SORLA is required for insulin-induced expansion of the adipocyte precursor pool in visceral fat

Aberrant expansion of white adipose tissue (WAT) mass causes obesity, a pathological condition that is prevalent in populations around the globe (NCD Risk Factor Collaboration [NCD-RisC], 2016; Global BMI Mortality Collaboration et al., 2016). Despite a high degree of functional resemblance, the main WAT depots underneath the skin (subcutaneous [sc]) and in the abdominal cavity (visceral [vis]) are quite distinct in their biological properties and their resulting cardiometabolic risk profiles (Pinnick et al., 2014). While adipogenesis in scWAT is largely confined to the prenatal period, adipogenesis in visWAT proceeds postnatally and contributes to the hyperplastic expansion of this fat depot in a state of energy surplus (Holtrup et al., 2017; Kim et al., 2014; Wang et al., 2013). The adipogenic potential of visWAT is particularly prevalent in the juvenile organism when the size of the adipocyte precursor cell pool in this tissue is determined by nutritional status (Wang et al., 2013). Remarkably, the precursor pool size defined at juvenile age remains constant into adult life (Spalding et al., 2008). Furthermore, recent work demonstrated the importance of these adipocyte progenitor cells in metabolic control (Shao et al., 2018). Still, the signals that promote precursor pool expansion in juveniles and why they operate in visWAT but not in scWAT are important yet unresolved questions.

Sorting-related receptor with A-type repeats(SORL1) is a gene associated with the longitudinal risk of obesity (Smith et al., 2010), i.e., the risk for obesity in adults who were overweight during early childhood (Geserick et al., 2018) or adolescence (reviewed in Singh et al., 2008). SORL1 encodes the type-1 transmembrane protein SORLA (also known as LR11) that acts as an intracellular sorting receptor for the insulin receptor (IR). Specifically, SORLA recycles internalized IR molecules from endosomal compartments back to the cell surface to increase the active surface pool of the IR and to enhance insulin signal reception in target cells (Schmidt et al., 2016). Here, we tested the hypothesis that SORLA activity represents a mechanism linking nutrition through insulin signaling with visceral progenitor cell expansion.

In line with our hypothesis, we document high levels of SORLA expression specific to adipocyte precursors in juvenile visWAT. In these cells, SORLA facilitates signaling by insulin to promote mitotic expansion, a determining step in adipogenesis. In mice lacking SORLA, blunted insulin signaling in response to overfeeding reduces the proliferative capacity of adipocyte precursor cells in juveniles and decreases the progenitor pool size in adults. Contrary to visWAT, this insulin-sensitizing activity of SORLA is absent in precursor cells from scWAT, explaining the distinct proliferative profiles of these two fat depots in response to dietary stimuli.

Previous studies documented that levels of SORL1 transcript and protein in visWAT are positively correlated with BMI in humans (Schmidt et al., 2016). In mice, overexpression of SORLA in WAT increased adipose tissue mass, while loss of receptor expression protected from diet-induced obesity (Schmidt et al., 2016). While these findings established the relevance of SORLA activity in WAT for body weight control, the mechanism whereby this receptor may affect the longitudinal risk of obesity remained unexplained.

Comparing SORL1 transcript levels in visWAT and scWAT biopsies in a cohort of 362 adult human subjects, we now show that correlation with BMI is specific for SORL1 transcript levels in visWAT of males (Fig. 1 A) and females (Fig. 1 B). Surprisingly, this correlation is not seen in scWAT of both sexes (Fig. 1, C and D).

In human WAT, SORL1 transcripts are found in mature adipocytes with slightly higher levels in vis as compared with sc adipose tissue (Fig. 1 E). Mature adipocytes constitute ∼20–40% of the total cell numbers in human WAT (Eto et al., 2009; Miao et al., 2020; Sun et al., 2020). Based on deposited single-cell RNA sequencing (scRNAseq) data (Vijay et al., 2020), the remaining stromal vascular fraction (SVF) of human WAT is composed of adipocyte precursor cells and various immune and endothelial cell types, with adipocyte precursors being more abundant in visWAT as compared with scWAT (76% versus 35% of SVF cells; Fig. 1 G). In the SVF, SORL1 transcripts are found in precursors and immune cell types at comparable levels (Fig. 1 F). However, whereas adipocyte precursors constitute the main SVF cell type in visWAT to express SORL1 (77% of cells), few precursors in scWAT express the receptor (3% of cells; Fig. 1 H). This striking observation suggests SORL1 expression as a feature discriminating adipocyte precursors in human visWAT from those in scWAT.

A distinctive characteristic of juvenile visWAT is the ability to expand its inherent pool of precursor cells, facilitating hyperplastic growth of this tissue mass in adult life (Wang et al., 2013). To query the contribution of SORLA to precursor cell proliferation, we compared the proliferative capacity in sc inguinal and epididymal (ep) fat pats of mice, either WT or genetically deficient (knockout [KO]) for Sorl1 (Andersen et al., 2005; Schmidt et al., 2016). Although the precise anatomical localization of these fat pats differs from that in humans, these murine tissues are commonly accepted model systems for human sc and vis fat (Cinti, 2012; Jeffery et al., 2016).

