A subset of dermal fibroblasts undergo rapid differentiation into adipocytes in response to infection and acutely produce the cathelicidin antimicrobial peptide gene Camp. Vitamin A and other retinoids inhibit adipogenesis yet can show benefit to skin disorders, such as cystic acne, that are exacerbated by bacteria. We observed that retinoids potently increase and sustain the expression of Camp in preadipocytes undergoing adipogenesis despite inhibition of markers of adipogenesis, such as Adipoq, Fabp4, and Rstn. Retinoids increase cathelicidin in both mouse and human preadipocytes, but this enhancement of antimicrobial peptide expression did not occur in keratinocytes or a sebocyte cell line. Preadipocytes undergoing adipogenesis more effectively inhibited growth of Staphylococcus aureus when exposed to retinoic acid. Whole transcriptome analysis identified hypoxia-inducible factor 1-α (HIF-1α) as a mechanism through which retinoids mediate this response. These observations uncouple the lipid accumulation element of adipogenesis from the innate immune response and uncover a mechanism, to our knowledge previously unsuspected, that may explain therapeutic benefits of retinoids in some skin disorders.
Retinoids comprise a large group of chemical compounds that are vitamin A (retinol [ROL]) analogs and/or derivatives and can influence multiple physiological processes, including vision, cell proliferation, differentiation, and immunity (1, 2). Within the skin, retinoids have been shown to affect multiple cell types, modulating several processes, including wound healing, pigmentation, and immune defense (3–9), and have been used pharmacologically to treat chronological and photoaging, psoriasis, and acne (9–13). There are few cases in which retinoids have exhibited direct antimicrobial activity (14, 15); however, under most circumstances, immune modulation by retinoids is thought to only occur indirectly (5–7, 9, 16–18). For example, the mechanism by which retinoids are therapeutically useful in acne, a disorder that is driven in part by the bacterium Cutibacterium acnes, is thought to occur through their effect to decrease sebum production and alter epidermal function rather than through an influence on immune defense against bacterial growth (9).
It has been recently observed that dermal fibroblasts of the preadipocyte lineage will undergo rapid, local proliferation and differentiation in response to bacterial infection of the skin (19, 20). This process, dubbed reactive adipogenesis, results in an increase in the expression of the cathelicidin gene CAMP, an antimicrobial peptide that peaks during this process as the local adipocytes mature (20). Reactive adipogenesis is essential for optimal innate immune defense because impairment of the adipogenic process decreases total skin cathelicidin expression and increases the severity of bacterial infections (19–21).
Retinoids have been shown to inhibit adipogenesis and promote lipid breakdown (22–27) and therefore would be predicted to decrease cutaneous antimicrobial defense capacity. In this study, we sought to understand the effects of retinoids on antimicrobial peptide production during adipogenesis to better understand the therapeutic mechanism of action of retinoids. Despite exhibiting the anticipated antiadipogenic effects, we unexpectedly observed that retinoids enhanced and sustained cathelicidin levels in developing preadipocytes, ultimately translating to increased antimicrobial activity. This process was found to depend in part on hypoxia-inducible factor 1-α (HIF1α), a broadly active transcription factor within the CAMP gene promoter. These observations reveal an uncoupling between adipogenesis and antimicrobial expression that could be exploited as a novel means of modulating innate immune defense.
Materials and Methods
Abs, chemicals, and reagents
Vitamin A (ROL), all trans-retinal (RAL), all trans-retinoic acid (RA), 9-cis RA, 13-cis RA, 1α,25-dihydroxyvitamin D3, l-mimosine, tazarotene, 3-isobutyl-1-methylxanthine (IBMX), dexamethasone, indomethacin, and recombinant human insulin were purchased from Sigma-Aldrich (St. Louis, MO). Acriflavine hydrochloride was purchased from Thermo-Fisher Scientific (Pittsburgh, PA). AM580 was purchased from Abcam. MALP-2 was purchased from Enzo Life Sciences. Rabbit anti-CRAMP Abs were made in our laboratory, as previously described (28). Mouse anti-FABP4 was purchased from Santa Cruz Biotechnology (Santa Cruz, CA).
