Reagent and resource sharing

Reagents created for this study, including the Hsd3b6^Δskin mice and bioluminescent strains of S. aureus, are available upon request from the corresponding author and may require a completed materials transfer agreement if there is potential for commercial application. Requests for resources should be directed to and will be fulfilled by the corresponding author.

Experimental model and participant details

Participant recruitment

This study was approved by the University of Texas (UT) Southwestern Institutional Review Board (STU 2019-0145), and written informed consent was obtained from study participants. Recruitment occurred from January 2024 to June 2024, and the human samples were collected between June 2024 and August 2024. Inclusion criteria included participants aged 18‒40 years, a body mass index between 20 kg m⁻² and 35 kg m⁻², and willingness and ability to comply with the requirements of the protocol. Exclusion criteria included participants outside the ages of 18‒30 years, the use of antibiotics in the last 6 months, the use of topical medications, chronic skin disorders or other chronic medical conditions, the use of lipid or hormone-altering medications (except for oral contraceptives), the use of immunomodulator medications, women with irregular menstrual periods, or participants with a history of surgery to endocrine organs. Basic demographic and medical information was collected from each participant. Two healthy control participants, a 29-year-old male and a 26-year-old female, were selected and informed consent was obtained. Study participants were compensated in accordance with the approved study protocol.

Human skin sampling

Skin excretions were collected via Sebutape after completion of skin preparation to decrease external variables⁵. At the time of sampling, the forehead was cleaned with alcohol and four Sebutapes from Clinical and Derm LLC were applied for 15 minutes. Participants were sampled daily for 6 days between the hours of 12 pm and 3 pm. Following collection, tapes were stored in glass vials at −20 °C until processing. Hormones were extracted into 3 ml of LC–MS-grade methanol, and quantified via LC–MS/MS using a SCIEX QTRAP 6500+ (ref. ⁵). Outliers within the four biological replicate samples were identified using a dual-sided Grubbs’ test, also called the extreme studentized deviate method, and were removed. The remaining biological replicate values for each day were averaged. These averaged values were translated from picogram per tape to nanomoles per litre (nM) using the total volume of the tape as provided in US patent US4532937A. Using GraphPad Prism v.10.4.1 (GraphPad Software), a linear regression model was created and plotted for each participant, along with the 95% confidence interval of each regression. An ordinary two-way analysis of variance (ANOVA) was performed to determine statistical significance of time and sex.

Human peripheral blood

Human peripheral blood samples were obtained from fully consented and institutional-review-board-approved donors through BioIVT, Innovative Research and IQ Biosciences.

Mice

Conventionally raised C57BL/6 male and female mice aged 6–9 weeks old were purchased from the Jackson Laboratory. C57BL/6 wild-type, K14-Cre^+/− and Hsd3b6^fl/fl (Extended Data Fig. 2a) mice were bred and maintained in the specific pathogen-free barrier facility at the UT Southwestern Medical Center at Dallas⁵⁹. The generation of Hsd3b6^Δskin (K14-Cre^+/−; Hsd3b6^fl/fl) is described below. Mice were co-housed with three to five mice per cage in all experiments. All mice were housed under a 12-hour light:12-hour dark cycle, at an ambient temperature of 22 ± 2 °C and relative humidity of 40–60%. Mice were fed ad libitum with free access to drinking water according to protocols approved by the Institutional Animal Care and Use Committees of the UT Southwestern Medical Center. All experiments involving live animals were approved by the Institutional Animal Care and Use Committees of the UT Southwestern Medical Center (protocol number 2015-101064).

Hsd3b6^fl/fl (C57BL/6), with loxP sites surrounding the first coding exon of Hsd3b6, were generated using CRISPR–Cas9 genome editing with guide RNAs targeting regions of the Hsd3b6 locus (Extended Data Fig. 2a). Guide RNAs were injected into fertilized C57BL/6J embryos by the Children’s Research Institute Mouse Genome Engineering facility at UT Southwestern. Healthy blastocytes were implanted in pseudo-pregnant mice. The resulting litter was screened by genomic sequencing to detect insertion of loxP sites and mice were bred to homozygosity and backcrossed with wild-type C57BL/6 mice. To generate K14-Cre^+/−; Hsd3b6^fl/fl (Hsd3b6^Δskin) mice, Hsd3b6^fl/fl mice were crossed with K14-Cre^+/− mice to generate K14-Cre^+/−; Hsd3b6^fl/+ mice; K14-Cre; Hsd3b6^fl/+ mice were crossed to Hsd3b6^fl/fl mice to obtain experimental mice K14-Cre^+/−; Hsd3b6^fl/fl (Hsd3b6^Δskin) and corresponding controls, Hsd3b6^fl/fl. Hsd3b6^fl/fl and Hsd3b6^Δskin status was determined using PCR primers (Extended Data Table 2) and resolving on a 3% agarose gel.