Initially, we studied mice on a normal chow diet at 5 wk (juvenile) or 10 wk (adult) of age. To determine the number of proliferative cells in WAT depots, we quantified EdU⁺ (5-ethynyl-2'-deoxyuridine) cells in these mice following supplementation with EdU in drinking water for 7 d (Jeffery et al., 2015). In WT mice, proliferative (EdU⁺) cells were detected in epWAT and scWAT in the adult (Fig. 2, A and C) and in the juvenile stage (Fig. 2, B and C). However, the proliferative capacity was highest in juvenile epWAT, exceeding that of other juvenile or adult fat tissues by fourfold (Fig. 2 C). SORLA deficiency decreased the number of EdU⁺ cells in epWAT and scWAT of KO mice, documenting the relevance of receptor activity for cell proliferation in these tissues (Fig. 2 C). To specifically investigate the impact of SORLA on nutrition-induced cell proliferation, we fed juvenile mice a high-fat diet (HFD) for 3 d and quantified the resulting numbers of proliferating cells in WAT thereafter using EdU incorporation. In juvenile WT animals, short-term HFD feeding increased the number of EdU⁺ cells in epWAT by fivefold as compared with normal chow feeding (Fig. 2 D). This substantial increase in proliferating cells was not seen in juvenile scWAT, or in epWAT and scWAT from adult WT mice (Fig. 2 D). This diet-induced increase in cell proliferation in epWAT was significantly reduced in juvenile mice lacking SORLA (Fig. 2 E). By contrast, no effect of SORLA deficiency on HFD-induced cell proliferation was seen in scWAT (Fig. 2 F). Thus, while SORLA deficiency impairs basal levels of cell proliferation in scWAT and epWAT to a similar extent, its impact on nutrient-induced cell proliferation is unique to juvenile epWAT, and not seen in scWAT.

Quantification of EdU⁺ cells by immunohistology (as in Fig. 2) encompassed adipocyte precursors but also other proliferating cell types in the murine WAT potentially impacted by receptor deficiency. Thus, we refined further analysis by focusing on adipocyte precursors isolated from these tissues. To do so, we applied a multicolor FACS scheme to isolated a (Lin⁻CD29⁺CD34⁺Sca1⁺) cell population that constitutes the early adipocyte precursor cell type in WAT. This precursor cell population can be further distinguished into adipocyte progenitors (CD24⁺) and committed preadipocytes (CD24⁻; Berry and Rodeheffer, 2013; Rodeheffer et al., 2008). Both cell populations express PdgfRα, a marker of the adipogenic lineage (Berry and Rodeheffer, 2013). Here, we applied multicolor FACS to characterize these two precursor cell populations in epWAT and scWAT of juvenile WT and KO mice (Fig. S1, A–D).

Sca1⁺ adipocyte precursors express SORLA as documented by quantitative RT-PCR and Western blotting (Fig. 3 A). Sorl1 transcript levels were highest in precursors from juvenile epWAT as compared with adult epWAT (3-fold) or juvenile or adult scWAT (7- or 15-fold; Fig. 3 B). The proliferative capacity of CD24⁺ progenitors and CD24⁻ preadipocytes in juvenile epWAT was significantly reduced by SORLA deficiency as shown by treating WT and KO mice with BrdU and by subsequent quantification of BrdU⁺ cells in the sorted Sca1⁺ cell populations (Fig. S1 E and Fig. 3 C). No impact of SORLA deficiency on the proliferative capacity of progenitors or preadipocytes was seen in scWAT (Fig. 3 D). Impaired precursor cell proliferation coincided with a significant reduction in the number of CD24⁺ progenitors in epWAT (Fig. 3 E) but not in scWAT (Fig. 3 F) of juvenile KO mice. This fat depot–specific impact of SORLA deficiency on the proliferative capacity of Sca1⁺ precursors in epWAT, but not scWAT, was also seen in juvenile mice fed an HFD for 1 wk (Fig. 3 G). Apart from an apparent decrease in the number of adipocyte precursors in mutant epWAT, SORLA deficiency did not cause any discernable alteration in the numbers of various immune and endothelial cell types in the sorted SVF of epWAT or scWAT (Fig. S2).

In mature adipocytes, SORLA acts as a transmembrane receptor to promote functional expression of the IR (Schmidt et al., 2016) . However, SORLA in adipocytes is also subject to ectodomain shedding, releasing the extracellular portion of the receptor that acts as a paracrine factor in control of thermogenic gene expression (Whittle et al., 2015). To query whether SORLA promotes nutrient-induced proliferation of adipocyte precursors mainly through a cell-autonomous or a non–cell-autonomous mechanism, we performed cell transplantation experiments in mice (Fig. 4 A). In detail, we purified the Sca1⁺ precursor cell population from SVF of the epWAT of juvenile mice and transplanted them into the scWAT pad of recipient animals using Matrigel plugs (Schwalie et al., 2018). Subsequently, we treated the recipients with an HFD for 5 wk and with EdU in drinking water for the first 2 wk of the protocol. Finally, we quantified the number of EdU⁺ cells in the Matrigel plugs recovered from the recipients using immunohistology (Fig. 4 B). Dissection was meant to assure that mainly transplanted cells were quantified, although infiltration of some host cells into the Matrigel plug could not be excluded. In these transplantation experiments, WT precursors from epWAT showed a robust proliferative capacity when introduced into the scWAT of WT recipients (68% EdU⁺ cells; Fig. 4 C, WT in WT). WT precursors showed a comparable proliferative capacity when transplanted in scWAT of KO recipients (59% EdU⁺ cells; Fig. 4 C, WT in KO). By contrast, the number of proliferating cells was significantly lower when epWAT precursors from KO donors were implanted into the scWAT of KO recipients (41% EdU⁺ cells; Fig. 4 C, KO in KO). Although transplanting these cells into the scWAT of WT recipients improved their proliferative capacity (51% EdU⁺ cells; Fig. 4 C, KO in WT), this increase was not statistically significant when compared with the KO in KO condition. This observation argued that a cell-intrinsic mechanism of SORLA action plays an important role in diet-induced precursor cell proliferation.