3T3-L1 primary mouse preadipocytes were purchased from the American Type Culture Collection (CL-173). Neonatal primary dermal fibroblasts were isolated by our laboratory, as previously described (19). Cells were grown in DMEM supplemented with 10% FBS, glutamax, and antibiotic-antimycotic (Thermo-Fisher Scientific) in a humidified incubator at 5% CO2 and 37°C under sterile conditions. To induce differentiation, 2-d postconfluent 3T3-L1 cells were switched to adipocyte differentiation medium containing 2 μM dexamethasone, 250 μM IBMX, 200 μM indomethacin, and 10 μg/ml recombinant human insulin. For time course experiments, differentiation media were refreshed at indicated times. Immortalized nondiabetic human preadipocytes (29) were a generous gift from Dr. P. M. Schliever (Carver College of Medicine, Iowa City, IA). Human preadipocytes were grown and differentiated, as previously described (29), and differentiation media were refreshed on indicated days. Human epidermal keratinocytes (no. C-001-5C; Life Technologies) were cultured in EpiLife medium supplemented with 60 μM CaCl2, EpiLife Defined Growth Supplement (Life Technologies), and antibiotic/antimycotic. SEB-1 sebocytes were cultured in Sebomed basal medium (Millipore) supplemented with recombinant human epidermal growth factor (5 ng/ml) (Sigma-Aldrich). To examine the effects of retinoids on cathelicidin expression, retinoids (1 μM, unless otherwise noted) or vehicle (ethanol) were added to differentiation media or cell culture media on day 0 (unless indicated otherwise). Retinoid-containing media were subsequentially refreshed every 2–3 d throughout the course of the experiment. Treatment of cells with 1α,25-dihydroxyvitamin D3 (vitamin D) (10 nM) or vehicle (ethanol) was conducted identical to that of retinoids. To examine the role of HIF-1α, cells were cotreated with l-mimosine (400 μM), acriflavine (5 μM), or vehicle (water) for 24 h.
RNA isolation, cDNA synthesis, and real-time quantitative PCR analysis
Cultured cells were lysed with TRIzol reagent (Life Technologies) or PureLink Lysis Buffer (Ambion/Life Technologies), and RNA was isolated using the PureLink isolation kit. Up to 1 μg of RNA was reverse transcribed to cDNA using the iScript cDNA synthesis kit (Bio-Rad). Quantitative real-time PCR was performed with the CFX96 Real-Time System (Bio-Rad) using SYBR Green Mix (Biomake, Houston, TX). The housekeeping gene TAT-binding box protein (Tbp) was used to normalize gene expression in samples. Specific primer sequences are shown in Supplemental Table I.
Supernatants from treated cells were enriched for protein using Oasis HLB Plus Light Cartridges (Waters, Milford, MA) and lyophilized. Lyophilized protein from 500 μl of supernatant was separated on a Novex 10–20% Tricine Gel (Invitrogen) and transferred to a PVDF membrane using Trans-Blot Turbo Transfer Pack (Bio-Rad), according to the manufacturer’s instructions. Membranes were blocked with Odyssey blocking buffer (LI-COR) and incubated overnight at 4°C with primary Abs. Immunoblots were washed and incubated with fluorescent secondary Abs (LI-COR) for 1 h at room temperature and subsequentially imaged using the Odyssey Infrared Imager (LI-COR).
For immunofluorescence analysis, 3T3-L1 cells were grown, differentiated, and treated in eight-well chamber slides, as described earlier. Cells were fixed in 4% PFA for 10 min and subsequentially incubated with primary Abs overnight at 4°C. Cells were washed and incubated for 1 h with secondary Abs conjugated to Alexa Fluor 594 (Invitrogen), and nuclei were counterstained with DAPI. All images were taken with an Olympus BX41 microscope.