Method details

Quantification of hormones in mouse samples

Age- (7–8 weeks) and sex-matched Hsd3b6^fl/fl and Hsd3b6^Δskin mice were analysed. Blood samples were obtained from the retro-orbital vein of anaesthetized mice followed by serum isolation with the micro sample tube Serum Gel (catalogue number 41.1378.005; Sarstedt). Skin hormones were quantified from skin secretions. After anaesthesia with isoflurane, hair was removed using depilatory cream and shaving. After 24 hours, Sebutape (Clinical and Derm LLC) was applied to the dorsal surface for 15 minutes (Extended Data Fig. 2e). Steroid extraction was performed as previously described⁵. Sebutape was removed and placed in 3 ml of chromatography–mass spectrometry grade methanol (catalogue number A456-500; Thermo Fisher Scientific) in an 8 ml polytetrafluorethylene/rubber-lined vial (catalogue number 03-343-3E; Thermo Fisher Scientific). The sample was then dried by vacuum centrifuge at 40 °C and stored at −20 °C until analysis. For analysis, samples were reconstituted with 100 μl of appropriate hormone kit assay buffer. Steroid hormone quantification of progesterone, testosterone and DHT was measured by mouse-specific immunoassay (catalogue numbers MBS7606191, MBS266250 and MBS760829; My BioSource) following the manufacturer’s instructions.

Immunofluorescence microscopy

Mouse skin samples were fixed in formalin and embedded in paraffin by the UT Southwestern Histology Core. Samples were deparaffined with xylene followed by rehydration with decreasing concentrations of ethanol. Heat-induced antigen retrieval was attained in 10 mM sodium citrate buffer. Sections were washed briefly and blocked for an hour in blocking/permeabilization buffer (PBS, 5% goat serum and 0.5% Triton X-100). Sections were incubated in blocking/permeabilization buffer overnight with the following antibodies: anti-HSD3B6 (2.5 µg ml⁻¹; orb592071; Biorbyt), anti-cytokeratin-14 (1 µg ml⁻¹; sc-53253; Santa Cruz Biotechnology). After a brief wash in PBST (PBS + 0.2% Tween-20), sections were incubated with corresponding secondary antibodies: donkey anti-rabbit Alexa Fluor 647 (2 µg ml⁻¹; 711-605-152; Jackson Immunoresearch) and donkey anti-mouse Alexa Fluor 594 (2 µg ml⁻¹; A-21203; Thermo Fisher Scientific). The slides were then washed briefly in PBST and mounted with DAPI containing mounting medium (0100-20; SouthernBiotech). Images were processed using a Zeiss 780 confocal microscope.

Bacterial strains and plasmids

S. aureus strains (Extended Data Table 1) were streaked on tryptic soy agar plates and grown overnight at 37 °C. Single colonies were selected and cultured in tryptic soy broth (TSB) at 150 rpm at 37 °C in a shaking incubator overnight, followed by a 1:100 subculture at 37 °C in a shaking incubator to obtain bacteria from the mid-log phase (optical density at 600 nm (OD_600 nm) = 0.6). For reporter strains, all in vitro cultures were performed using TSB in the presence of 10 µg ml⁻¹ of chloramphenicol. Bacteria were pelleted, washed with PBS and resuspended in either TSB for in vitro experiments or PBS for in vivo experiments. HG003, ΔagrC, ΔagrBD and ΔagrA mutant strains were obtained from Dr Ferric Fang, University of Washington School of Medicine¹⁶. S. aureus strains from AD skin were obtained from Dr Julie Segre and Dr Heidi Kong³⁰. All other strains from the collections of the laboratories of A.R.H. and T.A.H.-T.

Construction of lux-expressing strains

The integrated luxCDABEG cassette was transduced into S. aureus strains HG003, ΔagrBD and ΔagrC obtained from the laboratory of Dr. Ferric Fang, University of Washington School of Medicine^16,40 using phage 11 generating strains AH6222 (lux+), AH6224 (lux+) and AH6223 (lux+), respectively.

Construction of complemented ΔagrC strain

The HG003 ΔagrC strain was complemented by introducing the pAgrC1AgrA plasmid⁶⁰ by phage transduction⁶¹ using phage 11.