Murine embryonic fibroblasts (MEFs) are a well-established model to study mechanisms of adipogenesis (Ruiz-Ojeda et al., 2016). When treated with an induction cocktail containing insulin, rosiglitazone, 3-isobutyl-1-methylxanthine, and dexamethasone, MEFs recapitulate an adipogenic program that entails synchronous entry of progenitors into cell cycle, followed by several rounds of mitosis (mitotic expansion). The resulting committed preadipocytes undergo differentiation to mature adipocytes as evidenced by expression of adipocyte-specific genes and by lipid filling.

We compared the adipogenic potential of MEFs derived from WT or SORLA KO mice to dissect the exact step in which SORLA promotes adipogenesis. After 7 d of differentiation, Oil Red O staining for lipids was substantially reduced in KO MEF cultures compared with WTs (Fig. 5, A and B). We also observed a strong decrease in expression of mature adipocyte markers using Western blotting (Fig. 5, C and D) and quantitative RT-PCR analysis (Fig. S3 D), indicating reduced numbers of adipocytes being produced in the absence of SORLA. To delineate whether these changes were due to defects in cell differentiation, we analyzed expression of markers of early CCAAT/enhancer binding protein (Cebp) Δ and β (Cebpd, Cebpb) and later Cebpa stages of differentiation. Marker levels were comparable between the two genotypes (Fig. S3, A–C), arguing that the reduced lipid accumulation in KO MEF cultures was not due to an impairment in adipocyte maturation from committed preadipocytes.

Multiple signaling pathways are involved in generation of mature adipocytes from progenitors, including signaling through IR (Suryawan et al., 1997) and insulin-like growth factor 1 receptor (IGF1R; Smith et al., 1988), as well as through downstream MAPK and PI3K/protein kinase B (AKT) pathways (Liu et al., 2015; Moseti et al., 2016). Therefore, we compared the levels of pathway components of these differentiation programs in SORLA WT and SORLA KO MEFs treated with the induction cocktail. Prior to treatment, the total levels of IR and IGF1R, as well as of AKT, extracellular-signal regulated kinase (ERK), mitogen-activated protein kinase kinase 4 (MKK4), and JNK, were identical in both genotypes (Fig. S3, E and F). Also, the levels of phosphorylated (active) forms of pAKT, pERK, and pJNK were unchanged, while levels of pMKK4 were slightly increased in SORLA KO MEFs (Fig. S3, E and F). However, 5 min after addition of the induction cocktail, KO MEFs failed to show a robust increase in phosphorylation (i.e., activation) of several pathway components, including pIR, pIGF1R, pAKT, pERK, pMKK4, and pJNK that was readily observed in WT cells (Fig. 5, E and F).

An initial step in adipogenesis entails mitotic expansion of progenitors (Reichert and Eick, 1999). Progression through several rounds of mitosis is coordinated by key regulators of the cell cycle including cyclins D1 and E1 (Baldin et al., 1993) as well as transcription factors FOXO1a and 3a (Schmidt et al., 2002). Levels of cyclin D1, cyclin E1, and FOXO1a/3a were comparable in WT and KO MEF preparations before induction as shown by quantitative Western blot analysis (Fig. 6, A and B). However, 10 min after addition of the induction cocktail, a failure to increase levels of these proteins was evident in KO MEFs as compared with WT cells (Fig. 6, C and D). In line with impaired induction of MAPK and PI3K/AKT pathways and cell cycle regulators, stimulated KO MEFs failed to initiate robust cell proliferation as documented by EdU incorporation (Fig. 6, E and F). By contrast, pronounced cell proliferation was evident in MEFs derived from WT mice following treatment with the induction medium (Fig. 6, E and F).

