Abstract

   Adipogenesis is a tightly orchestrated multistep process wherein
   preadipocytes differentiate into adipocytes. The most studied aspect of
   adipogenesis is its transcriptional regulation through timely
   expression and silencing of a vast number of genes. However, whether
   turnover of key regulatory proteins per se controls adipogenesis
   remains largely understudied. Chaperone-mediated autophagy (CMA) is a
   selective form of lysosomal protein degradation that, in response to
   diverse cues, remodels the proteome for regulatory purposes. We report
   here the activation of CMA during adipocyte differentiation and show
   that CMA regulates adipogenesis at different steps through timely
   degradation of key regulatory signaling proteins and transcription
   factors that dictate proliferation, energetic adaptation, and signaling
   changes required for adipogenesis.
     __________________________________________________________________

   Timely remodeling of a subset of the preadipocyte proteome by CMA
   promotes differentiation into adipocytes.

INTRODUCTION

   The adipose tissue is a major player in fat storage and metabolism.
   Adipocyte differentiation is a dynamic, tightly coordinated, and
   multifactorial process wherein preadipocytes transform into mature
   fat-storing adipocytes in response to lipid availability ([42]1). The
   process of adipocyte differentiation consists of two phases: the
   commitment of multipotent mesenchymal stem cells to preadipocyte and
   the terminal differentiation of preadipocytes into adipocytes ([43]2).
   Adipogenesis involves a temporally regulated pattern of gene expression
   events that drive the activation of pathways including cell
   proliferation, cellular energetics, and lipid synthesis. Sequential
   engagement of these processes is attained by the complex interplay
   between signaling molecules and transcription factors. Most of our
   understanding of adipocyte differentiation is through studies of gene
   expression profiles of the involved factors in cultured preadipocytes.
   However, whether coordinated and timed degradation of key regulatory
   proteins controls adipocyte differentiation remains undetermined.

   Chaperone-mediated autophagy (CMA) is a selective protein degradation
   pathway in which proteins bearing a KFERQ-like motif are degraded in
   lysosomes upon recognition by the cytosolic chaperone heat shock
   cognate protein of 71 kDa (HSC70) ([44]3). Upon arrival to lysosomes,
   substrates bind to the CMA receptor lysosome-associated membrane
   protein 2A (LAMP2A), which multimerizes to form a translocation complex
   for substrates to reach the lumen ([45]4, [46]5). CMA remodels the
   cellular proteome through the turnover of fully functional essential
   regulatory proteins to terminate their function. Previous studies
   identified as CMA substrates proteins involved in the regulation of
   glucose and lipid metabolism and cell cycle and in the initiation or
   termination of specific transcriptional programs ([47]3). Given the
   diversity of proteins that can be selectively turned over by CMA, and
   the fact that some of these cellular pathways are timely activated or
   repressed during adipocyte differentiation, we hypothesized that CMA
   could participate in adipocyte differentiation through the timed
   degradation of regulatory proteins. CMA activity decreases gradually
   with age, and in response to sustained dietary lipid challenges, both
   conditions associated with defective adipogenesis, reduced preadipocyte
   capacity to differentiate, dysfunctional adipocyte-like cells, and
   adipose tissue dysfunction ([48]6).

   Here, we show that inactivation of basal CMA leads to blockage of
   adipocyte differentiation at the early commitment steps. Blockage of
   CMA in adipocyte precursors in mouse results in reduced fat mass with
   abnormally augmented cellular proliferation rates, increased
   inflammation, and fibrosis. We demonstrate that CMA drives adipogenesis
   by sequentially regulating cell cycle arrest, glycolysis, and specific
   transcriptional adipogenesis programs, through the coordinated
   degradation of limiting proteins in each of these processes. Our
   findings underscore the importance of functional CMA for adipogenesis.

RESULTS

CMA activity increases with adipocyte differentiation

   To determine possible changes in CMA activity during adipogenesis, we
   used the preadipocyte line 3T3L1 that differentiates into adipocytes
   when treated with the adipogenic pharmacological agents, insulin,
   dexamethasone, and 3-isobutyl-1-methylxanthine ([49]7). Adipocyte
   differentiation associated with increased LAMP2A mRNA and protein
   levels, but we did not observe an increase in other LAMP2 isoforms
   [[50]Fig. 1, A and B; note that the apparent delay in the increase of
   LAMP2A protein levels relative to mRNA changes is an expected result of
   the complex maturation of the protein through the Golgi and its long
   half-life once in lysosomes ([51]8)]. We found a time-dependent
   increase in CMA activity with adipocyte differentiation using a
   fluorescent CMA reporter ([52]9) that, when delivered through CMA,
   highlights lysosomes as fluorescent puncta ([53]Fig. 1C and fig. S1A).
   Costaining with the lysosomal marker LAMP1 confirmed KFERQ-Dendra^+
   puncta as lysosomes and also revealed that the number of total
   lysosomes remained relatively constant through differentiation (fig.
   S1B). These findings support the combination of two different
   mechanisms, shown before to regulate CMA ([54]10, [55]11), to
   contribute to the observed increase in CMA during differentiation: (i)
   recruitment of lysosomes previously not engaged in CMA
   (LAMP1^+Dendra^−) to perform CMA (LAMP1^+Dendra^+), followed by (ii)
   transcriptional up-regulation and de novo synthesis of LAMP2A, the
   limiting CMA component. These sequential mechanisms for CMA
   up-regulation explain the observed activation of CMA ([56]Fig. 1C) at
   times before the increase in LAMP2A protein ([57]Fig. 1B) because
   LAMP2A is present in all type of lysosomes (active and inactive for
   CMA) and it is the lysosomal acquisition of the luminal HSC70 chaperone
   that makes them competent for CMA ([58]11).

Fig. 1. CMA activity increases during adipocyte differentiation.

   [59]Fig. 1.
   [60]Open in a new tab

   (A) mRNA levels of indicated genes from 3T3L1 cells during adipocyte
   differentiation. n = 6 independent experiments (i.e.). (B)
   Representative immunoblots of indicated proteins from 3T3L1 cells
   during adipocyte differentiation (left) and quantification expressed as
   arbitrary densitometric units (ADU) relative to day 0 (right). n = 4
   i.e. (C) Representative images of 3T3L1 cells stably expressing
   KFERQ-PS-Dendra2 reporter at the indicated times during adipocyte
   differentiation. Insets show boxed area at higher magnification. Nuclei
   are highlighted with 4′,6-diamidino-2-phenylindole (DAPI). Full-field
   images are shown in fig. S1A. Right: Quantification of CMA activity as
   number (left) or total area (right) of fluorescent puncta per cell. n =
   6 i.e. with 40 cells each. (D) Immunostaining for PreF1 of eWAT from
   KFERQ-Dendra-mice maintained on regular diet (RD) or high-fat diet for
   7 days (7d HFD). Left: Representative full-field images. Insets show
   higher-magnification images of preadipocytes and perinuclear region of
   adipocytes. Right: quantification of percentage of PreF1^+ cells
   displaying fluorescent puncta (top) and average number of fluorescent
   puncta per cell (bottom). Nuclei are highlighted with DAPI. n = 4 mice.
   Values are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ns P >
   0.05 using one-way analysis of variance (ANOVA) (A to C), unpaired t
   test (D, top), and two-way ANOVA (D, bottom).

   Similar activation of CMA also occurred in vivo during adipogenesis. We
   induced adipogenesis in a mouse model expressing the same CMA reporter
   (KFERQ-Dendra2 mice) ([61]12) by feeding for 1 week a high-fat diet
   (HFD) ([62]13). While as previously reported for other cells, HFD
   decreased CMA activity of differentiated adipocytes (preadipocyte
   factor 1; PreF1-negative), it induced CMA in adipocyte precursors
   (PreF1-positive) ([63]Fig. 1D; only the perinuclear region of
   adipocytes is shown in the inserts to better appreciate the Dendra^+
   puncta). We conclude that preadipocytes display up-regulated CMA
   activity as they differentiate into adipocytes.

