Abstract

Objective

   The capacity of mature adipocytes to de-differentiate into
   fibroblast-like cells has been demonstrated in vitro and a few, rather
   specific in vivo conditions. A detailed comparison between
   de-differentiated fat (DFAT) cells and adipose stem and progenitor
   cells (ASPCs) from different adipose depots is yet to be conducted.
   Moreover, whether de-differentiation of mature adipocytes from
   classical subcutaneous and visceral depots occurs under physiological
   conditions remains unknown.

Methods

   Here, we used in vitro "ceiling culture", single cell/nucleus RNA
   sequencing, epigenetic anaysis and genetic lineage tracing to address
   these unknowns.

Results

   We show that in vitro-derived DFAT cells have lower adipogenic
   potential and distinct cellular composition compared to ASPCs. In
   addition, DFAT cells derived from adipocytes of inguinal origin have
   dramatically higher adipogenic potential than DFAT cells of the
   epididymal origin, due in part to enhanced NF-KB signaling in the
   former. We also show that high-fat diet (HFD) feeding enhances DFAT
   cell colony formation and re-differentiation into adipocytes, while
   switching from HFD to chow diet (CD) only reverses their
   re-differentiation. Moreover, HFD deposits epigenetic changes in DFAT
   cells and ASPCs that are not reversed after returning to CD. Finally,
   combining genetic lineage tracing and single cell/nucleus RNA
   sequencing, we demonstrate the existence of DFAT cells in inguinal and
   epididymal adipose depots in vivo, with transcriptomes resembling
   late-stage ASPCs.

Conclusions

   These data uncover the cell type- and depot-specific properties of DFAT
   cells, as well as their plasticity in response to dietary intervention.
   This knowledge may shed light on their role in life style
   change-induced weight loss and regain.

   Keywords: Adipocyte de-differentiation, DFAT cell re-differentiation,
   NF-κB, Diet intervention, scRNA-seq/snRNA-seq, DNA methylation

Highlights

     * •
       In vitro-derived DFAT cells show lower adipogenic potential and
       distinct transcriptional profiles compared to ASPCs.
     * •
       Inguinal DFAT cells show higher adipogenic potential than those of
       epididymal origin, which requires NF-κB signaling.
     * •
       HFD enhances DFAT cell colony formation and re-differentiation into
       adipocytes, switching to CD only reverses the latter.
     * •
       HFD feeding induces lasting epigenetic changes in DFAT cells and
       ASPCs that are not reversed by switching to CD.
     * •
       DFAT cells are present in inguinal and epididymal adipose depots
       in vivo and resemble late-stage ASPCs.

1. Introduction

   White adipose tissue (WAT) serves as a major storage site for excess
   energy in the form of triglyceride and acts as an endocrine organ that
   secretes various hormones and factors to maintain energy homeostasis.
   In addition to the white adipocytes that contain triglycerides, WAT
   also comprises various other cell types in the stromal vascular
   fraction (SVF), including adipose stem and progenitor cells (ASPCs),
   immune cells, vascular endothelial cells and small proportions of other
   cell types [[29][1], [30][2], [31][3]]. Obesity develops when calorie
   intake regularly exceeds the body's needs, during which WAT expands in
   an unhealthy way characterized by excessive hypertrophy of existing
   adipocytes, impaired ASPC differentiation and accumulation of
   proinflammatory macrophages [[32]4,[33]5]. The overloaded adipocytes
   are under mechanical stress, become hypoxic and eventually undergo
   apoptosis, leading to influx of macrophages into the tissue that
   activates inflammation and efflux of lipids into the neighboring
   non-adipose organs that causes local and systemic insulin resistance
   [[34]6]. On the other hand, weight loss via calorie restriction
   promotes adipocyte size reduction by activating lipolysis and
   increasing tissue sensitivity to neural and hormonal stimulation
   [[35]7].