We substantiated the necessity of SORLA for insulin-induced adipogenesis in the SVF isolated from murine WAT. In agreement with data obtained in MEFs, SVF cells from juvenile epWAT of KO mice failed to increase pIR, pERK, and pJNK levels to the same extent as WT cells following application of induction medium (Fig. 7, A and B, time point 5 min). No difference in baseline levels of pERK and pJNK was observed in the SVF of juvenile epWAT before induction (Fig. 7, A and B, time point 0 min). Importantly, no impact of SORLA deficiency on insulin pathway stimulation (i.e., pIR, pERK, or pJNK levels) was observed in cells isolated from scWAT of juvenile mice (Fig. 7, A and B, time point 5 min). The tissue type–specific action of SORLA in epWAT was confirmed by quantifying cell proliferation using EdU incorporation. Following application of induction medium, the proliferative capacity of SVF cells from epWAT of juvenile KO mice was significantly lower than that of cells from juvenile WT epWAT (Fig. 7 C). No impact of SORLA activity on proliferation was observed comparing cells isolated from scWAT of WT or KO mice (Fig. 7 D). In line with a decreased proliferative capacity, we observed reduced transcript levels for adipocyte differentiation markers cEBPα, LPL, and FABP4 in SVF preparations from KO as compared with WT epWAT 1 wk after treatment with induction medium (Fig. 7 E). No difference in adipocyte marker gene expression was observed comparing SVF from scWAT of WT and KO animals (Fig. 7 F). To corroborate that SORLA specifically impacted the cellular response to insulin, we treated the SFV from epWAT with induction cocktails containing or lacking this hormone. A substantial increase in pIR and pAKT levels in WT cells was only seen with the full cocktail (w/) but not with induction medium lacking insulin (w/o; Fig. 7, G and H). Insulin action on pIR and pAKT levels was blunted in the SVF from KO epWAT treated with the full induction cocktail (Fig. 7, G and H).

To ultimately document the necessity of SORLA for insulin-dependent proliferation and differentiation of adipocyte precursor cells, we repeated the differentiation experiment in flow-sorted Sca1⁺ precursors from juvenile WAT. In agreement with the results obtained in the SVF, Sca1⁺ cells from juvenile KO epWAT failed to increase pIR, pAKT, and pERK levels to the same extent as WT cells following application of induction medium for 5 min (Fig. 8, A and B, stimulated). No difference in baseline levels of pAKT and pERK was observed before induction (Fig. 8, A and B, unstimulated). By contrast, no insulin-induced increase in the levels of pIR, pAKT, or pERK was seen in Sca1⁺ precursors from juvenile scWAT WT as compared with KO mice (Fig. 8, C and D, unstimulated or stimulated). The relevance of this tissue type–specific action of SORLA for adipogenesis was substantiated by comparative expression analysis of adipocyte markers perilipin-1, adiponectin (Adipoq), and PPARγ (Pparg). After 1 wk of differentiation, expression of these markers was significantly lower in Sca1⁺ cell preparations from KO as compared with WT epWAT (Fig. 8, E and F). No impact of receptor deficiency on adipocyte marker expression was observed in Sca1⁺ cells from scWAT of WT or KO animals (Fig. 8, E and G).

To document that the level of SORLA activity is a molecular determinant of the ability of adipocyte precursors to undergo mitotic expansion in response to insulin, we made use of a mouse strain carrying a Cre-inducible SORLA transgene inserted in the endogenous ROSA26 gene locus (Caglayan et al., 2014; Schmidt et al., 2016). To achieve SORLA overexpression in adipocyte precursors, we isolated the SVF from scWAT of juvenile WT mice and treated the cells with a recombinant cell–permeant version of Cre recombinase fused with the protein translocation peptide from HIV-TAT (TAT, trans-activator of transcription) and a nuclear localization sequence (Peitz et al., 2002). Cre-induced activation of the transgene strongly increased levels of SORLA in scWAT cells compared with untreated cells as shown by immunocytochemistry (Fig. 9 A) and Western blotting (Fig. 9 B). Increasing the already high levels of SORLA in the SVF from epWAT did not further raise the high proliferative capacity of these cells in response to insulin (Fig. 9 C). However, raising SORLA levels in scWAT cell preparations significantly enhanced their proliferative capacity when compared with nontreated controls (Fig. 9 C).

Taken together, our data identified SORLA as a molecular factor defining the unique proliferative property of the juvenile epWAT in response to diet. High levels of SORLA expression in adipocyte progenitors of the juvenile epWAT sensitize these cells to nutritional stimuli provided through insulin, promoting expansion of the vis precursor cell pool during overfeeding. By contrast, SORLA activity is relatively low in scWAT progenitors, blunting their response to insulin and preventing diet-induced progenitor pool propagation.

Early studies in humans have demonstrated the different biological properties of visWAT versus scWAT, defining their distinct cardiometabolic risk profiles. Distinctions with particular relevance are timing and conditions that trigger expansion of these fat depots from a resident pool of early adipocyte progenitors. While expansion of scWAT mainly proceeds prenatally, visWAT expansion occurs in two periods in the early postnatal and the subsequent adolescent phase of life (Hirsch and Han, 1969; Holtrup et al., 2017; Kim et al., 2014; Tang et al., 2008; Wang et al., 2013). Also, dietary surplus causes hyperplastic expansion of vis fat, whereas sc fat appears resistant to dietary stress (Jeffery et al., 2015; Kim et al., 2014; Kulenkampff and Wolfrum, 2019).