Blocking CMA impairs adipocyte differentiation

   To determine whether preadipocyte CMA activity is required for their
   differentiation, we stably knocked down lamp2a in the preadipocyte cell
   line [L2A(−)] ([64]Fig. 2A). By day 8 of induction of adipogenesis, 70%
   of control (CTR) cells had differentiated into adipocytes, but L2A(−)
   cells failed to differentiate as demonstrated by the absence of lipid
   droplets (LDs) in phase contrast, 1,6-diphenyl-1,3,5-hexatriene (DPH)
   staining, and undetectable perilipin 1 (PLIN1; specific adipocyte LD
   marker) ([65]Fig. 2, B to G). Immunoblot and immunofluorescence
   revealed abnormally elevated levels of PLIN2 (small LD coat protein in
   preadipocytes) and no PLIN1 in the small LD observed in L2A(−) cells
   ([66]Fig. 2, D and E).

Fig. 2. Blocking CMA impairs adipocyte differentiation.

   [67]Fig. 2.
   [68]Open in a new tab

   (A) Immunoblot of 3T3L1 cells control (CTR) or stably knocked down for
   L2A [L2A(−)] and densitometric quantification. n = 5 i.e. (B) Phase
   contrast images at the indicated differentiation days. Green
   fluorescent protein (GFP) shows transduction efficiency. (C) Percentage
   of differentiated cells at day 8 from (B). n = 10 i.e. (D) Immunoblot
   of the indicated proteins and densitometric quantification during
   adipocyte differentiation. n = 3 − 6 i.e. (E) DPH, PLIN1, and PLIN2
   staining at day 9 of adipocyte differentiation. (F and G)
   Quantification of area occupied by LD (F) and PLIN1 intensity (G) from
   (E). n = 4 (F) and 3 (G) i.e. (H) % eWAT/body weight for WT and L2AKO
   mice. n = 10 mice. (I and J) Hematoxylin and eosin (H&E) staining (I)
   and size distribution of cells (J) of eWAT. Zoom: Higher magnification.
   n = 3 mice. (K) eWAT immunostaining for Mac2 and caveolin1 and
   quantification of number of macrophages. Arrowheads: Macrophages
   (Mac2^+ cells). Nuclei are highlighted with Hoechst. n = 4 mice. (L and
   M) mRNA expression of indicated genes (L) and H&E staining (left) and
   quantification of crown-like structures (CLS; arrowheads) (right) (M)
   in eWAT from WT and L2AKO mice on HFD. n = 4 mice. (N) Staining for
   BODIPY 493/503 (to stain LD) and quantification of LD area in in vitro
   differentiated primary preadipocytes from WT and L2AKO mice eWAT at day
   9. Nuclei are highlighted with DAPI. Full-field images are in fig. S1K.
   n = 4 mice with 15 fields each. All insets are boxed areas at higher
   magnification. Values are means ± SEM. *P < 0.05, **P < 0.01, and ***P
   < 0.001 using two-way ANOVA (D and J) or unpaired t test (the rest of
   the panels).

   To study the contribution of CMA to adipogenesis in vivo, we knocked
   out lamp2a specifically in the fibro/adipogenic progenitor cells that
   give rise to white adipocytes ([69]14) by crossing LAMP2A^flox/flox
   mice ([70]15) with platelet-derived growth factor receptor–α
   (PDGFRα)–directed CRE mice (henceforth, LAMP2A knockout mice, ^L2AKO
   mice) that express CRE in most types of adipocyte progenitor cells
   (fig. S1C) ([71]14, [72]16). To study postnatal adipogenesis, we
   performed most of our studies in mice at 1 month of age. ^L2AKO mice
   had decreased epididymal white adipose tissue (eWAT) depot size and
   higher proportion of smaller cells compared to control ([73]Fig. 2, H
   to J) but displayed no size or morphological changes in inguinal WAT
   (iWAT) (fig. S1, D and E). Control and ^L2AKO mice eWAT showed
   comparable oxygen consumption rates (OCR) (fig. S1F) excluding changes
   in energy expenditure as the cause of smaller eWAT cells in ^L2AKO
   mice. Chow-fed ^L2AKO mice eWAT showed a mild proinflammatory phenotype
   with increased macrophage content ([74]Fig. 2K) and transcriptional
   up-regulation of proinflammatory genes, but not in the
   anti-inflammatory genes IL10 and Arginase 1 (fig. S1G). The ^L2AKO mice
   inflammatory phenotype was exacerbated when fed HFD, with increased
   eWAT levels of F4/80 and tumor necrosis factor–α ([75]Fig. 2L) and of
   crown-like structures, a histological signature of macrophage
   infiltration in inflamed adipose tissue ([76]Fig. 2M). Triglyceride
   accumulation in the liver from HFD-fed ^L2AKO mice (fig. S1H) suggests
   a possible failure of the inflamed adipose tissue to adequately store
   lipid. Because we show that there is no L2A loss in the liver of these
   ^L2AKO mice (fig. S1C), this hepatic lipid accumulation is likely due
   to the defect in adipose tissue, causing ectopic lipid deposition.
   Although one of the most widely used CRE drivers to study adipogenesis,
   PDGFRα expression is not exclusive to fibro/adipogenic progenitors and
   can also be detected in cells in the vasculature and in the central
   nervous system ([77]17, [78]18); however, the fact that we did not find
   any difference in vasculature in eWAT (fig. S1I) or in
   hypothalamus-regulated functions such as food intake (fig. S1J) in the
   ^L2AKO mice, at least at the age of our studies, suggests no
   contribution from these extra-adipose sites of potential L2A knockout.
   In further support of the lack of contribution of other cell types or
   organs to the observed adipose phenotype, we isolated primary
   preadipocytes from ^L2AKO mice. As we observed in the 3T3-L1 cell line,
   in vitro differentiation of primary preadipocytes from ^L2AKO mice eWAT
   depots also displayed reduced number of differentiated cells at day 9
   ([79]Fig. 2N and fig. S1K). These findings suggest that CMA controls
   adipogenesis in a cell-autonomous manner.

   As in other cells and tissues ([80]15, [81]19–[82]22), selective
   elimination of LAMP2A in preadipocytes both in vitro or in vivo did not
   interfere with macroautophagy (measured as light chain protein 3, LC3
   II flux) ([83]23) and did not result in compensatory up-regulation of
   this pathway, as in other cell types (fig. S1, L and M) ([84]15,
   [85]19). In summary, our findings show that blocking CMA in
   preadipocytes results in impaired adipocyte differentiation that
   associates with reduced fat mass and increased adipose inflammation.

CMA is important for preadipocyte commitment to differentiation

   To identify the step(s) in the adipogenesis program dependent on CMA,
   we used the preadipocyte cell line because the adipogenic steps and
   timelines are well defined in this cell line ([86]1). In support of
   defect at an early step of differentiation, L2A(−) cells showed
   down-regulation of the terminal adipocyte differentiation markers PLIN1
   and adipocyte protein 2 (aP2) and of the master adipogenesis drivers
   CCAAT/enhancer-binding protein alpha (CEBPα) and peroxisome
   proliferation-activated receptor γ (PPARγ), while CEBPβ expression by
   day 2 of differentiation was elevated and CEBPδ expression remained
   unchanged ([87]Fig. 3A). The elevated levels of the preadipocyte genes
   delta-like noncanonical notch ligand 1 (DLK1) and GATA binding protein
   3 (GATA3) in L2A(−) cells further support a defect in the initial step
   of adipogenesis (days 0 to 2) in these cells. Furthermore, using a
   small-molecule inhibitor of CMA ([88]24) that we confirmed efficiently
   inhibits CMA in preadipocytes (fig. S2A), we found that blockage of CMA
   at different days after commitment (day 2, 4, or 6 onward) did not
   affect the differentiation of 3T3L1 cells, determined by BODIPY
   staining of LD and analysis of expression of PLIN1 and aP2 (fig. S2, B
   and C).

Fig. 3. CMA is required for preadipocyte commitment to differentiation.