Histology and immunohistochemistry
Tissue biopsies were directly embedded in OCT compound. Frozen sections were fixed in 4% PFA for 10 min prior to H&E staining or immunofluorescence staining. For immunohistochemistry, fixed and permeabilized frozen tissue sections were stained with BODIPY solution to visualize lipid droplets, and nuclei were counterstained with DAPI. All images were taken with an Olympus BX41 microscope.
In vitro bacterial killing assay
Antibiotic-free conditioned media were collected from differentiating dermal fibroblasts, as previously described (20). Conditioned media (200 μl) were mixed with 105 CFU of Staphylococcus aureus stain USA300 in deep-well 96-well plates and incubated at 37° with shaking at 275 rpm for 12 h prior to plating for CFU counting.
All animal experiments were approved by the University of California, San Diego (UCSD), Institutional Animal Care and Use Committee. SKH-1 hairless mice were originally purchased from Charles River Laboratories and bred and maintained in the animal facility of UCSD. Animals in all experimental models were age and sex matched. Skin infections were conducted as previously described (20), using S. aureus strain USA300 (methicillin-resistant S. aureus) (S. aureus strain AH4807 was used for IVIS experiments) or PBS control. Mice were distally administered one s.c. injection with RA (10 mg/kg) or vehicle (DMSO) dissolved in olive oil on the day of infection and two additional injections on subsequent days. Mice were sacrificed on day 3, and a skin biopsy of the infected region was harvested. Skin biopsies were homogenized in PureLink Lysis Buffer (or PBS for CFU counting) with 2 mM zirconia beads in a minibead beater (Biospect, Bartleville, OK) for RNA extraction. To quantify CFU, homogenized skin samples were serially diluted and plated onto Baird-Parker agar and grown for 24 h to quantify the CFU per gram of tissue. For in vivo live bacterial imaging, mice were imaged under isoflurane inhalation anesthesia (2%). Photons emitted from luminescent bacteria were collected during a 1-min exposure, using the Xenogen IVIS Imaging System and living image software (Xenogen, Alameda, CA). Bioluminescent image data are presented on a pseudocolor scale overlaid onto a grayscale photographic image. Using the image analysis tools in living image software, circular analysis windows (of uniform area) were overlaid onto regions of interest, and the corresponding bioluminescence values (total flux) were measured.
Experiments conducted were performed in triplicate with at least three technical replicates. Statistical significance was calculated using a Student two-tailed t test in which *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001, as indicated in the figure legends.
Retinoids increase cathelicidin expression during adipogenesis
To determine the effects of retinoids on Camp expression during reactive adipogenesis, we first examined differentiation of preadipocytes in vitro under defined conditions. Differentiation was associated with increased Camp mRNA expression, and treatment with the TLR2/6 ligand, MALP-2, further stimulated Camp expression in both undifferentiated and differentiating 3T3-L1 preadipocytes (Fig. 1A). The addition of vitamin A (ROL) greatly amplified the Camp response to differentiation, resulting in an increase in steady-state Camp mRNA levels compared with untreated 3T3-L1 cells, with the greatest increase occurring with Malp-2 and ROL (Fig. 1A).
Vitamin A (also known as ROL) is an inactive precursor that can be enzymatically converted to RA or RAL to exhibit physiological activity. Addition of ROL, RAL or RA had similar effects on Camp induction when added to 3T3-L1 cultures at the start of an 8-d differentiation experiment. RA appeared most potent when added between day 2 and 8 after differentiation was initiated, and induction was least potent when added later during adipogenesis at day 5–8 (Fig. 1B). These findings suggest adipocyte precursors can metabolize vitamin A to active forms.