   Mature adipocytes have long been considered as a terminally
   differentiated cell type. Development of the “ceiling culture” method
   by Sugihara et al. in the late 1980s demonstrated the ability of mature
   adipocytes to de-differentiate into fibroblast-like cells in vitro
   [[36]8,[37]9]. During de-differentiation, adipocytes undergo
   morphological changes from round-shaped cells loaded with lipids to
   spindle-shaped fibroblast-like cells without apparent lipid droplets,
   accompanied by increased expression of preadipocyte marker genes
   [[38]10]. The resulting de-differentiated fat (DFAT) cells possess stem
   cell-like properties and can give rise to multiple cell lineages,
   including cardiomyocyte, osteoblast and chondrocyte [[39]11,[40]12].
   During pregnancy and lactation, adipocytes in mammary gland have been
   shown to de-differentiate into preadipocyte- and fibroblast-like cells,
   and subsequently proliferate and re-differentiate into adipocytes after
   weaning [[41]13]. In addition, reversible de-differentiation of dermal
   adipocytes is required for hair cycling and skin repair
   [[42]14,[43]15]. Moreover, adipocyte de-differentiation has been shown
   to contribute to liposarcoma [[44]16] and mammary tumor progression
   [[45]17]. So far, the extent to which DFAT cells resemble ASPCs at the
   functional and transcriptional levels, as well as their plasticity in
   response to dietary interventions remain to be elucidated.

   The family of NF-κB transcription factors has been shown to play
   important roles in adipose tissue homeostasis and plasticity. This
   group of proteins encompasses transcriptional activators, such as RelA
   (p65), RelB and c-Rel, as well as repressors, including NF-κB1 (p50)
   and NF-κB2 (p52) [[46]18]. Inactive heterodimers of RelA and NF-κB1 are
   sequestered in the cytosol by IκBα [[47]19]; upon stimulation by
   cytokines and growth factors, IKKβ kinase becomes activated and
   phosphorylates IκBα, resulting in the release and nuclear translocation
   of RelA and NF-κB1. Adipocyte-specific deletion of IKKβ has been shown
   to exacerbate tissue inflammation induced by HFD [[48]20] and adipocyte
   cell death, as well as reduce adaptive adipose tissue remodeling and
   elevate lipolysis in visceral adipose depots [[49]21]. Protein levels
   of several NF-κB subunits, including RelA, RelB and NF-κB2 have been
   shown to increase during adipogenesis of 3T3-L1 cells [[50]22].

   In the present study, we used cell culture models, single cell/nucleus
   RNA-sequencing (sc/snRNA-seq) methods, epigenetic analysis and genetic
   lineage tracing to characterize DFAT cells derived from subcutaneous
   and visceral adipose depots under different dietary interventions and
   uncovered the role of NF-κB signaling in regulating the
   re-differentiation of DFAT cells from inguinal WAT (iWAT).

2. Results

2.1. DFAT cells derived from subcutaneous and visceral fat depots have lower
adipogenic potential and distinct transcriptional profiles compared to ASPCs