   [89]Fig. 3.
   [90]Open in a new tab

   (A) mRNA expression of indicated genes of 3T3L1 cells control (CTR) or
   stably knocked down for lamp2a [L2A(−)] at indicated times of adipocyte
   differentiation. n = 3 to 10 i.e. (B and C) mRNA expression of
   indicated preadipocyte (B) and adipocyte (C) genes of eWAT from WT and
   L2AKO mice. n = 5 mice. (D) Flow cytometry for WT and L2AKO eWAT
   adipose progenitor cells showing percentage of Lin^− PDGFR^+ adipose
   progenitors. n = 3 mice. (E) Left: Immunostaining for PreF1 and
   caveolin1 of eWAT from WT and L2AKO mice. Arrowheads: PreF1^+ cells.
   Right: Quantification of percentage of PreF1^+ cells. Nuclei are
   highlighted with Hoechst. n = 4 mice. Values are means ± SEM. *P <
   0.05, **P < 0.01, ***P < 0.001, and ns P > 0.05 using two-way ANOVA (A)
   and unpaired t test (B to E).

   Similar to L2A(−) 3T3L1 cells, eWAT depots from the ^L2AKO mice also
   displayed aberrant abundance of preadipocytes—increased expression of
   preadipocyte markers DLK1, KROX20, GATA2, GATA3, and TCF7L1 and lower
   expression of differentiated adipocyte genes Leptin, glucose
   transporter 4 (GLUT4), and Adipoq ([91]Fig. 3, B and C). We attribute
   the absence of changes in adipocyte genes such as PLIN1, CEBPα, and
   PPARγ in the ^L2AKO mice to the known heterogeneity of adipocyte
   progenitors in vivo. iWAT depots from the ^L2AKO mice at the same age
   show no significant differences in expression of adipocyte genes,
   although they display a significant increase in KROX20 expression and
   trends to higher levels of other preadipocyte markers, albeit not
   significant and more modest than the ones observed in the eWAT of these
   animals (fig. S2D). Consistent with differentiation failure, ^L2AKO
   mice eWAT depots showed increased number of adipocyte precursors
   [Lin^−CD140^+ cells ([92]25) in stromal vascular fractions or PreF1^+
   cells in depot sections] ([93]Fig. 3, D and E). Once again, we propose
   these changes in gene expression in the eWAT observed upon CMA blockage
   to be mostly of cell autonomous nature, because primary preadipocytes
   from ^L2AKO mice eWAT when differentiated in vitro displayed a
   significantly lower expression of most adipocyte genes analyzed,
   although, in this setting, only one preadipocyte gene (GATA3) was
   elevated in the ^L2AKO mice at day 9 of differentiation (fig. S2E).
   Consistent with these molecular and cellular aberrations, loss of CMA
   resulted in age-associated insulin-resistance (fig. S2F). Our studies
   show that CMA loss impairs the early stages of adipogenesis and results
   in accumulation of undifferentiated preadipocytes.

CMA regulated pathways in adipogenesis

   To identify the protein substrates of CMA that accumulate in
   CMA-deficient preadipocytes, which could contribute to halted
   adipogenesis, we first focused on the persistence of PLIN2 in LD from
   L2A(−) preadipocytes ([94]Fig. 2, D and E, and fig. S3A). As
   differentiation progresses, large LD formation requires exchange of
   PLIN2/3 by PLIN1 ([95]26). PLIN2 degradation by CMA has been described
   in nonadipose cells ([96]27), whereby HSC70 detaches PLIN2 from the LD
   surface and promotes its transfer to lysosomes for degradation without
   exposure to the cytosol through kiss-and-run encounters between CMA
   active lysosomes and LD ([97]27, [98]28). The significant decrease in
   PLIN2 in control preadipocytes after day 2 of differentiation ([99]Fig.
   2D), coinciding with the increase in CMA activity, suggested that PLIN2
   may undergo CMA degradation in preadipocytes. To test whether PLIN2
   degradation by CMA was required for the PLIN2-to-PLIN1 exchange, we
   stimulated adipogenesis in control preadipocytes and, at day 3 (right
   before the time when PLIN1 expression starts), transfected them with a
   plasmid expressing WT-PLIN2 or a previously described CMA-resistant
   mutant PLIN2 [missing the CMA targeting motif ([100]27) and fig. S3B].
   Cells expressing CMA-mutant PLIN2 failed to differentiate into
   adipocytes (fig. S3C), suggesting that the removal of PLIN2 from the
   growing LD occurs via CMA and it is required to progress to full
   differentiation. PLIN2 is also known to be degraded by the
   ubiquitin-proteasome system in differentiated adipocytes ([101]29), and
   we also observed a discrete increase in levels of PLIN2 in our in vitro
   preadipocyte cultures upon addition of the proteasome inhibitor
   lactacystin, in support of some level of basal degradation of PLIN2
   through the proteasome (fig. S3D). Proteasome degradation of PLIN2 was
   comparable between control and L2A(−) cells up to day 4 of
   differentiation when we found it to be decreased in L2A (−) cells (fig.
   S3D). This decrease in proteasome degradation seems specific for PLIN2
   because the degradation rates of the pool of K48-specific ubiquitinated
   proteins were unchanged in CMA-deficient preadipocytes at all times
   tested (fig. S3E).

   Because, besides PLIN2 persistence, we identified pronounced
   transcriptional changes in master adipogenesis drivers ([102]Fig. 3),
   an event that precedes changes in LD morphology, we next tested whether
   CMA may also facilitate timely removal of key components earlier in the
   adipogenesis program. To identify proteins that may act as inhibitors
   of adipogenesis when not removed by CMA, we first analyzed cellular
   pathways disrupted in L2A(−) preadipocytes at days 0 and 2 of
   differentiation. Because the transcriptional profile of adipocyte
   differentiation in 3T3L1 cells has been well characterized and CMA has
   been shown to degrade specific transcription factors or their
   inhibitors ([103]3) and to regulate transcriptional programs in other
   settings ([104]21, [105]30), we performed comparative transcriptomics.
   Transcriptional analysis revealed that L2A(−) cells had a blunted
   response to differentiation with only approximately 18 and 42% of the
   up-regulated and down-regulated genes coinciding with control cells
   ([106]Fig. 4, A and B). We identified 143 genes that increased at basal
   levels and that failed to decrease in L2A(−) cells with
   differentiation, which were related to cell cycle, signaling pathways,
   and metabolism ([107]Fig. 4, C to G). Many of the signaling pathways
   up-regulated in L2A(−) cells [such as Sirtuin, Wnt, and protein kinase
   A (PKA) signaling] are known to suppress preadipocyte differentiation
   ([108]Fig. 4, D to G). The 52 genes that basally decreased in L2A(−)
   cells and that failed to increase at day 2 coded for proteins that
   participate in cytoskeleton, tight junction, and gap junction signaling
   among others (fig. S4, A to C).

Fig. 4. CMA blockage deregulates functional pathways involved in adipocyte
differentiation.

   [109]Fig. 4.
   [110]Open in a new tab

   (A) Differentially expressed genes in CTR and L2A(−) 3T3L1 cells,
   comparing days 0 and 2. Venn diagrams depict the up-regulated (top) and
   down-regulated (bottom) genes in the two cell lines. (B) Heatmap and
   hierarchical clustering analysis between CTR and L2A(−) cells at days 0
   and 2 of differentiation. (C) Venn diagram of genes increased basally
   in L2A(−) cells and those that fail to decrease in L2A(−) with
   differentiation when compared to CTR. (D and E) Gene ontology
   enrichment analysis of genes with constitutively (basal) increased
   expression (D) or that fail to decrease with differentiation (E) in
   L2A(−) cells when compared to CTR. (F and G) Pathway enrichment
   analysis of genes with constitutively (basal) increased expression (F)
   or that fail to decrease expression with differentiation (G) in L2A(−)
   cells when compared to CTR. (H to K) Predicted upstream activators of
   genes with constitutively (basal) increased expression (H) or that fail
   to decrease expression with differentiation (J) in L2A(−) cells when
   compared to CTR. Hierarchical arrangement places Myc on top of upstream
   regulators constitutively activated in L2A(−) and TGFβ on top of
   upstream regulators that fail to decrease during differentiation [see
   detailed maps in fig. S4 (D and E)]. Percentage of proteins bearing CMA
   motifs (KFERQ-like motifs) among the proteins shown in (H) and (J) are
   shown in (I) and (K), respectively. All gene ontology terms are
   significant for P < 0.001.