In vitro luminescence assays

S. aureus strains, HG003, ΔagrBD and ΔagrC with constitutive Lux (φ11::LL29luxCDABEG) and quorum-sensing-dependent lux (pAmiAgrP3lux) plasmids (HG003 (AH6225) agr type I, USA100 (AH430) MRSA type II and MW2 (AH1747) MRSA type III)^14,18 were grown in TSB supplemented with antibiotic selection and subcultured 1:200 into fresh TSB containing steroid hormone. The assay was completed in opaque-sided, 96-well, clear-bottom, tissue-culture-treated plates with a final well volume of 200 μl. Bioluminescent signals (photons per 0.1 second acquisition time) were measured using a BioTek H1 Synergy plate reader. Experiments were completed in triplicate, with the agr-type-specific AIPs AIP-I (catalogue number 4515-v; Peptide Institute), AIP-II (catalogue number 4516-v; Peptide Institute) and AIP-III (catalogue number 4517-v; Peptide Institute) as positive control.

In vitro growth assay

S. aureus strains HG003 (agr type I), ΔagrBD, ΔagrC, USA100 (AH430; MRSA type II) and MW2 (AH1747; MRSA type III) were cultured overnight in TSB at 37 °C with shaking. Overnight cultures were diluted 1:100 in fresh TSB and grown to mid-log phase. Bacterial suspensions were inoculated into 96-well clear-bottom plates containing 10 nM testosterone, 10 nM DHT or vehicle control. OD₆₀₀ was measured over a 24-hour incubation period to assess bacterial growth using a BioTek H1 Synergy plate reader.

Haemolysis assay

Overnight cultures of HG003, ΔagrBD and ΔagrC strains were inoculated 1:200 into 10 ml of TSB containing testosterone, AIP-I or vehicle alone at concentrations of 10 nM. Cells were grown to mid-log phase. The supernatant of 1 ml of culture was filter sterilized using Millex sterile syringe filters with a pore size of 0.22 µm (catalogue number SLGV033RS). Filtered supernatant diluted 1:1 with PBS was added to 25 µl of human blood (catalogue numbers HUMANWBK2-0110649 and HUMANWBK2-0110717; BioIVT; catalogue number IWB1K2E-10ML; Innovative Research) to a 96-well V-bottom plate and incubated with agitation at 37 °C for 1 hour. After spinning at 1,000 rpm for 10 minutes, the supernatant was transferred to a flat-bottom 96-well plate. Absorbance was read at 541 nm (A₅₄₁) for haemoglobin using a BioTek H1 Synergy plate reader. The percentage of haemolysis was calculated using the following formula: (A₅₄₁ of RBC-treated sample − A₅₄₁ of buffer)/(A₅₄₁ of H₂O − A₅₄₁ of buffer); where buffer (PBS) = baseline, H₂O = 100% haemolysis⁷⁰.

Skin cell cytotoxicity assay

Cultures of HG003, ΔagrBD and ΔagrC strains were grown overnight with TSB containing testosterone, AIP-I or vehicle alone at concentrations of 10 nM. Bacteria were pelleted, followed by filter sterilization of the supernatant using Millex sterile syringe filters with a pore size of 0.22 µm (catalogue number SLGV033RS). Human keratinocytes (HaCaT cells; catalogue number T0020001; AddexBio) were used for cytotoxicity assays. These cells are spontaneously transformed keratinocytes derived from histologically normal skin of a male donor (62 years old). Cells were cultured in AddexBio-optimized DMEM supplemented with 10% FBS under standard conditions. HaCaT cells were treated with sterile-filtered bacterial supernatants at 5% by volume for 24 hours. After PBS washing, the resulting supernatants were used to measure lactate dehydrogenase (LDH) release from damaged cells using the LDH Cytotoxicity Detection Kit (catalogue number 2570393; Invitrogen).

Neutrophil killing assay

HG003, ΔagrBD and ΔagrC strains of bacteria were treated with testosterone, AIP-I or vehicle at concentrations of 10 nM and allowed to grow to mid-log phase (OD₆₀₀ = 0.6). Purified human neutrophils (catalogue number IQB-Hu1-Nu10; IQ Biosciences; catalogue numbers HUMANNEUT-0127862 and HUMANNEUT-0127879; BioIVT) were seeded at 1 × 10⁵ cells per well into a 96-well plate in 90 μl of RPMI. Bacterial supernatants (10 μl) were added (final concentration of 10%). After 3 hours incubation at 37 °C, 5% CO₂, the plates were centrifuged at 250g for 10 minutes, and the resulting supernatants were used to measure LDH leakage from damaged cells as the marker of neutrophil lysis with an LDH Cytotoxicity Detection Kit (catalogue number 2570393; Invitrogen). The percentage of neutrophil lysis was calculated using neutrophils incubated with 10% RPMI as 0% lysis control, and neutrophils incubated with 0.2% Triton X-100 were defined as 100% lysis⁷¹.