Our data recapitulate the enhanced precursor cell proliferation in juvenile murine epWAT compared with scWAT in vivo (Fig. 2 C) and in vitro (Fig. 7), and the susceptibility of this process to dietary stress (Fig. 2, D–F). Importantly, we uncovered that this age- and depot-specific effect of the diet on precursor cell proliferation is dependent on SORLA as loss of this receptor reduces the prominent expansion of precursor cell numbers in juvenile epWAT in response to overfeeding (Fig. 2, E and F). SORLA-dependent precursor cell proliferation coincides with the generation of mature adipocyte in MEFs (Fig. 5, A and B), as well as in SVF (Fig. 7, E and F) and Sca⁺ precursor cell preparations (Fig. 8, E–G). Although not shown directly, the SORLA-dependent increase in precursor cell proliferation seen in the epWAT of WT mice in vivo is likely to also translate into larger numbers of mature adipocytes being produced. This conclusion is based on work from others documenting a direct correlation between precursor cell proliferation and generation of mature adipocytes (Berry and Rodeheffer, 2013; Wang et al., 2013).

In earlier studies, inherent mechanisms, such as distinct identities of adipocyte precursors (Gesta et al., 2006; Macotela et al., 2012), but also differences in the microenvironment (Jeffery et al., 2016; Kulenkampff and Wolfrum, 2019), have been held responsible for the unique adipogenic properties of visWAT as compared with scWAT. Our data now identify SORLA as a molecular factor that defines the ability of epWAT precursors to proliferate in response to dietary input, an activity most prevalent in the juvenile organism (Fig. 2, D–F). Because Sorl1 transcript levels correlate with the extent of cell proliferation in juvenile versus adult fat tissues (Fig. 3 B) and because precursor cell proliferation in scWAT can be promoted by increasing its expression levels (Fig. 9), the activity level of SORLA appears to be a crucial determinant of the depot-specific proliferative properties of adipocyte precursors. Our hypothesis is supported by scRNAseq data documenting robust expression of SORL1 in adipocyte precursors in human visWAT, but not scWAT (Fig. 1 H).

Several signaling factors have been implicated in promoting proliferation of adipocyte precursors. For example, transcription factor forkhead box C2 protein enhances cyclin E expression and increases proliferation in vis preadipocyte, likely acting downstream of mammalian target of rapamycin (Gan et al., 2015). Also, estrogen receptors (Zhang et al., 2016b) and inhibitor of metalloprotease-1 (TIMP-1; Zhang et al., 2016a) have been shown to control adipose-derived stem cell proliferation. Whether these mechanisms act differently on precursors from vis and scWAT has not been explored so far. We now document that SORLA promotes the mitotic expansion of precursor cells in vitro and that this mechanism is operable in precursors from epWAT but not scWAT (Fig. 7, C and D), providing a molecular explanation for the inherent proliferative properties of both tissues. By contrast, the receptor activity seems dispensable for subsequent steps in adipocyte maturation as shown in MEFs (Fig. S3, A–C). The latter findings also argue against a nonspecific effect of SORLA deficiency on cell viability, although such an effect cannot be fully excluded.

We also document that the signal(s) that act through SORLA to promote progenitor pool expansion in vivo derive from a nutritional input (Fig. 2, D–F). We cannot deduce the identity of the responsible signaling molecule with absolute certainty, but our findings strongly argue for insulin as shown by the blunted response of SORLA-deficient cells to this hormone in terms of phosphorylation of the IR and AKT (Fig. 7, G and H). In support of our model, insulin conveys signals from nutritional input to adipose tissue (Blüher et al., 2002; Boucher et al., 2010), and short-term HFD exposure induces AKT phosphorylation, a prominent hub of insulin signaling in vis adipocyte precursors (Jeffery et al., 2015). Also, stimulation of the insulin pathway is a major mechanism whereby progenitors are committed into adipogenesis (Green and Kehinde, 1975; Tang et al., 2008), and its function in promoting cell proliferation in adipose tissues was shown in diabetic rats after insulin infusion (Géloën et al., 1989). Conceptually, SORLA deficiency may impact adipogenesis by other defects more fundamental than insulin signaling. Such a hypothesis may be tested by comparing the residual adipogenic potential of WT and KO precursors in response to an induction medium lacking insulin. However, in our hands, even WT cells failed to respond to such conditions with induction of adipogenic gene expression and lipid accumulation (data not shown), in line with data reported by others before (Guller et al., 1988; Négrel et al., 1978).

Foremost, a proposed cell autonomous effect of SORLA on insulin signaling in adipocyte precursor cells is supported by its established function as retrieval factor for the IR. Specifically, SORLA promotes cell surface recycling of internalized IR molecules, thereby facilitating insulin signal reception in target cells (Schmidt et al., 2016). In line with a role for SORLA in promoting insulin signaling, loss of the receptor blunts activation of MAPK pathway components ERK and JNK as well as the PI3K/AKT pathway in adipocyte precursors (Figs. 5, E and F; Fig. 7, A and B; and Fig. 8, A and B). These pathways act downstream of insulin in stimulating adipogenesis (Accili and Taylor, 1991; Tomiyama et al., 1995; Bost et al., 2005; Tang et al., 2005; Park et al., 2014). According to our model, high levels of SORLA specific to juvenile epWAT precursors sensitize these cells to insulin derived from nutritional signals. Consequently, low receptor levels, as in scWAT or in adult epWAT, render these cells relatively insensitive to insulin-derived mitogenic stimuli. While our data provide evidence for the importance of a cell-autonomous action of SORLA in control of precursor cell proliferation, they do not rule out non–cell-autonomous mechanisms whereby the epWAT contributes to a pro-mitogenic milieu. This assumption is supported by our transplantation experiments where precursors in the KO in WT condition trended with intermediate levels of proliferation between KO in KO and WT in WT, although this trend did not reach statistical significance (Fig. 4 C). Conceptually, such non–cell-autonomous mechanisms may be dependent (i.e., ectodomain shedding) or independent of SORLA expression in this tissue.