   To identify the specific CMA substrates important for adipogenesis
   regulation, we used the Ingenuity Pathway Analysis tool that predicts
   differentially activated upstream regulators from the transcriptional
   profile of L2A(−) and control cells. We identified 55 and 54 upstream
   activators (with z score ≥ 2), respectively ([111]Fig. 4, H and J),
   with a majority of them (78 to 80%) bearing the CMA-targeting motif in
   their sequence ([112]Fig. 4, I and K). Hierarchical mapping of these
   upstream regulators identified Myc and transforming growth factor–β
   (TGFβ) ([113]Fig. 4, H and J, insets, and fig. S4, D and E) as the
   major proteins responsible for the transcriptional changes in
   downstream proliferation, metabolism, transcription, and signaling
   events required for adipogenesis.

   Changes in the transcriptional profile of CMA-deficient preadipocytes
   matched in terms of overall z score, upstream regulators, and
   downstream effectors with profiles reported in human adipose tissue
   under experimental proinflammatory conditions or adipose from patients
   with common metabolic disorders such as obesity or diabetes (fig. S5A).
   Furthermore, comparison of the signature heatmap of CMA-deficient
   preadipocytes and the human disease adipose tissues also revealed
   matching in many entities (upstream regulators, downstream effectors,
   canonical pathways, and causal networks) (fig. S5B).

   Overall, these findings support that CMA failure during adipogenesis
   prevents engagement and repression of specific transcriptional pathways
   previously proven essential for adipogenesis, namely, the proliferative
   burst, the metabolic switch, and the termination of TGFβ signaling.

CMA loss leads to sustained preadipocyte proliferation

   Central to adipogenesis is the transition between cell proliferation
   and differentiation ([114]31). Because our gene expression analyses
   predicted the upstream regulator of proliferation Myc, to be activated
   in L2A(−) cells (fig. S4D), we first evaluated the cell cycle state in
   control and L2A(−) cells during early stages of differentiation—when
   growth-arrested cells undergo one or two rounds of cell division for
   initiation and commitment to differentiation ([115]31). Contrary to
   control cells that shift from 10 to 30% proliferating cells 24 hours
   after initiation of differentiation, 30% of L2A(−) preadipocytes were
   proliferating on day 0 and remained at 35% on day 1 ([116]Fig. 5A).
   Similarly, eWAT depots from ^L2AKO mice also show elevated
   proliferation based on increased expression of Cyclin D1 and increased
   Ki67 staining when compared to wild-type (WT) littermates ([117]Fig. 5,
   B and C).

Fig. 5. Lack of CMA in preadipocytes leads to sustained proliferation and
glycolysis during differentiation.

   [118]Fig. 5.
   [119]Open in a new tab

   (A) Cell cycle analysis of 3T3L1 cells control (CTR) or stably knocked
   down for lamp2a [L2A(−)] at days 0 and 1 of differentiation. Left:
   Representative histograms. Right: Quantification of phases of cell
   cycle. n = 3 i.e. (B) mRNA expression of indicated gene of eWAT. n = 5
   mice. (C) Left: Immunostaining of eWAT for Ki67 and quantification of
   percentage of Ki67^+ cells (arrowheads). Insets: Higher magnification.
   n = 4 mice. (D) Immunoblot of CTR and L2A(−) 3T3L1 cells at indicated
   times of adipocyte differentiation (left) and quantification of protein
   levels (right). n = 4 i.e. (E to G) Immunostaining for Myc (E) and
   quantification of Myc levels in cytosol (F) and nuclei (G) in 3T3L1
   cells at day 0 untreated or treated with inhibitors of lysosomal
   proteolysis (Lys Inh). Nuclei are highlighted with DAPI. n = 3 i.e. (H
   to J) Extracellular acidification rates (ECAR) (H), energy phenotype
   (I) and glycolysis, glycolytic capacity, and glycolytic reserve (J) in
   3T3L1 cells CTR or L2A(−) at days 0 to 4 of differentiation. n = 4 i.e.
   (K) Immunoblot and quantification of 3T3L1 cells at indicated times of
   differentiation. n = 9 i.e. (L) Immunoblots and lysosomal degradation
   rates for indicated proteins in 3T3L1 cells untreated or treated with
   Lys Inh at indicated times of differentiation. LC3 is a positive
   control for Lys Inh. n = 7 i.e. Values are means ± SEM. *P < 0.05, **P
   < 0.01, ***P < 0.001, and ns P > 0.05 using two-way ANOVA (A, D, F, J,
   and L) and unpaired t test (B, C, G, and K).

   Because Myc bears two canonical CMA-targeting motifs (^267VEKRQ and
   ^361VLERQ) and its levels were elevated in L2A(−) cells at day 0
   ([120]Fig. 5D), we explored whether Myc could undergo CMA degradation
   during adipogenesis. Addition of lysosomal inhibitors in control cells
   revealed lysosomal degradation of Myc during adipogenesis that mirrored
   Myc transcriptional changes, whereby maximal transcriptional expression
   coincided with a decrease in Myc lysosomal degradation and
   down-regulation of transcription with a return to basal rates of
   lysosomal degradation (fig. S6B compared to fig. S6A). In contrast,
   lysosomal turnover of Myc was undetectable in L2A(−) cells, thus
   explaining its higher levels in these cells (fig. S6B). In agreement
   with previous studies ([121]32), we noted that a substantial proportion
   of Myc is also degraded by the proteasome with a similar pattern to
   lysosomal degradation in control cells (fig. S6C). Despite proteasomal
   degradation of Myc remaining constant in L2A(−), we observed higher
   basal Myc nuclear levels in these cells, consistent with their
   proliferative phenotype ([122]Fig. 5, E to G). Degradation of cytosolic
   Myc in lysosomes, observed in control cells, was not detected in L2A(−)
   cells, thus explaining their higher cytosolic levels ([123]Fig. 5, E
   and F). We conclude that failure of CMA to degrade Myc, a key driver of
   cellular proliferation, is behind the increase in proliferation and
   dysregulation of cell cycle observed in CMA-defective preadipocytes.

Aberrant glycolysis contributes to defective differentiation of CMA-deficient
preadipocytes

   The clonal expansion of preadipocytes in the initial steps before
   differentiating into adipocytes requires a substantial energy burst
   ([124]33). Glycolysis was the top transcriptionally up-regulated
   pathway in L2A(−) cells at the basal state that remained elevated upon
   differentiation ([125]Fig. 4, D and E). We found that, while control
   cells showed an increase in glycolysis at day 2, which decreased by day
   4 of differentiation to basal levels ([126]Fig. 5, H to J),
   CMA-deficient cells displayed dysregulated glycolysis rates that did
   not respond to adipogenic cues of differentiation ([127]Fig. 5, H to
   J). L2A(−) preadipocytes remained in the energetic phenotype quadrant
   throughout differentiation, instead of the transition from quiescent to
   energetic and back to the quiescent phenotype of control cells
   ([128]Fig. 5I).

   We propose that these rapid changes in glycolysis in control cells
   during adipogenesis may require a combination of transcriptional
   down-regulation and timely removal of the glycolytic enzymes to
   terminate their function, as reported in other tissues ([129]15).
   Although most glycolytic enzymes in L2A(−) cells displayed higher basal
   protein levels ([130]Fig. 5K), not all of them were transcriptionally
   up-regulated (fig. S7, A and B). Lysosomal degradation assays revealed
   that each of these glycolytic enzymes was degraded through CMA in
   control cells at a rate and timing unique for each enzyme ([131]Fig.
   5L; note that plotting represents the lysosomal degradation rate for
   each protein). By contrast, L2A(−) cells had negligible lysosomal
   degradation for these three glycolytic enzymes ([132]Fig. 5L) that did
   not correlate with changes in their gene expression (fig. S7, A and B)
   and that explain why levels of these enzymes remained elevated by day
   4, when they had returned to basal levels in control cells (fig. S7C).
   Studies with the proteasome inhibitor lactacystin revealed no
   significant differences between control and L2A(−) cells in the
   fraction of glycolytic enzymes degraded by the proteasome (fig. S7, D
   and E). While we found similar glycolysis activity in control and
   L2A(−) cells at day 2 of differentiation ([133]Fig. 5J, top), in
   agreement with glycolytic enzymes reaching similar levels in both cell
   types ([134]Fig. 5K), glycolytic capacity and reserve ([135]Fig. 5J,
   middle and bottom) were markedly elevated in L2A(−) cells, suggesting
   that blockage of CMA may promote the higher activity of the glycolytic
   enzymes through mechanisms other than changes in their cellular levels
   [i.e., through changes in activator factors or in posttranslational
   modifications ([136]34–[137]36)].