Quantitative real-time PCR

HG003, USA100 (AH3684), MW2 (AH843), ΔagrBD, ΔagrA, ΔagrC and AD strains were treated with testosterone and/or the respective AIPs at concentrations of 10 nM and allowed to grow to mid-log phase. Cells were pelleted and lysed using lysis matrix B tubes containing 0.1 mm silica spheres (catalogue number 174701; MP Lysing Matrix Tubes) and lysostaphin (catalogue number L7386; Sigma) at room temperature, and RNA was purified using the RNeasy Mini Kit (catalogue number 74104; Qiagen). RNA was quantified by absorbance at 260 nm, and its purity was evaluated by the ratios of absorbance at 260 nm:280 nm. RNA was used as a template to generate cDNA using the High-Capacity Reverse Transcription Kit (catalogue number 01071619; Applied Biosystems). Quantitative real-time PCR was performed by amplifying cDNA using Power SYBR Green Master Mix (catalogue 2749999; Applied Biosystems) and QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems). Relative expression values were calculated using the comparative Ct (ΔΔCt) method, and transcript abundances were normalized to gyrA transcript abundance for S. aureus and Rplp0 for mouse tissue samples. Primer sequences are shown in Extended Data Table 2.

RNA sequencing

Briefly, cultures of HG003 were grown in TSB with 10 nM or 100 nM of testosterone, 10 nM pregnenolone or DMSO alone in triplicate to an optical density of 0.4 at OD₆₀₀. Cells were collected and treated with RNA Protect Bacteria Reagent (catalogue number 76526; Qiagen). Cells were lysed using lysostaphin (catalogue number L7386; Sigma) and RNA purified using the RNeasy Mini Kit (catalogue number 74104; Qiagen). Sample quality was affirmed via Bioanalyzer (Agilent). Ribosomal RNA was depleted using RiboCop from the Bacterial META Removal Kit (Lexogen). cDNA libraries were generated at the University of Michigan Microbiome core using the CORALL RNA-seq Library Prep Kit (Lexogen). Samples were barcoded, pooled and sequenced in 125 × 125 paired-end reads on an Illumina HiSeq 2000 sequencer. Raw sequencing reads in FASTQ format were aligned and annotated to the S. aureus NCTC8325 reference genome with annotated small RNA⁶² using QiagenCLC Genomics Workbench default settings (v.21.0.5): mismatch cost, 2; insertion and deletion cost, 3; length and similarity fraction, 0.8. Normalization and differential expression calculations of uniquely mapped bacterial transcripts were performed using CLC. All transcripts with a false-discovery-rate-adjusted P 

S. aureus epicutaneous skin infections

Before infection studies, mice were acclimatized to the animal biosafety level 2 animal housing facility. Age- (7–8 weeks), strain- and sex-matched C57BL/6 male and female mice, Hsd3b6^fl/fl and Hsd3b6^Δskin were used in the study. A previously described mouse model of epicutaneous S. aureus exposure was followed^21,46. Briefly, the dorsal skin of anaesthetized mice (2% isoflurane) was shaved and depilated (Nair cream). After 24 hours, bioluminescent S. aureus strains were grown to mid-log phase, pelleted and resuspended in PBS to achieve inoculum containing 1 × 10⁶ CFU. A 100 μl volume of PBS containing 1 × 10⁶ CFU with or without 10 nmoles of testosterone, AIP-I or the same volume of vehicle was placed on a sterile gauze pad and attached to the shaved skin with transparent bio-occlusive dressing (catalogue number 1622W; Tegaderm 3M; Henry Schein Medicals), and secured with adhesive bandages (catalogue number 1275033; Band-Aid; Johnson & Johnson, American White Cross) for 4 days. Photons emitted from luminescent bacteria were collected during an auto-exposure using the IVIS Lumina 3 imager machine and living image software (Xenogen) over the course of 1 minute. Bioluminescent image data are presented on a pseudo-colour scale (blue representing least intense and red representing the most intense signal) overlaid onto a greyscale photographic image. Using the image analysis tools in living image software, circular analysis windows (of uniform area) were overlaid onto dorsal regions of the infection area, and the corresponding bioluminescence values (total flux) were measured and plotted versus time after infection. Mice were randomly assigned to treatment groups, and at experimental endpoints, mice were euthanized using carbon dioxide inhalation.