In humans, SORL1 has been specifically associated with the longitudinal risk of obesity (Smith et al., 2010). Prenatal or juvenile mechanisms implicated in the longitudinal risk of obesity in rodent models include maternal overfeeding during pregnancy or weaning (Guberman et al., 2013; Kayser et al., 2015). Juvenile overfeeding may impact adult risk of obesity for several reasons, potentially including the determination of the adipocyte precursor pool size. In this respect, SORLA-dependent expansion of the adipocyte precursor pool in juvenile epWAT in response to overfeeding may pose a unique mechanism whereby the juvenile metabolism impacts on adult energy homeostasis. Whether increased numbers of precursors are beneficial by increasing the number of mature adipocytes in obesity (Arner et al., 2010; Kim et al., 2014) or detrimental by contributing to diet-induced hyperplasia of vis fat (Muir et al., 2016; Spalding et al., 2008) remains a matter of debate. Irrespective of whether an increased precursor cell pool is ultimately good or bad for adult health, it is tempting to speculate that genetically determined SORL1 levels specific to visWAT precursors define the response of an individual to dietary input at the juvenile age, a process that impacts the progenitor pool size in visWAT and, ultimately, adult metabolism.

Relative protein levels in cells and tissue were determined by densitometric scanning of replicate Western blots using ImageJ software. Antibodies against SORLA were generated in house. The following antibodies were purchased from Cell Signaling: pIR/pIGF1R (#3024), IR (#3025), IGF1R (#9750), pIGF1R (#4568), adiponectin (#2789), perilipin (#9349), GLUT4 (#2213), AKT (#4691), and pAKT-Thr (#2965), ERK (#9102), pERK (#4695), MKK4 (#3346), pMKK4 (#4514), JNK (#9258), pJNK (#9251), FOXO1a (#2880), pFOXO1a/3a (#9464), cyclin D1 (#2978), cyclin E1 (#20808), calnexin (#2433), and β-tubulin (#5568).

For quantitative PCR (qPCR) analysis of SORL1 transcripts, paired samples of vis or sc whole adipose tissue were obtained from 362 individuals (246 women and 116 men). The age ranged from 19 to 93 yr (Klöting et al., 2010). BMI was calculated by weight (kg) divided by square of height (m). All adipose tissue samples were collected during open or laparoscopic abdominal surgery, immediately frozen in liquid nitrogen, and stored at −80°C. The study on human adipose specimens was approved by the Ethics Committee of the University of Leipzig (approval no. 159–12-21052012) and performed in accordance with the Declaration of Helsinki. All subjects gave written informed consent before taking part in this study.

Normalized scRNAseq data reported by Vijay et al. (2020) of the SVF of human visWAT and scWAT were downloaded from GEO (accession no. GSE136230). In brief, data were available from 14,371 cells derived from nine nondiabetic but obese middle-aged individuals (male/female, 2/7; age, mean ± SD, 43.3 ± 11.4; BMI, mean ± SD, 43.3 ± 11.4). Based on expression characteristics of reported cell type markers (Vijay et al., 2020), individual cells were annotated to the indicated subsets of cell types using hierarchical clustering (Euclidian, Ward). SORL1 expressing cells were defined as cells with SORL1 transcripts per million ≠ 0.

Mice with targeted disruption of Sorl1 (referred to as SORLA KO) have been described before (Andersen et al., 2005). All studies were performed comparing male SORLA KO with SORLA WT control animals on an inbred C57BL/6-J background. The animals were kept on a standard mouse chow (4.5% crude fat, 39% carbohydrates; Sniff Deutschland; V1124-3) or on an HFD (35% crude fat, 10% carbohydrate; Sniff Deutschland; E15741-34). All animal experimentation was conducted following approval by local authorities of the Federal State of Berlin (X9007/17, G0156/19).

The SVF was isolated from the epWAT of juvenile donor mice, sorted for adipocyte precursors, and transplanted into the scWAT pad of adult recipients via Matrigel (Corning; #354234) injections. Four animals were used per Matrigel pad. Subsequently, recipients received EdU (ab146186) at 200 µg/ml for 14 d in drinking water to label proliferative cells while being on an HFD for 5 wk (35% fat, 23.9% protein, 4.9% fibers, 5% ashes; Provimi Kliba SA) to induce proliferation in the transplanted tissue pads. Thereafter, the Matrigel pads were carefully excised and fixed overnight in 4% paraformaldehyde. The pads were washed in water and processed with the STP120 Spin Tissue Processor (Thermo Fisher Scientific), followed by paraffin embedding. Histological sections of transplanted tissue pads were immunostained for EdU (SULFO-Cy3-Azide method) and DAPI. The numbers of EdU⁺ cells and total cell count (DAPI) in transplanted tissue pads were determined using ImageJ software.