RA can isomerize between all-trans RA, 9-cis RA, and 13-cis RA (32). Similar activity was observed by each of these isoforms on Camp expression (Fig. 1C). Using RA analogs specific for the retinoid acid receptors (RAR), agonists for both RARα and RARβ/γ potently stimulated Camp expression to an extent similar to RA (Supplemental Fig. 1A). RA demonstrated a dose-dependent induction of Camp at 24 h (Fig. 1D) but significantly induced Camp expression in as little as 6 h and resulted in sustained high levels of expression at day 8, a prolonged increase that contrasted with the normal downregulation of Camp seen during late adipogenesis in the absence of a retinoid (Fig. 1E). Previous observations have shown that Camp is the only AMP induced during adipogenesis, whereas expression of α- and β-defensins was not altered (20). Our current studies are consistent with these observations, as addition of RA was unable to induce expression of other AMPs (Supplemental Fig. 1B). Contrary to the increase in antimicrobial peptide gene expression that was promoted by retinoids but consistent with prior observations of the effect of RA to inhibit adipogenesis (27, 33), we observed that RA inhibited expression of adipocyte maturation markers Adipoq, Fab4, and Rstn during the time course of this experiment (Fig. 1F, 1G). These observations suggest that RA promoted an uncoupling between antimicrobial peptide expression and classic markers of adipocyte differentiation that are associated with lipid accumulation.
In agreement with our transcriptional studies, significantly elevated levels of the murine cathelicidin protein (Cramp) precursor protein were detected in RA-treated 3T3-L1 cell supernatants (Fig. 2A). Cramp immunostaining complimented these findings, as more Cramp staining was seen in the cytoplasm of 3T3-L1 cells during differentiation into adipocytes in the presence of RA (Fig. 2B). On day 12, numerous round areas in the cytosol that were unstained by DAPI or for Cramp are maturing lipid droplets.
Human and mouse cathelicidin genes exhibit several regulatory parallels but also have important differences. The human CAMP promoter possesses an active vitamin D response element that is lacking in mice (34, 35), and vitamin D can upregulate CAMP expression in several human cell types, including human keratinocytes (34, 36). To determine if the observations made in murine-derived 3T3-L1 cells could be found in human preadipocytes, we investigated the effects of RA and vitamin D in a cultured human preadipocyte line (19, 20). Both 9-cis RA and vitamin D increased CAMP expression in human preadipocytes and exhibited an additive effect when cells were treated with both together (Fig. 2C). However, RA was unable to stimulate CAMP expression under similar culture conditions in a human sebocyte cell line or in primary cultures of normal human keratinocytes (Fig. 2D). Thus, we conclude that retinoids can increase expression of cathelicidin in both human and mouse preadipocytes, but the effects of RA on this response are specific to cell type.
Given the significant cathelicidin upregulation observed with retinoid treatment, we next wished to determine whether these results correlate with antimicrobial activity. As our observations have shown that mouse neonatal dermal fibroblasts (MSFBs) express significantly more Cramp protein compared with 3T3-L1 cells (Supplemental Fig. 2A), we selected to use MSFBs to determine antimicrobial activity. MSFB cells were stimulated to initiate adipogenesis in culture within the presence of RA, and this culture supernatant was then added to bacterial cultures of S. aureus (USA300). The growth of S. aureus was maximally inhibited from culture supernatants of cells treated with RA (Fig. 3A). Despite promoting Camp expression and enhancing antimicrobial activity, RA’s effects on cytokine and chemokine expression were comparatively minor under the conditions we examined (Supplemental Fig. 2B).
Given our in vitro findings, we next wished to determine how RA might contribute to antimicrobial activity during S. aureus infection. Systemic treatment of mice with 10 mg/kg RA for 3 d decreased dermal adipose in uninfected skin as well as in skin after challenge with S. aureus (Fig. 3B, 3C). Camp mRNA expression was induced by S. aureus and RA (Fig. 3D). However, no consistent change in the susceptibility to S. aureus survival was seen in mice treated with RA by manual CFU counting of live bacteria (Fig. 3E) or as detected by IVIS analysis of a reporter strain of S. aureus (Fig. 3F).