   In support of the timely nature of the CMA-imposed changes in
   glycolysis, which occurs only at a specific time during adipogenesis,
   the levels of these enzymes in adipose tissue of ^L2AKO mice were no
   longer elevated when compared to WT tissue (fig. S7F). The observed
   trend toward decreased levels of glycolytic enzymes in eWAT of ^L2AKO
   mice could be reactive to the reduced fat mass and more inflamed state
   of the tissue ([138]Fig. 2, H to M), as also supported by the observed
   decreased expression of glycolytic and lipogenic enzymes (fig. S7G).
   Our findings indicate that CMA regulates glycolysis rates during
   adipogenesis and that failure to timely degrade glycolytic enzymes
   leads to uncontrolled glycolysis in preadipocytes and adipogenesis
   failure.

Timed inhibition of TGFβ/SMAD signaling by CMA is required for adipocyte
differentiation

   Hierarchical mapping of the comparative transcriptomics data from
   control and L2A(−) preadipocytes predicted TGFβ signaling to be the
   major upstream mediator of the downstream defects during the
   differentiation of CMA-deficient preadipocytes ([139]Fig. 4J and fig.
   S4E). TGFβ inhibits adipogenesis in vitro ([140]37) and associates with
   adipose tissue inflammation and fibrosis ([141]38). We found that eWAT
   depots from ^L2AKO mice exhibited up-regulated expression for TGFβ
   ([142]Fig. 6A), and this activation of TGFβ associated with
   significantly increased eWAT fibrosis as determined by trichrome
   staining ([143]Fig. 6B) and with increased expression of
   fibrosis-related genes and of CD9, known to associate with increased
   fibrogenesis ([144]Fig. 6C) ([145]39). Expectedly, the fibrogenic
   phenotype of ^L2AKO eWAT was exacerbated when mice were fed HFD
   ([146]Fig. 6D).

Fig. 6. Dysregulated TGFβ signaling is responsible for the adipogenesis
defect in CMA-deficient preadipocytes.

   [147]Fig. 6.
   [148]Open in a new tab

   (A) mRNA expression of indicated gene of eWAT from WT and L2AKO mice. n
   = 8 mice. (B) Trichrome A staining of eWAT and quantification of
   percentage of Trichrome^+ area. n = 5 mice. (C) mRNA expression of
   indicated genes in eWAT. n = 5 mice. (D) Trichrome A staining of eWAT
   from WT and L2AKO mice on HFD and quantification of percentage of
   Trichrome^+ area. n = 4 mice. (E) Representative immunoblots for
   indicated proteins from eWAT of WT and L2AKO mice and quantification. n
   = 10 mice. (F) Representative immunoblots for indicated proteins and
   quantification of CTR and L2A(−) 3T3L1 cells at indicated times of
   adipocyte differentiation. n = 4 i.e. (G) Representative immunoblots
   and quantification of SMAD2/3 lysosomal degradation rates in CTR and
   L2A(−) 3T3L1 cells untreated or treated with Lys Inh at indicated times
   of adipocyte differentiation. n = 7 i.e. (H to J) Immunostaining of
   SMAD2/3 (H) and quantification of SMAD2/3 levels in cytosol (I) and
   nuclei (J) in 3T3L1 cells at day 0 untreated or treated with Lys Inh.
   Image shows color mapping by gradient intensity. n = 3 − 4 i.e. (K) DPH
   staining and quantification of percentage of differentiated cells in
   3T3L1 cells at day 7 untreated or treated with TGFβ inhibitor for 12
   hours at day 1.5. n = 4 i.e. Values are means ± SEM. *P < 0.05, **P <
   0.01, and ***P < 0.001 using unpaired t test (A to E and J), two-way
   ANOVA (F, G, and I), and one-way ANOVA (K).

   eWAT depots of ^L2AKO mice and L2A(−) preadipocytes at day 2 of
   differentiation displayed elevated levels of phosphorylated SMAD
   homolog 2/3 (SMAD2/3), downstream effectors of TGFβ that inhibit the
   positive feedback loop between the master regulators of adipogenesis
   PPARγ and CEBPα ([149]Fig. 6, E and F) ([150]40, [151]41). Because
   SMAD2/3 have a phosphorylation-dependent CMA-targeting motif
   (^409KVLTQ), we assessed their lysosomal degradation and found
   substantial degradation of this protein by lysosomes at days 0 and 2 in
   control cells detected both by immunoblot and immunofluorescence and as
   higher colocalization with lysosomal markers, which were largely
   abolished in L2A(−) cells ([152]Fig. 6, G to I, and fig. S8A). Changes
   in proteasomal degradation cannot account for the increase in SMAD2/3
   levels in L2A(–) cells (fig. S8B). As anticipated, persistently higher
   cytosolic levels of SMAD2/3 in L2A(−) cells resulted in significantly
   elevated nuclear content of SMAD2/3 in these cells ([153]Fig. 6J and
   fig. S8C).

   To determine whether sustained elevated TGFβ/SMAD signaling was
   responsible for the differentiation failure of L2A(−) preadipocytes, we
   pharmacologically attenuated this signaling with the TGFβ inhibitor
   SB525334 administered for 12 hours before the adipocyte commitment
   phase and found that this intervention was enough to partially rescue
   the adipogenic defect of L2A(−) cells ([154]Fig. 6K and fig. S8D).
   These findings suggest that CMA plays a key role in adipocyte
   differentiation by temporally suppressing TGFβ SMAD signaling in
   preadipocytes. To analyze the possible interdependence of the observed
   changes in TGFβ signaling and glycolysis in preadipocytes upon CMA
   blockage, we first up-regulated glycolysis in control cells with ML265,
   a glycolysis activator. We found that glycolysis up-regulation, to
   mimic the one observed in L2A(−) cells, did not result in significant
   changes in TGFβ signaling (fig. S8, E and F). Next, we inhibited TGFβ
   signaling in control cells and found a mild increase in glycolysis
   (fig. S8G) in support of the idea that the elevated TGFβ signaling
   observed in L2A(−) cells was not directly responsible for the observed
   up-regulated glycolysis in these cells. In summary, our studies have
   identified CMA as a major orchestrator of three distinct time-dependent
   nodes—proliferation, metabolism, and transcriptional regulation—each of
   which contributes to essential aspects of adipocyte differentiation.

DISCUSSION

   We have identified a key role for CMA in adipocyte differentiation that
   relies on the selective and timely degradation to terminate the
   function of proteins involved in cellular proliferation (Myc),
   energetics (glycolytic enzymes), and transcriptional regulation
   (SMAD2/3) (fig. S9). Although maximal CMA activity is reached by day 6
   of adipogenesis, functional CMA is necessary from the earlier stages
   during the determination phase of adipogenesis. Our data support that
   termination of specific cellular pathways required for adipogenesis
   progression is achieved not only by the transcriptional down-regulation
   of genes in that pathway but also through the selective degradation of
   key regulatory proteins coded by those genes. In this study, we
   demonstrate that CMA regulates the duration of diverse cellular
   processes critical for adipogenesis completion.