For proliferation analysis, mice were treated with BrdU (800 µg/ml) in drinking water for 3 d before sacrifice. SVFs were obtained from sc and visWAT for flow-cytometric analysis as follows. Tissues were excised, minced, and digested in collagenase/Hepes solution for 75 min at 37°C. HBSS (Sigma-Aldrich; H8264) with 3% BSA was added to the homogenates and centrifuged for 3 min at 300 × g. Floating adipocytes were removed and the SVFs homogenized in HBSS/3% BSA, then centrifuged again, and the pellets were homogenized and filtered through a 40-µm mesh filter. For counting and sorting of adipocyte precursors, cells were stained for 60 min on ice in HBSS/3% BSA with the following antibodies: CD45 APC-eFluor 780 (eBioscience; 47–0451-82; 1:3,000), CD31 PE-Cy7 (eBioscience; 25–0311-82; 1:1,000), CD29 BV421 (Becton Dickinson; 564131; 1:100), CD34 Alexa Fluor 647 (BioLegend; 119314; 1:200), Sca1 brilliant violet (BioLegend; 108129; 1:200), and CD24 PE (BioLegend; 101808; 1:100). For counting and sorting of other cell types in SVFs, cells were stained for 60 min on ice in HBSS/3% BSA with the following antibodies: CD11b FITC (Miltenyi Biotec; 130–113-234; 1:50), CD8 BV650 (BioLegend; 100741; 1:100), CD19 PE (BioLegend; 152408; 1:200), and CD4 APC (BioLegend; 100412; 1:100). Following antibody incubation, the cell preparations were washed with HBSS/3% BSA and filtered into a FACS tube.

For BrdU analysis, cells were stained as above and then fixed and permeabilized with Phosflow lyse/fix (Becton Dickinson; 558049) and Perm Buffer III (Becton Dickinson; 558050) according to the manufacturer’s recommendations. Cells were treated with DNase (Invitrogen; 18047019; final concentration, 50–375 U) in PBS with 6 mM MgCl₂ and 1 mM CaCl₂ for 2 h in a 37°C water bath. Then the cells were washed in HBSS/3% BSA and incubated with anti-BrdU antibody (Alexa Fluor 488; BioLegend; 364106; 1:100) overnight at 4°C. Finally, the cells were washed and analyzed on a BD FACS Aria 2 analyzer. Data analysis was performed using BD FACS Diva Software and FlowJo V10.

The SVF was cultured from ep and scWAT of mice carrying a Cre-inducible SORLA expression construct inserted into the endogenous Rosa26 gene locus (Caglayan et al., 2014). To induce SORLA overexpression, replicate cell layers were treated with 6.6 µM of purified Cre recombinase protein consisting of an N-terminal 6x His tag, a TAT peptide, and an NLS sequence (HTNCre) and with 100 µM chloroquine for 3 h (Excellgen; EG-1001). Cells were washed and further incubated until confluency. After 3 d, the adipocyte differentiation cocktail was applied to the confluent cell layers and EdU treatment started 18 h later for 6 h before fixing the cells and performing immunodetection of SORLA, EdU, and DAPI staining. Quantification of EdU⁺ and DAPI⁺ cells was done using ImageJ software.

Mice were treated with 200 µg/ml EdU in drinking water for 3 or 7 d. Thereafter, adipose tissues were harvested and fixed in 4% PFA/PBS for 8–16 h at 4°C. After 1 h of washing in PBS, the tissues were dehydrated, embedded in paraffin, and cut in 10-µm sections. For EdU detection, sections were incubated for 5 min in 100 mM Tris, permeabilized in PBS with 0.5% Triton X-100 for 30 min, and washed three times for 5 min in PBS. Then the tissues were incubated with detection solution (2 mM CuSO₄, 20 mg/ml ascorbic acid, and 8 µm SULFO-Cy3-Azide [Lumiprobe; #81330] in PBS) for 30 min, followed by three washes for 5 min in PBS. Finally, the sections were incubated for 5 min in DAPI/PBS, washed again in PBS, and mounted with DAKO Fluorescence Mounting Medium. The number of EdU⁺ cells and the total cell count (DAPI) were determined from three sections per animal using ImageJ software.

Primary adipocytes were differentiated from preadipocytes or Lin⁻CD34⁺CD29⁺Sca1⁺-sorted cells isolated from WAT or from MEFs. In brief, WAT was removed, minced, and digested with collagenase/Hepes solution for 45–60 min at 37°C. Cells were incubated with red blood cell lysis buffer (155 mM NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA) for 5 min and filtered through a 20-µm mesh filter to obtain the SVF. To generate MEFs, embryonic day 14.5 mouse embryos were dissected and trypsinized for 30 min at 37°C. After centrifugation (1,000 ×g, 10 min), the cell pellets were resuspended in DMEM supplemented with 10% FBS and plated on culture dishes.

