3-SMS1 cycle regulates SM homeostasis and the Cer-and PLC-independent DAG pool in the GC. Current techniques preclude time and space resolution between the transfers of de novo synthesized Cer from the ER into the GC, and Cer, released locally by SMPD3 hydrolysis from SM of Golgi membrane domains. To get insight into the metabolic interrelationship of SM/Cer and PC/DAG, we focused on their analysis in control and smpd3 − / − primary chondrocytes in culture. We next analyzed the PL-classes of control and smpd3 − / − chondrocytes, which were separated by highperformance thin layer chromatography (HPTLC).
Basically, Cer and DAG concentrations in the lipidome are very low. In smpd3 − / − chondrocytes, the molar ratios of SM and Cer were reduced to one-half and one-third, respectively (Figures 2i-k).
Mass spectrometry, using selective ion monitoring (SIM), was applied for identification and quantification of HPTLCseparated Cer bands (m/z 264 and m/z 266), respectively. In GC, the pattern of fatty acid substituents of DH-Cer of control and smpd3 − / − chondrocytes represents de novo synthesized DH-Cer, which is markedly different from that of Cer. IHC of the other dominant growth regulating cell type, hypothalamic neurosecretory neurons in arcuate N. and periventricular N., using antibodies against growth hormonereleasing hormone (GHRH) and melanocyte-stimulating hormone (MSH), revealed smpd3 − / − neurons heavily loaded with proteo-hormones, but low abundance in control neurons (Figures 3f and g), which confirms previous immunohistochemical results. 13 Inhibited protein transport, dysproteostasis, ER stress and apoptosis in SMPD3-deficient chondrocytes. We next studied the cellular response to the stalled protein transport along the GSP. ER stress is measured by activation of unfolded protein response (UPR) and visualized by the accumulation of misfolded proteins in the lumen of the tubular ER system directly by transmission electron microscopy (EM) (for review Oslowski    Equal amounts of lipid total extracts of enzyme assays were separated by HPTLC and visualized by charring lipid bands (Figure 5l), different from enzyme assay of kidney, which shows equal amounts of de novo synthesized fluorescent SM (Figures 5m and n).

Discussion
This study is focused on the mechanism underlying SMPD3 deficiency, which causes systemic and cell-specific growth inhibition, a novel form of juvenile dwarfism. We used the unbiased smpd3 − / − mouse model. In this study, primary chondrocytes in culture of control and smpd3 − / − miceas compelling in vitro systemwere instrumental in unraveling the molecular mechanism underlying the crucial function of SMPD3 in the GSP. The absence of SMPD3 suppressed ECM protein transport and secretion, disrupted proteostasis, activated UPR and ER stress and finally apoptosis, which is reflected phenotypically in skeletal growth inhibition and joint malformation. Translation of this mechanism into disruption of the GSP in hypothalamic proteo-hormones secreting neurons provides a molecular interpretation of the previously described hypothalamus induced combined pituitary hormone deficiency, underlying the systemic hypoplasia of the smpd3 − / − mutant. 13  Exploring the function of SMPD3 in the non-canonical GSP, we quantified cytokine secretion in the medium of control and smpd3 − / − peritoneal macrophages. The unaltered secretion indicated the restriction of SMPD3 function to the canonical GSP (Figures 1o and p).
Conclusive data on the subcellular topology are a prerequisite for functional studies, as this issue has been controversially discussed and the PLM proposed as the scaffold of SMPD3. 18,19 The immune-histochemical and biochemical studies, reported here, further suggested the localization and association of SMPD3 and SMS1 in the Golgi complex.
SMPD3 and SMS1 are concentrated in DIMs of the GC of control chondrocytes. In the absence of SMPD3, SMS1 segregates from DIM domains, which disturbed structure of DIMs and deregulate SM synthesis in the GC (Figure 1n). SMS1 protein concentration and enzyme activity were increased in lysates of smpd3 − / − chondrocytes (Figures 5k and l). SM and Cer concentrations were reduced (Figure 2i).