   A role for macroautophagy in adipocyte differentiation has been
   reported ([155]42–[156]44), but CMA and macroautophagy are
   nonredundant, and each of them contributes to specific aspects of
   adipogenesis. Macroautophagy-defective mice display reduced WAT mass,
   increased fat browning, mitochondrial enrichment, elevated fatty acid
   oxidation, and no changes in adipocyte progenitor numbers ([157]42,
   [158]43). Inhibition of macroautophagy after development results
   instead in eWAT inflammation, mitochondrial dysfunction, and insulin
   resistance without affecting adiposity ([159]44). Studies on mouse
   models with macroautophagy inhibition specifically in PDGFRα^+
   progenitor cells (using PDGFRα-CreERT2) have reported protection
   against HFD-induced adipose fibrogenesis and increase in adipose
   browning ([160]45). In contrast, the phenotype of CMA deficiency in
   PDGFRα^+ progenitors is dominated by reduced fat mass, disruption of
   early phase of adipogenesis with increased number of adipose
   progenitors, elevated proliferation, increased inflammation and
   fibrogenesis, and no changes in browning or fatty acid oxidation.
   Similarly, the ubiquitin proteasome system also contributes to
   adipogenesis ([161]46) through degradation of CEBPα and PPARγ
   ([162]47–[163]51). In the case of CMA-deficient preadipocytes, we
   observed rerouting of some of the identified CMA substrates to the
   proteasome (i.e., Myc and SMAD2/3) but not for others (i.e., enolase,
   GAPDH and hexokinase). However, even for the CMA substrates rerouted to
   the proteasome because proteasome degradation occurred at different
   times than by CMA, the functional defect in adipogenesis was not
   prevented. These findings highlight the unique roles of distinct
   autophagy pathways and the proteasome in adipogenesis, with each
   pathway catering to a specific cellular function. In this work, we
   noted that CMA blockage in iWAT led to a more discrete phenotype than
   the one observed in eWAT. Future studies are needed to discriminate
   whether these differences are a consequence of other cell lineages
   compensating early on in iWAT development or due to adipose
   depot-specific roles of CMA.

   Despite the gradual increase in CMA activity as preadipocyte
   differentiation progresses, the fact that blockage of CMA even before
   it becomes up-regulated blunts the adipogenesis process highlights that
   basal CMA activity is necessary and sufficient for the initiation of
   adipogenesis. Later stages in the adipogenic program may require
   up-regulation of preadipocyte CMA above the normal basal levels to
   accommodate the larger number of proteins targeted for degradation, as
   remodeling of the proteome of a preadipocyte to become an adipocyte
   becomes more extensive. In this work, we have identified that
   adipogenesis-related changes in three different processes—signaling,
   proliferation, and glycolysis—require fully functional CMA. A
   multifunctional effect of this nature is expected for a proteolytic
   pathway when considering the functional variety of substrate proteins,
   and it highlights the role of protein degradation in the regulation of
   cellular functioning.

   Reduced adipogenesis, like the one we observed in both in vitro and in
   vivo models of CMA-deficiency, associates with up-regulation of the
   preadipocyte marker, PreF1, which per se is known to inhibit
   adipogenesis ([164]52). In addition, many of the known inhibitors of
   adipocyte differentiation (such as TGFβ, SMAD3, Wnt, Sirtuin, PKA, and
   GATA2/3) ([165]53) were also up-regulated in our CMA-deficient models.
   We propose that termination of TGFβ/SMAD3 signaling may be one of the
   main mechanisms behind the regulatory effect of CMA on adipocyte
   differentiation, because pharmacological inhibition of TGFβ was
   effective in rescuing adipocyte differentiation in the CMA-defective
   models.

   The remarkable coincidences between the transcriptional profile of
   CMA-defective preadipocytes and human adipose tissue from common
   metabolic conditions closely related to the metabolic syndrome of
   aging, and the fact that CMA blockage in preadipocytes results in
   impaired adipogenesis, increased adipocyte progenitors, hyperplasia,
   inflammation, and fibrosis in the adipose tissue, make attractive the
   possibility that age-related decline of CMA activity in preadipocytes
   may contribute to the adipose tissue dysfunction characteristic of old
   organisms.

MATERIALS AND METHODS

Mice

   Adipose progenitor-specific LAMP2A KO mice (L2AKO) were generated by
   crossing C57BL/6 PDGFRα-Cre mice (JAX mice, stock 013148) with C57BL/6
   L2A^f/f mice ([166]15). The control experimental group (named WT) was
   composed of either L2A^f/f or PDGFRα-Cre littermate mice as no
   differences were detected among them. Most experiments were done with
   4- to 5-week-old mice (when eWAT development is completed, and to rule
   out any compensatory changes that may occur with age in L2AKO mice).
   HFD experiments were done with 14-week-old mice that were on HFD (60%
   kcal from fat; Research Diets, D12492) for 8 weeks. KFERQ-Dendra2
   transgenic mice ([167]12) at 4 weeks of age were given HFD for 7 days.
   Only males were used in this study owing to the important differences
   in female metabolism that will require a full study on its own. All
   mice were genotyped at weaning, and genotyping was reconfirmed
   postmortem to correct for any possible misplacement during husbandry.
   Mice were all in the C57BL/6J background and housed in ventilated cages
   with no more than five mice per cage on a 12-hour light/12-hour dark
   cycle at 23°C with free access to water and food in the institutional
   barrier facility along with sentinel cages and were specific
   pathogen–free. All mouse procedures were approved by the Institutional
   Animal Care and Use Committee of Albert Einstein College of Medicine.

Cells

   3T3-L1 murine preadipocyte cells (American Type Culture Collection,
   CL173) stably transduced with lentiviral vectors for control, knockdown
   of lamp2a ([168]54), or KFERQ-Dendra2 CMA reporter ([169]9, [170]12)
   were maintained in Dulbecco’s modified Eagle’s medium (DMEM)
   (Sigma-Aldrich, D5648) supplemented with 10% fetal bovine serum (FBS)
   (Thermo Fisher Scientific, 35010CV) and antibiotics (Thermo Fisher
   Scientific, 15240112) at 37°C and 5% CO[2]. The 3T3-L1 cell line was
   authenticated by karyotyping at regular intervals to confirm their
   identity and tested for mycoplasma contamination using a DNA staining
   protocol with Hoechst 33258 dye (Invitrogen, H3570) (weekly) or the
   MycoSenser PCR Assay Kit (monthly). All cells used in the present study
   were free of such contamination.

Reagents

   Sources of chemicals and antibodies including dilutions are indicated
   in table S1. We have also provided catalog numbers or clone numbers for
   all the commercial antibodies in table S1. Most antibodies used are
   commercial and have been chosen on the basis of extensive use in the
   literature. When available, antibodies were validated using cell lines
   knocked down for the antibody target (i.e., in the case of antibodies
   against LAMP2A, we used cell lines and tissues knocked down or tissues
   from knockout mice as negative controls).

Cell treatments

   Adipocyte differentiation was initiated by culturing confluent cells in
   adipogenic induction medium consisting of the growth medium
   supplemented with 160 nM insulin (Sigma-Aldrich, I9278), 250 nM
   dexamethasone (Sigma-Aldrich, D4902), and 0.5 mM
   1-methyl-3-isobutylxanthine (IBMX, Sigma-Aldrich, I7018) (day 0).
   Forty-eight hours later (day 2), the cells were switched to maintenance
   medium supplemented just with insulin. After 48 hours (day 4), the
   cells were placed back in unsupplemented maintenance medium, which was
   changed every 48 hours. For lysosomal flux assays, cells were treated
   with 20 mM ammonium chloride (Sigma-Aldrich, AX12701) or 100 μM
   leupeptin (Thermo Fisher Scientific, [171]BP266225) for 12 hours at the
   indicated time points in differentiation. For proteasomal flux assays,
   cells were treated with lactacystin (5 μM; Enzo Life Sciences,
   BMLPI104) for 12 hours at the indicated time points, and efficacy of
   inhibition was monitored by tracking changes in K48-ubiquitinated
   proteins. TGFβ inhibitor (SB525334, Tocris, 3211) was added to
   differentiating cells at day 1.5 at 1 μM for 12 hours. ML265 (Cayman
   Chemical, 13942) was added to activate glycolysis in control cells at 1
   and 2 μM for 12 hours.