   Using a previously described in vitro “ceiling culture” method
   [[51]8,[52]9], we obtained DFAT cells from adipocytes isolated from
   iWAT and epididymal WAT (eWAT) of chow diet (CD)-fed C57BL/6J male
   mice. Appearance of DFAT cells was characterized by the morphological
   transition from round to spindle shape and gradual loss of Oil Red O
   staining ([53]Figure S1A), as well as elevated expression of ASPC
   marker genes (Pdgfra and Pdgfrb) and reduced expression of adipocyte
   marker genes (AdipoQ and Lep) ([54]Figure S1B). In addition, we crossed
   the AdipoQ^CreERT2 mouse strain that expresses a tamoxifen-inducible
   Cre recombinase under the control of the AdipoQ gene [[55]23] with the
   Ai14 strain that expresses tdTomato under the control of the Rosa26
   locus in a Cre-dependent manner [[56]24] for genetic labelling of
   adipocytes. Six-week-old AdipoQ^CreERT2;Ai14 mice were injected with
   three daily consecutive shots of tamoxifen and eWAT adipocytes were
   collected one day after the last injection for in vitro
   de-differentiation. No overlap between tdTomato and PDGFRA
   immunostaining that labels ASPCs was observed in whole-mount eWAT
   tissue sections, confirming the adipocyte-specific labelling of
   AdipoQ^CreERT2 ([57]Figure S1C). tdTomato^+ adipocytes were found in
   the isolated adipocyte suspension ([58]Figure S1D), which allows us to
   specifically follow their in vitro de-differentiation process. By
   monitoring the round-shaped tdTomato^+ adipocytes in the same visual
   field on a daily basis, we found that most of the spindle-shaped DFAT
   cells retained the tdTomato signal ([59]Figure S1E), indicating that
   these cells are actually derived from the adipocytes. To further
   exclude a potential contamination of ASPCs, we also derived DFAT cells
   from adipocytes of Pdgfra^CreERT2;Ai14 mice whose ASPCs were labelled
   right before collection by tamoxifen injection. We did not find any
   tdTomato^+ cells in such cultures from two independent experiments
   (data not shown), thereby excluding a potential contamination of ASPCs
   in the isolated DFAT cells.

   To compare the adipogenic potential of DFAT cells with ASPCs, we
   extracted adipocytes and SVF cells in parallel from the same adipose
   depots (iWAT and eWAT) of CD-fed C57BL/6J mice for de-differentiation
   and ASPC expansion, respectively, followed by (re)-differentiation into
   adipocytes ([60]Figure 1A). Six days after induction of in vitro
   differentiation, Oil Red O staining revealed that DFAT cells have a
   lower adipogenic potential than ASPCs from their cognate fat depot and,
   among these, cells derived from eWAT always showed lower
   differentiation potential than those from iWAT (i.e.
   eWAT_DFAT < eWAT_ASPC < iWAT_DFAT < iWAT_ASPC) ([61]Figure 1B). In
   agreement with this, Lep mRNA levels followed the same trend
   ([62]Figure 1C). These results suggest that differences in adipogenic
   potential exist at both the cell type and the depot levels.

Figure 1.

   [63]Figure 1
   [64]Open in a new tab

   DFAT cells and ASPCs are different in terms of adipogenic potential and
   transcriptional profiles. (A) Schematic illustration of the
   experimental design for DFAT cell and ASPC in vitro culture. (B) Bright
   field images and quantification of Oil Red O staining of DFAT and SVF
   cells on DIV 8. n = 4 independent experiments, represented by a dot in
   the graph. (C) qPCR analysis of Lep mRNA level in cells that had been
   induced for adipocyte differentiation for 6 days. n = 4 independent
   experiments, represented by a dot in the graphs. Statistical
   significance was assessed by ordinary one-way ANOVA Tukey's multiple
   comparisons test in (B–C). ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns,
   not significant. Values were normalized to the iWAT_DFAT group and were
   presented as mean ± standard deviation (SD) in (B–C). (D) Uniform
   Manifold Approximation and Projection (UMAP) of all cell types in
   cultured DFAT cells and ASPCs obtained from iWAT and eWAT of C57BL/6J
   mice fed with CD. (E) Feature plots showing the expression patterns of
   ASPC and adipocyte marker genes in the ASPC population of the four
   datasets. Z-score was used for presentation. (F) UMAPs and cellular
   composition of ASPC subpopulations in the four datasets. (G) Heatmap
   showing the average expression of DEGs in the three ASPC
   subpopulations. Z-score was used for presentation. (H) Pathway
   enrichment analysis of the DEGs of the three ASPC subpopulations using
   the Gene Ontology-Biological Process gene set. (For interpretation of
   the references to color in this figure legend, the reader is referred