For adipocyte differentiation, MEFs or stromal vascular cells including preadipocytes were cultured to confluency. 3 d later, the medium was replaced with induction medium containing DMEM supplemented with 5% FBS, 17 nM insulin (Sigma-Aldrich; I6634), 0.1 µM dexamethasone (Sigma-Aldrich; D4902), 250 µM 3-isobutyl-1-methylxanthine (Sigma-Aldrich; I5879), 1 µM rosiglitazone (Sigma-Aldrich; R2408), and 60 µM indomethacin (Sigma-Aldrich; I7378; only for preadipocyte preparation). After the indicated time points, cells were harvested and snap-frozen in liquid nitrogen for further analysis of protein and mRNA levels. For complete adipocyte differentiation, 2 d after induction, the cells were grown in DMEM/10% FBS with insulin for another 2 d. After 7–14 d of culture in normal growth medium (DMEM/10% FBS), the adipocytes were lipid-filled and ready for Western blot analysis, qPCR, or Oil Red O staining.

Differentiated cells were rinsed three times with HBSS (Sigma-Aldrich; H1387) and fixed in cold Baker's Formalin (3.7% formaldehyde and 1% CaCl₂ in distilled water) for 30 min at 4°C. Thereafter, the fixation solution was removed, and filtered 60% Oil Red O solution (Sigma-Aldrich; O0625) was applied to the cells for 10 min at RT. Afterward, the staining solution was removed and the cell preparation cleaned in running tap water (3–5 min). Finally, the cell layers were recovered and the Oil Red O staining intensity determined using a standard Photometer at OD 500 nm.

MEF and adipocyte precursor cells from SVF were grown in 48-well plates until 2 d after confluency. Cells were treated with induction medium to start adipocyte differentiation. After 18 h, EdU was added to the differentiation medium at a final concentration of 10 µM for the indicated time points. Cells were then washed in PBS/3% BSA, fixed in 3.7% formaldehyde/PBS for 15 min, washed again, and permeabilized in 0.5% Triton X-100/PBS for 20 min at RT. The Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen; C10337) was used to detect EdU incorporation into the genomic DNA according to the manufacturer’s instructions.

For transcriptomics analysis, RNA from MEFs or SVF at various stages of adipocyte differentiation were extracted using the RNeasy Lipid Tissue Mini Kit (QIAGEN). The quantity and integrity of the RNA preparation were determined using a NanoVue plus Spectrophotometer (GE Healthcare). 1 μg of total RNA was reverse-transcribed using standard reagents (Life Technologies) and cDNA processed for TaqMan probe–based qPCR using the QuantStudio 6 Flex Real-Time PCR System (Life Technologies). Expression of Cebpd (Mm00786711_s1), Cebpb (Mm00843434_s1), Cebpa (Mm00514283_s1), Pparg (Mm00440940_m1), Lpl (Mm01345521_m1), Fabp4 (Mm00445878_m1), and adiponectin (Mm00456425_m1) was calculated by the standard curve method and normalized to the expression of 18S ribosomal RNA (rRNA; 4352930E) or Gapdh (Mm99999915_g1; Life Technologies) as internal controls.

To determine Sorl1 (Mm01169526_m1) transcript levels in adipocyte precursor cells, the SVF was isolated from ep and sc adipose tissues and FACS-sorted for Lin⁻CD34⁺CD29⁺Sca1⁺ cells, followed by RNA isolation.

The microscopic images were acquired using a Leica Microsystems CMS GmbH apparatus (model DMi8) with 10×, 40×, or 63× objectives and a Leica K5 or DMC2900 camera at RT. Image analysis was performed using Leica Application Suite X (LAS X version 3.7.4.23463) and ImageJ.

Statistical analyses were performed using GraphPad Prism 8 Software. Data are presented as mean ± SEM. For comparing two groups, unpaired two-sided Student’s t test, unpaired two-sided Mann–Whitney test, or paired Wilcoxon test was used. For experiments with more than two groups, unmatched two-sided one-way ANOVA followed by Tukey’s multiple comparisons test was used. For experiments with more than two parameters, unmatched two-sided two-way ANOVA with Sidak’s multiple comparison test was applied. For association studies, linear regression analysis with Pearson correlation coefficient was used. Data distribution for the parametric tests was assumed to be normal, but this was not formally tested.

Fig. S1 shows the multicolor FACS strategy and exemplary scatter plots of cells from the murine SVF of WAT during FACS sorting. Proliferating adipocyte progenitors are identified using the markers CD45⁻CD31⁻CD29⁺CD34⁺Sca1⁺BrdU⁺. Fig. S2 shows exemplary scatter plots and the percent distribution of different cell types after FACS sorting of the murine SVF of WAT. Fig. S3 shows the adipogenic differentiation potential of MEFs from SORLA WT and KO mice. Markers for early differentiation are unchanged while markers for mature adipocytes are reduced in KO animals.

We are indebted to T. Pasternack, R. Vogel, M. Ströhmann, and M. Kahlow for expert technical assistance.

Studies were funded in part by the European Research Council (BeyOND no. 335692), the Helmholtz Association (iCEMED HA-314, AMPro ZT-0026), and the Novo Nordisk Fonden.

The authors declare no competing financial interests.

Author contributions: V. Schmidt designed and performed all experiments; C. Horváth and H. Dong performed the transplantation experiment; P. Qvist analyzed the scRNAseq data; V. Schmidt, C. Horváth, C. Wolfrum, and T.E. Willnow evaluated data. M. Blüher provided essential human biospecimens. T.E. Willnow wrote the manuscript with input from all authors.