Cell imaging

   Cells grown on coverslips were fixed with 4% paraformaldehyde for 30
   min, permeabilized, and blocked with 1% bovine serum albumin (BSA) and
   0.01% Triton X-100 in phosphate-buffered saline (PBS). Incubation with
   primary and secondary antibodies conjugated to fluorophore in 0.1% BSA
   in PBS was performed at room temperature for 1 hour each. For LD
   staining, cells were incubated with DPH (5 μM; Sigma-Aldrich, D208000)
   or BODIPY 493/503 (20 μg/ml; Thermo Fisher Scientific, D3922) for 20
   min at 37°C before fixation. For cells expressing CMA reporter
   (KFERQ-Dendra2), cells grown on coverslips were photoconverted (green
   to red) with a 405/20-nm LED array (Norlux) for 3 min using a light
   intensity of 50 mW/cm^2. Cells were fixed 16 hours later and processed
   for immunofluorescence as described above. Where indicated, coverslips
   were mounted in 4′,6-diamidino-2-phenylindole–Fluoromount-G (Southern
   Biotechnology, 0100-20) or incubated with DRAQ5 (Cell Signaling
   Technology, 4084) to stain the nucleus. For phase microscopy, live
   cells growing in six-well plates were used. Images were acquired with
   an Axiovert 200 fluorescence microscope (Carl Zeiss Microscopy) mounted
   with an ApoTome.2 slider or a Leica TCS SP8 confocal microscope using
   40× or 63× objective and 1.4 numerical aperture. Images were prepared,
   and channels were given pseudo-colors (in some instances) and
   quantified using ImageJ (National Institutes of Health). The number and
   area of puncta were quantified using “analyze particles” or “3D object
   counter” (ImageJ) after thresholding in nonsaturated images. The
   percentage of colocalization was determined by “JACoP” plugin (ImageJ)
   after thresholding of individual channels ([172]55). Fire lookup table
   (LUT) was used for color mapping by gradient intensity. Fluorescence
   intensity in nucleus/cytoplasm was measured by highlighting the
   nucleus/cell boundary using “free hand” tool (ImageJ). Antibodies used
   for cell imaging were LAMP2 (1:500; Developmental Studies Hybridoma
   Bank, GL2A7), Myc (1:800; Cell Signaling Technology, 5605), PLIN1
   (1:400; Cell Signaling Technology, 9349), PLIN2 (1:400; PROGEN
   Biotechnik, GP40), and SMAD2/3 (1:500; Cell Signaling Technology,
   8685).

Tissue imaging

   The Histology and Comparative Pathology Core at Albert Einstein College
   of Medicine performed the tissue embedding and processing according to
   standard protocols. Paraffin-embedded sections (5 μm thick) of
   formalin-fixed adipose tissues were subjected to hematoxylin and eosin
   (H&E) staining and trichrome staining as per standard protocols.
   Sections were imaged using Leica light microscope DMD108. Morphometric
   analysis of cell size was performed on H&E sections using the “cell
   magic wand” tool in ImageJ. Cell area of 250 to 300 cells per section
   was determined. Immunostainings of eWAT sections were performed by
   citric acid–based antigen retrieval at 95°C for 30 min (Vector
   Laboratories, H-3300) of paraffin-embedded tissues, followed by
   permeabilization and blocking in PBS with 5% goat serum, 3% BSA, and
   0.3% Triton X-100 for 1 hour at room temperature, and by incubation
   with primary antibody and fluorophore-conjugated secondary antibody
   prepared in 3% BSA and 0.1% Triton X-100 overnight at 4°C and 1 hour at
   room temperature, respectively. After antibody incubation, the tissue
   sections were washed five times in PBS and counterstained with Hoechst
   33342 (1:2000) for 1 hour at room temperature. For whole-mount tissue
   imaging, approximately 2-mm^3 sections of eWAT depots dissected from
   CMA reporter mice were fixed with buffered zinc formalin (ANATECH, 171)
   at room temperature overnight followed by permeabilization, blocking,
   and antibody stainings as mentioned above. Images were acquired with a
   Leica TCS SP8 confocal microscope as mentioned above. Antibodies used
   for tissue imaging were caveolin1 (1:1000; Cell Signaling Technology,
   3267), CD31 (1:200; Cell Signaling Technology, 77699), Ki67 (1:500;
   Abcam, ab15580), Mac2 (1:400; Thermo Fisher Scientific, 13-5301-82),
   and PreF1 (1:500; Sigma-Aldrich, MABN671).

Reverse transcription polymerase chain reaction

   Total RNA was isolated with RLT buffer (QIAGEN) for cells and with
   TRIzol (Invitrogen, 79306) for tissues (whole eWAT fat pad) and by
   using the RNeasy Plus mini kit (QIAGEN, 74134) according to the
   manufacturer’s instructions. Genomic DNA was excluded using a gDNA
   eliminator spin column. Total RNA (1 μg) was reverse-transcribed into
   cDNA using Superscript III (Invitrogen, 18080051), and quantitative
   reverse transcription polymerase chain reaction (RT-PCR) analyses were
   performed using Power SYBR Green PCR master mix (Applied Biosystems) on
   a StepOne Plus real-time PCR System (Applied Biosystems). Normalization
   of expression was performed with TATA-binding protein (TBP) as
   housekeeping gene. The mRNA expression in control samples was
   represented as 1, and mRNA expression in experimental samples was
   represented as fold change compared to expression in controls. The
   primers used are presented in table S2.

Immunoblot

   Cell or tissue (whole eWAT fat pad) lysates were prepared in
   radioimmunoprecipitation assay buffer [150 mM sodium chloride, 1%
   NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and 50 mM
   tris (pH 8)] containing protease and phosphatase inhibitors. Protein
   concentration was measured by the Lowry method using BSA as a standard,
   and samples were subjected to SDS–polyacrylamide gel electrophoresis
   and immunoblotted after transfer to nitrocellulose membranes using
   standard procedures. Proteins were visualized by using
   peroxidase-conjugated secondary antibodies and chemiluminescent reagent
   (PerkinElmer) in a G-BOX Chemi XX6 (Imgen). Unsaturated images were
   used for densitometry with ImageJ. The percentage of lysosomal
   degradation for specific proteins was calculated by discounting from
   the total amount of protein (given an arbitrary value of 100) the
   percentage of increase in protein levels upon addition of lysosomal
   inhibitors. Antibodies used for immunoblotting were actin (1:10,000;
   Abcam, AC15), aldolase (1:1000; Cell Signaling Technology, 3188),
   enolase 1 (1:1000; Cell Signaling Technology, 3810),
   glyceraldehyde-3-phosphate dehydrogenase (1:1000; Cell Signaling
   Technology, 2118), hexokinase (1:1000; Cell Signaling Technology,
   2024), LAMP1 (1:3000; Developmental Studies Hybridoma Bank, 1D4B),
   LAMP2A (1:5000; Invitrogen, 512200, AMC2), LAMP2B (1:2000; generated in
   our laboratory), LC3 (1:1000; Cell Signaling Technology, 2775), Myc
   (1:1000; Cell Signaling Technology, 5605), PLIN2 (1:2000; PROGEN
   Biotechnik, GP40), SMAD2/3 (1:1000; Cell Signaling Technology, 8685),
   phospho-SMAD2/3 (1:1000; Cell Signaling Technology, 8828), and
   K48-specific ubiquitin (1:2000; Millipore Sigma, 051307, Apu2).

Glycolysis stress test

   Extracellular acidification rates were measured using a 24-well
   Seahorse Bioanalyzer according to the manufacturer’s instructions
   (Agilent Technologies). Cells were grown on gelatin-coated XF24 plate
   and subjected to differentiation in a staggering manner—so that CTR and
   L2A(−) cells at different days of differentiation were on the same
   plate. For glycolysis stress test, cells were shifted to base medium
   supplemented with 2 mM l-glutamine and incubated in a CO[2]-free
   chamber at 37°C for 1 hour. Once in the bioanalyzer, plates were
   sequentially injected with 25 mM glucose, 2 μM oligomycin, and 100 mM
   2-deoxyglucose. Glycolysis, glycolysis capacity, glycolytic reserve,
   and energy phenotype were calculated using Wave software as per the
   manufacturer’s instructions.