Disease scoring

The severity of skin inflammation was assessed by a blinded observer from digital photographs and the total disease score was quantified²¹. The sum of the individual grades for erythema, oedema, erosion and scaling were each graded as 0 (none), 1 (mild), 2 (moderate) or 3 (severe).

Transepidermal water loss measurement

Transepidermal water loss, a measure of barrier function and integrity, of mice dorsal skin was measured using Vapometer (Delfin Technologies) according to manufacturer instructions⁶³.

Histology

Skin biopsy specimens were collected, fixed in 10% formalin and paraffin embedded. Skin cross-sections of 4 μm were mounted onto glass slides and stained with haematoxylin and eosin by the UT Southwestern Histology Core, according to guidelines for clinical samples. Epidermal thickness was measured by taking ten epidermal thickness measurements per mouse, averaged from images (Echo Revolve Microscopy) using ImageJ software.

Measurement of quorum sensing in vivo

S. aureus strains expressing quorum-sensing lux (pAmiAgrP3lux) plasmids were grown in TSB medium containing chloramphenicol overnight at 37 °C in a shaking incubator set to 150 rpm. Overnight cultures were diluted 1:100 TSB with chloramphenicol to mid-log phase and then pelleted and washed 2× in PBS and resuspended in sterile saline. PBS inoculum suspensions (100 μl) containing 1 × 10⁶ CFU were placed on a sterile gauze pad (1 cm × 1 cm) and attached to the shaved skin with transparent bio-occlusive dressing, with or without testosterone, ent-T, AIP-I or vehicle (3M; Tegaderm) and secured with 2 layers of adhesive bandages (Band-Aid; Johnson & Johnson). Beginning immediately after infection, mice were imaged hourly under isoflurane inhalation anaesthesia (2%). Photons emitted from luminescent bacteria were collected during auto-exposure using the IVIS Lumina 3 imager machine and living image software (Xenogen). Corresponding bioluminescence values (total flux) were measured and plotted versus time after infection⁴⁰.

In silico docking of testosterone to the AgrC receptor

The dimeric structure of AgrC was predicted using AlphaFold 2 (refs. ^33,34) as implemented in Google Colab. The sensory domain of a single subunit (residues 1–207) was selected for ligand docking. AIP-I was docked in silico using SwissDock^35,37 with nuclear-magnetic-resonance-derived coordinates for AIP-I. The docking pose for AIP-I that best aligned with reported structure–activity relationships^11,41 was selected as the AgrC–AIP-I complex for subsequent steroid docking. Stereospecific compound templates were retrieved from PubChem (testosterone; compound identifier 6013). Initial docking produced an AgrC model featuring a depression on the sensory domain surface, consistent with a putative membrane-exposed ligand-binding pocket. To refine the complex, co-folding of AgrC in the presence of both AIP-I and testosterone was performed using AF3⁶⁴, implemented in PXDesign^65,66 on the Protenix Server (https://protenix-server.com). AF3 successfully docked both ligands but introduced stereochemical distortions consistent with known AF3 limitations⁶⁴. These effects were also reproduced using Boltz-2 (ref. ⁶⁷). To correct for these artefacts, the co-folded AgrC–AIP-I complex was used as the target for in silico docking of testosterone with an updated version of AutoDock Vina^68,69. All structures were visualized and analysed using PyMOL v.2.5.2 (Schrödinger).

Quantification and statistical analysis

Statistical details of experiments can be found in the figure legends, including how significance was defined and the statistical methods used. Data represent mean ± s.e.m. Formal randomization techniques were not used; however, mice were allocated to experiments randomly. Mice that were determined to be in the anagen hair cycle at the initiation of the experiment were excluded. All statistical analyses were performed with GraphPad Prism software (v.10.0.1), except the bioluminescent imaging data that was analysed as described above. To assess the statistical significance of the difference between two treatments for in vitro models, we used unpaired two-tailed Student’s t-tests. To assess the statistical significance of the difference between two treatments for mouse models, we used the Mann–Whitney U-test and Kolmogorov–Smirnov test. To assess the statistical significance of differences between more than two treatments in vitro, we used one-way ANOVA with post-test corrections. To assess the statistical significance of differences between more than two treatment groups in mouse models, we used the Kruskal–Wallis test with post-test corrections. For experiments in vitro and in vivo where metrics were calculated over time, we used two-way ANOVA with post-test corrections. For the RNA-sequencing experiments, expression data were analysed with CLC. All transcripts with a false-discovery-rate-adjusted P 

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.