Transcriptomic response of preadipocytes to RA identifies HIF1α as mediator of antimicrobial response
To gain further insight into the mechanism through which preadipocytes increase Camp in response to RA, we performed RNA sequencing (RNA-seq) of 3T3-L1 cells. The expression of genes that were changed by more than 2-fold when 3T3-L1 cells were differentiated in the presence of RA were identified (Fig. 4A, 4B). An increase in genes associated with adipogenesis (Zfp423, Fabp4, Adipoq, Cebpa) was evident with differentiation, but several genes involved in hypoxic responses were also observed to be upregulated by RA (Supplemental Fig. 3A), particularly HIF1α (Fig. 4C). Quantitative PCR (qPCR) validated that Hif1a expression was increased in both mouse and human preadipocytes treated with RA (Fig. 4F, Supplemental Fig. 3B). Hif1a mRNA expression was also increased in skin samples from mice treated with RA and infected with S. aureus (Fig. 4D).
To ascertain the role of HIF1α in the response of preadipocytes to RA, we used an established HIF1α agonist (l-mimosine) and inhibitor (acriflavine). Camp expression was induced in preadipocytes treated with the HIF1α agonist and further enhanced by RA, and the HIF1α antagonist blocked Camp expression (Fig. 4E, 4F). Vascular endothelial growth factor A (VEGFA), a regulatory target of HIF1α, responded as expected with an increase in expression after agonist exposure and a decrease after antagonist treatment, thus serving as a control for the response to these chemical modifiers of HIF1α activity (Supplemental Fig. 3C). Given our observations that MALP-2 can stimulate Camp expression both in undifferentiated and differentiating 3T3-L1 cells, we validated the role of HIF1α in Camp regulation by cotreating cells with both acriflavine and MALP-2 under both conditions. Similarly, HIF1α inhibition effectively blocked MALP-2 induction of Camp expression and further evidenced the regulatory role of HIF1α (Supplemental Fig. 3D, 3E).
Retinoids are an effective therapy in the treatment of acne and other disorders (9–13). However, the clinical response to retinoids is dependent on mode of delivery, and the mechanisms responsible for the physiological effects of retinoids are incompletely understood. Retinoids indeed have been shown to exhibit both anti-inflammatory and proinflammatory effects on cellular immune responses of the skin (9, 37–39), but the activity of retinoids on innate host antimicrobial defense has not been well characterized. In this study, we present an unexpected observation that ROL and its metabolites potently enhance cathelicidin AMP expression during reactive adipogenesis. These findings suggest an alternative and previously underappreciated mechanism for the action of retinoids in the skin.
Retinoids inhibit adipogenesis and promote lipid breakdown (22–27). Inhibition of adipogenesis by inhibition of the Pparγ system will also inhibit the antimicrobial activity of dermal preadipocytes (19, 20). However, and contrary to expectations, we show in the current study that retinoids increase Camp expression despite exhibiting antiadipogenic activity. This observation demonstrates an uncoupling between the processes of adipogenesis and AMP expression.
Previous observations have shown that adipocytes produce the cathelicidin precursor protein that is upregulated during adipogenesis, which is processed into an active AMP (CRAMP); however, the enzyme that catalyzes this reaction in adipocytes has not been identified (19, 20). Our studies have shown that an increase in Camp gene expression then results in a significant upregulation of the CRAMP preprotein, yet the increase in antimicrobial activity was modest in comparison with the change in protein abundance. Whether this was indicative of a saturation of the system or modulation of CRAMP processing from inactive precursor protein to active peptide is unclear. We hypothesize that the contribution of other cells in vivo during the complex response of reactive adipogenesis may contribute to the maturation of the cathelicidin precursor that is enhanced by retinoids in preadipocytes.