Tissue respirometry

   OCR of eWAT were measured using Seahorse Bioanalyzer XF24 (Agilent
   Technologies). Briefly, freshly excised tissue explants (8 to 12 mg)
   were placed in an islet-capture plate in Krebs-Henseleit buffer (111 mM
   sodium chloride, 4.7 mM potassium chloride, 2 mM magnesium sulfate, 1.2
   mM disodium hydrogen phosphate, 0.5 mM carnitine, 2.5 mM glucose, and
   10 mM sodium pyruvate). Basal OCR was measured followed by injection of
   permeabilization agent (digitonin, 0.05 μg/ml). Values were normalized
   to tissue weight.

Flow cytometry

   For cell cycle analysis, centrifuged cells were resuspended in 1 μM
   DRAQ5 in PBS and incubated for 10 min at room temperature before
   analysis using a flow cytometer (BD LSR II analyzer). Histograms were
   deconvoluted by FlowJo software into G[0]-G[1], S, and G[2]-M peaks.
   Flow cytometric analysis for adipocyte precursors was performed as
   described previously ([173]25). Excised whole eWAT fat pad was digested
   in collagenase type 2 (0.8 mg/ml; Worthington Biochemical, LS4174) in
   Hanks’ balanced salt solution (HBSS) containing 3% BSA, 1.2 mM calcium
   chloride, 1.0 mM magnesium chloride, and 0.8 mM zinc chloride for 12
   hours in a 37°C shaking water bath. Centrifugation at 300g for 3 min
   yielded the stromal vascular fraction pellet that was resuspended in 3%
   BSA in HBSS buffer and sequentially filtered through 70-μm (BD
   Biosciences, 352350) and 40-μm (BD Biosciences, 352340) filters before
   staining with the following antibodies for 20 min on ice:
   CD45-allophycocyanin eFluor 780 (1:5000; Thermo Fisher Scientific,
   47-0451-80), CD31 phycoerythrin (1:500; Thermo Fisher Scientific,
   12-0311-81), and CD140a/PDGFRα fluorescein isothiocyanate (1:400;
   Thermo Fisher Scientific, 11-1401-80). Following antibody incubation,
   samples were washed, centrifuged at 300g for 3 min, and analyzed with a
   BD LSR II analyzer. Data analysis was performed using FlowJo software.
   Lin^− cells were defined as CD31^− CD45^−. Adipose progenitor cells are
   defined as Lin^− PDGFRα^+ cells. The gating strategy is shown in fig.
   S3F.

Microarray and analysis

   Total RNA was extracted from cells at days 0 and 2 using an RNeasy mini
   kit (QIAGEN) according to the manufacturer’s protocols and submitted to
   Phalanx Biotech for analysis. Quality check was performed by Agilent
   RNA 6000 Nano Assay, and the array was performed using Agilent Mouse
   Micro Array 8x60k. Array raw data files were processed with the Rosetta
   Resolver System. Gene clustering by average linkage algorithm was
   performed on selected expression gene lists and normalized by z score
   after data transformation and mean centering. Gene set enrichment
   analyses were run using the weighted enrichment. The gene sets were
   analyzed for gene ontology, pathway enrichment, upstream regulators,
   and analysis match using the Ingenuity Pathway Analysis software
   (QIAGEN) and STRING database ([174]https://string-db.org/). KFERQ-like
   motifs in predicted upstream regulator proteins were identified using
   the KFERQ Finder v0.8 database ([175]https://rshine.einsteinmed.edu)
   ([176]56).

Macroautophagy flux assay

   Cells were untreated or treated with lysosomal protease inhibitors,
   ammonium chloride (20 mM), and leupeptin (100 μM) for 12 hours. For ex
   vivo flux, freshly excised tissue explants were incubated with
   high-glucose DMEM in the presence or absence of lysosomal protease
   inhibitors, ammonium chloride (20 mM), and leupeptin (200 μM) at 37°C
   and 5% CO[2] for 2 hours. Afterward, cells or tissue explants were
   homogenized in a lysis buffer containing protease inhibitors and
   processed for immunoblot for LC3. Macroautophagy flux was calculated by
   dividing the densitometry value of LC3 II in lysosomal-inhibitor
   treated by that of untreated samples.

Isolation and differentiation of primary preadipocytes

   Primary preadipocytes were isolated from eWAT depots from WT and L2AKO
   mice using a preadipocyte isolation kit (Abcam, ab196988) according to
   the manufacturer’s instructions. The preadipocytes were cultured in
   DMEM/F12, supplemented with 10% FBS and 1% penicillin/streptomycin, at
   10% CO[2]. To induce adipogenesis, confluent cultures were treated with
   adipogenesis induction cocktail [growth medium supplemented with
   insulin (1.5 μg/ml), 1 μM dexamethasone, 500 μM IBMX, and 1 μM
   rosiglitazone] for 3 days. After 3 days, the cells were maintained in
   growth medium with insulin (1.5 μg/ml). Cells were used at day 9.

Other methods

   For transfection with PLIN2-tGFP WT or CMA-mutant (mutated KFERQ-motif)
   ([177]27), 3T3L1 cells were transfected with 1 μg of plasmid DNA using
   Lipofectamine 2000 (Invitrogen, 11668027). Liver triglyceride content
   was assessed using a commercial kit (BioVision, K622) according to the
   manufacturer’s instructions. Daily food intake was measured in
   individually housed mice for five consecutive days. Residual food after
   24 hours was subtracted from the added food to reveal daily food
   intake. Blood glucose in overnight fasted mice was measured using a
   glucometer (Bayer). Serum insulin in overnight fasted mice was measured
   using enzyme-linked immunosorbent assay (Crystal Chemicals, 90080).
   HOMA-IR was calculated using the standard calculations: HOMA-IR =
   Insulin (μIU/ml) × glucose (mmol/l)/22.5.

Quantification, statistics, and software

   As indicated in the figure legends, all quantitative data are presented
   as the means ± SEM of biologically independent experiments or samples
   (referred to as n in the figure legends). All measurements were taken
   from distinct samples. For data from mice, individual values are also
   shown. Power analysis was used to determine the number of animals
   required for each analysis based on the previous biochemical and
   histological differences that we have found when manipulating CMA in
   the in vitro system or in other in vivo studies. With the calculated
   sample size and a two-sided type 1 error rate of 5%, >80% power was
   predicted to detect effects >1.5 in the parameters analyzed. None of
   the animals were excluded from the study. Animals inside the group
   control and the group LAMP2A KO were randomly assigned to control or
   HFD using the randomization function in Excel. Every experiment was
   performed in at least three independent replicates. Experiments in
   cells in culture were performed on different days to confirm
   reproducibility of the procedures. In addition, markers of adipocyte
   differentiation were used for each of the differentiation experiments
   and percentage of cells differentiated in the control cells were
   monitored to assure success of the differentiation protocol. Data
   collection and analyses were done blinded to the samples. Blinding was
   lifted only for the final statistical analysis. The following software
   were used: ImageJ for imaging and immunoblot analysis, Ingenuity
   Pathway Analysis (QIAGEN) and STRING database
   ([178]https://string-db.org/) for cellular pathway enrichment,
   prediction of upstream regulators and analysis match, and FlowJo for
   flow cytometry analysis. Graphical plots were made using GraphPad Prism
   9.0, R, and FlowJo v10.6.1, and the images were assembled using ImageJ
   v1.53f. Statistical analyses were performed using GraphPad Prism 9.0.
   Before statistical testing, normality was assessed using the
   Shapiro-Wilk test. Outliers were determined by the ROUT method (Q =
   1%). Statistical significance was compared by two-tailed unpaired
   Student’s t test for two groups, one-way analysis of variance (ANOVA)
   followed by Dunnett’s multiple comparisons test, or two-way ANOVA
   followed by Šídák’s or Tukey’s multiple comparisons test for multiple
   groups with *P < 0.05, **P < 0.01, ***P < 0.001, and ns
   (nonsignificant) representing P > 0.05.

Acknowledgments