Free DAG species resembled that of the DAG core of PC, the donor substrate in the SMS1 reaction. SMPD3 deficiency caused no SM storage and the phospholipidome in all tissues of smpd3 − / − mice remained unimpaired, unlike the fatal lysosomal SM storage in the smpd1 − / − mouse, 14,15 a mimicry of human Niemann-Pick disease (Figures 2a-h). 14,15 Disrupted protein transport in GSP of SMPD3-deficient chondrocytes. Chondrocytes are competent secretory cells, actively synthesizing and secreting ECM proteins for epiphyseal cartilage architecture 28 during the growth phase. The impaired GSP in smpd3 − / − chondrocytes causes cytoplasmic accumulation of dominant ECM proteins: fibrillar Col2a, perifibrillar COMP, Matn3, ColVI and ColIX (Figures  3a-d). Dysproteostasis activates UPR (Figure 4g), unable to alleviate ER stress, but induces premature apoptosis in smpd3 − / − chondrocytes (Figures 4i-u).
EM convincingly unveiled the morphological changes, reflecting these processes: the dense coherent rER tubular system in control chondrocytes (Figures 4a-c) is contrasted by inflated cisternae of the rER network, loaded with polymorphic macro-vesicular, fibrillar and granular structures dispersed in the cytoplasm of smpd3 − / − chondrocytes (Figures 4d-f). This finally leads to skeletal growth retardation, malformation and chondrodysplasia (Figures 4t and u).
Gene and protein expression of representative ECM proteins Col2a and COMP, Golgi-micro vesicular transport proteins Arf and β-Cop1 and of growth factor Igf1 and IgfR were inconspicuous (Figures 5d-i). Expression of growth factors, bmp1, bmp4 and ECM protein comp, respectively, is downregulated (Figure 5c). Secreted BMP1 acts as Ca 2+ and Zn 2+ -dependent metalloproteinase processing procollagen I for ECM organization. SMPD3 is selectively required for the canonical Golgi-mediated secretory pathway. Evidence for unimpaired secretion of non-canonical secretory proteins was provided by multi-cytokine assay, quantifying of 23 cytokines, secreted into the growth-medium of unstimulated and LPS-stimulated peritoneal macrophages of control and smpd3 − / − mice, which revealed closely similar concentrations, presented in Figures 1o and p. SMPD3 transiently modifies the Golgi membrane for vesicular transport. The Golgi complex is a budding organelle with domain formation by lateral diffusion. Coatomer complex COPI and II drive budding, fission and fusion in the formation of small vesicular transport carriers. 29 The function of the SMPD3-SMS1 cycle in the formation of large pleiomorphic carriers for intracellular transport and secretion of ECM proteins in the GSP of chondrocytes, reported here, is a novel facet in studies addressing this enigmatic process.
The central role of SMPD3 and its concerted activity with SMS1 in the SMPD3-SMS1 cycle, which maintains homeostasis of SM metabolism in the GC, is delineated in Figure 6.
Current methodology precludes space-and time-resolved analysis of Golgi membrane domains undergoing remodeling during Golgi vesicular transport. Backed by biochemical, cellular and morphological finding, we conclude that the SMPD3-SMS1 cycle generates a PLC-independent DAG pool in the GC. It is of note that the species of this DAG pool largely reflect those of the major PC species, donor substrates in the  (Figures 1i-k). Sophsticated atomic force microscopy and fluorescence correlation spectroscopy have been applied to phase-separated lipid bilayer model systems, consisting of ternary SM/PC/C domains, embedded in a liquid disordered PC phase, to investigate the effect of Cer on liquid-ordered membrane domains. [33][34] Treatment with bacterial nSMase released Cer from SM of the SM/C complex and simultaneously displaced cholesterol Complementary and in support of our results are the observations that protein trafficking to the cell surface is retarded in response to downregulation of SMS1, 36 and the Golgi secretory function inhibited at reduced DAG levels. 30 SMPD3 induced changes in the homeostasis of SM/PC/ Cer/DAG concentrations in the lipid pool of the GC are expected to be minor and quantification impeded by insufficient current methodology. Kinetic studies of Cer and DAG release, SM hydrolysis and phosphoryl-choline group transfer from PC in control and SMPD3-deficient GC membrane stacks may give insight into the molecular mechanism underlying the potential role of SMPD3 in the Golgi SMPD3/SMS1 cycle in remodeling the Golgi lipid bilayer membrane for vesicle formation, transport and secretion.
Collectively, these studies suggest a novel mechanism underlying the function of SMPD3 in the GSP and necessitate future experiments addressing the concerted action of SMPD3, SMS1, and of Cer and DAG lipid-driven remodeling of the membrane bilayer in the GSP.