Excess vitamin A results in epidermal hyperplasia and a reduction in white adipose tissue (WAT), whereas infection results in an expansion of dermal WAT (20, 27, 40). Interestingly, RA treatment attenuated the dermal WAT expansion associated with S. aureus infection, yet Camp was still significantly upregulated in comparison with the individual stimuli. RA improved the capacity of differentiating preadipocytes to inhibit S. aureus. However, despite these observations that RA was active in vivo and induced antimicrobial activity in vitro, there was no clear increase in the antimicrobial activity in vivo of mouse skin against S. aureus. This suggests that the immunomodulatory role of RA in the skin is complex and may be dependent on dose, timing, or target organism. Additional work is needed to define the physiological significance of retinoid exposure on innate immune defense.
To gain further insight and understand the complexity of RA-mediated regulation, we examined the transcriptomic response to RA in the context of differentiating preadipocytes. RNA-seq analysis identified HIF1α as a potential regulator of Camp expression. The human CAMP promoter possesses an HIF1α response element and has been implicated in cutaneous immune defense as well as CRAMP expression in the colon (41). The Hif1a promoter possesses an RA response element (42), and RA has been reported to enhance HIF1α expression under normoxic conditions (42–44). These past observations are in agreement with our RNA-seq and qPCR data in which we found that RA induced elevated Hif1a. Similarly, activation of HIF1α by the iron chelator l-mimosine mimicked the effects of RA on Camp expression. The combinatorial effect of both activation and increased Hif1a expression exhibited a significant additive effect, thus providing further evidence that HIF1α and RA work in concert to promote Camp expression. Interestingly, HIF1α inhibition abolished Camp expression under all conditions. This may suggest Camp transcription is entirely dependent on HIF1α or that the global effects of HIF1α further attenuate other mechanisms involved in the expression of Camp. Support for this comes from observations that HIF1α inhibition also downregulates Cebpb transcription and inhibits initiation of adipogenesis (45). Inhibition by acriflavine similarly downregulated Cebpb expression, whereas l-mimosine downregulated Cebpa, the latter of which has also been proposed to regulate CAMP expression. Taken together, this could suggest that C/EBPβ indirectly regulates Camp expression; however, our data do not define if either C/EBPβ or C/EBPα directly regulates this process.
Induction of adipogenesis is mediated by C/EBPβ, which in turn mediates the expression of the preadipogenic factors C/EBPα and PPARγ (46, 47). Although RA does not inhibit the expression of C/EBPβ, it has been shown to inhibit its function (22). Our observations have shown that initiation of differentiation, yet not final maturation of adipocytes, is sufficient and necessary to enhance Camp expression by retinoids. However, there appears to be a limited timeframe during which this effect can be observed, as the potency of retinoid stimulation was markedly reduced when treatment was initiated at later timepoints during differentiation. This suggests that access to the Camp promoter is limited during adipogenesis; such a limited access period may partially explain why acute, high-dose RA administration to mice did not improve resistance to S. aureus infection. Under these conditions, initiation of adipogenesis is also inhibited and may counteract the beneficial host defense function of RA that we have observed in vitro. Analysis of low-dose vitamin A exposure, as well as correlation of responses in human subjects on RA therapy, is needed.
In summary, these observations have revealed a novel disassociation between adipogenesis and the process through which fibroblasts in the dermis that become adipocytes can produce cathelicidin. Reactive adipogenesis is a recently discovered and important physiological response to infection of skin and gut (19–21, 48). These findings provide further insight into how Camp is regulated in adipocytes, a mechanism with important therapeutic implications.
We thank members of the Gallo Laboratory for helpful advice pertaining to experimental procedures and design. We also thank Dr. Patrick M. Schliever (Carver College of Medicine, Iowa City, IA) for the immortalized human adipocyte cell line.
This work was supported by National Institutes of Health Grant R01AR069653.
The online version of this article contains supplemental material.
R.L.G. is a co-founder, scientific advisor, consultant and has equity in MatriSys Biosciences and is a consultant, receives income, and has equity in Sente. The other authors have no financial conflicts of interest.