This unprecedented aspect of SMPD3 function necessitates a reconsideration of the current tenet regarding the role of Smases and SM metabolites in intracellular signal transduction pathways. Expansion of this concept to other rapidly growing cells, for example, in tumors, the immune system and inflammation, awaits further studies and may open new perspectives for therapeutic strategies.

Materials and Methods
Mouse experiments. All experiments were approved by the Institutional Animal Care and Research Advisory Committee of the University of Cologne, Cologne, Germany.
Real-time PCR. RNA was isolated for RT-PCR from p16 control and smpd3 − / − chondrocytes using Trizol (Invitrogen), reverse-transcribed using a Transcriptase kit (Life Technologies Inc., Darmstadt, Germany). 37 Quantitative PCR reactions were performed with the ABI Prism 7900HT using a 96-well format, the Fast SYBR Green Master Mix (Applied Biosystems, Waltham, MA, USA), following the manufacturer's protocol. Data analysis was performed using the 2-ΔΔCt method.
Acid and neutral sphingomyelinase assays. Total aSMase-and Mg 2+ -dependent nSMase activity were determined as described. 38 SM synthase assay. SMS1 activity in cell lysates was determined following an established procedure. 23 Cell fractionation. Cell fractionation and isolation of Golgi fractions and Triton X-100 insoluble DIMs were performed by established procedures [39][40][41] Cytokine analysis. Cytokines, secreted by peritoneal macrophages via the non-Golgi-secretory pathway, were quantitated in the medium using the BioPlexPro mouse cytokine 23-plex assay (Bio-Rad, Bioplex #M600009RDPD, Hercules, CA, USA) following the manufacturer´s protocol. Macrophages were isolated from adult (4mo) male control and smpd3 − / − mice 72 h after intraperitoneal injection of 1.5 ml 4% thioglycolate, washed with PBS. In all, 1.5 × 10 6 macrophages unstimulated and stimulated with 100 μg LPS/ml medium for 48 h were used. The supernatant was centrifuged for the cytokine assay.
FACS analysis of single-cell chondrocytes. For cell survival experiments, cultured primary chondrocytes were analyzed by flow cytometry as previously described. 45,46 Briefly, chondrocytes were isolated by collagenase type 2 treatment (Worthington Biochemical Corporation, Lakewood, NJ, USA) and subsequently stained with fluorescently labeled annexin A5 (AnxA5-Dye490) 2 and Sytox Blue Dead Cell Stain (Thermo Fisher Scientific, Schwerte, Germany) in binding buffer (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 , pH 7.4) followed by flow cytometry analysis (FACSCantoII, Becton Dickinson, Heidelberg, Germany). Using the FlowJo 7.6 software (LLC Ashland, OR, USA), cell debris were excluded and by comparing the area and width of the FSC and SSC signals only single cells were considered for further analysis. Viable (AnxA5 -/Sytox Blue − ), apoptotic (AnxA5 + /Sytox Blue − ) and dead cells (AnxA5 + /Sytox Blue + ) were quantified and data represented as mean ± S. D. of n = 4 individual mice with three technical replicates each. Statistical analysis was performed with Student´s t-test using the unpaired two tails method.
Electron microscopy. Primary chondrocyte were grown on Aclar-membranes (Fa. TED Pella Inc., Redding, CA, USA) for 6 days, washed 3x with PBS, fixed for 1 h at 4°C with 2% glutaraldehyde, 2% PFA, 0.2% picric acid in 0.1 M cacodylate buffer, pH 7.35. Fixation buffer was removed and cells washed 3x with 0.1 M cacodylate buffer, pH 7.35 post-fixed in 1% OsO 4 solution for 1 h, and stained in 1% uranyl acetate for 1 h at room temperature. After dehydration, specimens were embedded in Araldite (Serva, Heidelberg, Germany). Ultra-thin sections (about 70 nm) were stained with uranyl acetate and lead citrate and were examined by EM (Zeiss 902A, Zeiss, Oberkochem, Germany). The semi-thin sections (1 μm) were stained with methylene blue for light microscopy.
Statistical analyses. Results are expressed as mean ± S.D. Statistical significance of differences between two groups was determined by a two-tailed Student's t-test using GraphPad QuickCalcs (La Jolla, CA, USA): t-test calculator. A P-value of * ⩽ 0.05, ** ⩽ 0.01, *** ⩽ 0.001 was considered significant. Sizes of animal cohorts are listed under respective figures.