Our understanding of the neuroendocrine regulation of the hair follicle (HF) has advanced over the past two decades. The hypothalamic-pituitary-thyroid (HPT) axis couples neuroendocrine control to peripheral tissue gene expression and metabolism: hypothalamic thyrotropin-releasing hormone (TRH) drives pituitary thyroid-stimulating hormone (TSH), TSH stimulates thyroidal release of thyroid hormones (TH) triiodothyronine (T3) and thyroxine (T4). The active ligand T3 enters target cells via dedicated transporters – MCT8/10 and LAT1/2 in human skin and hair – to bind nuclear receptors TRα and TRβ and regulate thyroid hormone-responsive gene expression. More common than direct transport in vivo, local deiodinases generate receptor-active T3 from the inactive T4 via the DIO2 enzyme. The skin functions as a peripheral neuroendocrine organ and, within the human HF, recapitulates key elements of this axis: extrathyroidal TSH-receptor signaling, epithelial TR activity, intrafollicular deiodination, and a local TRH system. After all, epithelial barrier systems preceded the evolution of centralized brains, and the HF and brain arise from a common ectodermal primordium developmentally [3]. Clinically, both hypo- and hyperthyroidism perturb human hair cycling and fibre quality, underscoring that thyroid hormone signaling is a bona fide regulator of follicle homeostasis in vivo.

The human HF is a compartmentalized mini-organ that both senses and generates hypothalamic cues. Its epithelium expresses TRH and the TRH receptor (TRH-R): TRH is produced by outer-root-sheath (ORS) keratinocytes, while TRH-R localizes predominantly to the inner root sheath (IRS) [1,4]. The HF ORS and IRS keratinocytes together constitute a peripheral analogue of the HPT-axis: together, they synthesize TRH and mount TRH-driven programs, such as mitochondrial biogenesis (PGC-1α, TFAM, HSP60, MTCO1), increased mitochondrial mass and respiration, and coordinated transcriptional responses [1,4–6]. TRH has been found to elevate intrafollicular proopiomelanocortin (POMC) and its melanocortin products, both adrenocorticotropic hormone (ACTH) and alpha-melanocyte-stimulating hormone (α-MSH), in the IRS and hair matrix and, via these paracrine cues, stimulate hair bulb melanocyte proliferation and melanogenesis. The most parsimonious pathway for TRH-driven hair pigmentation is: TRH from ORS keratinocytes → TRH-R activation on IRS cells → POMC processing → IRS release of ACTH + α-MSH → melanocyte MC1R activation → cAMP - PKA signaling → MITF-mediated proliferation and melanogenesis [4,6]. In organ-cultured female human scalp HFs, TRH prolonged the anagen phase, increased proliferation and reduced apoptosis of hair matrix cells, antagonized TGF-β2-induced catagen, and remodeled epithelial programs (reduced K6 in ORS). Consistent with these effects, TRH attenuated ATM- and ATR-dependent phosphorylation of p53, indicating TRH dampens localized DNA-damage checkpoint activation within the follicular epithelium [1].

Beyond these growth-cycle and mitochondrial effects, TRH has been found to modulate epithelial hormonal and cytokine signalling in intact follicles. In organ-cultured female human scalp HFs, TRH reshaped prolactin (PRL) - prolactin receptor (PRLR) signaling, a cytokine-receptor pathway with both endocrine and local autocrine and paracrine roles, and of much interest to prior hair loss research. Specifically, TRH increased whole-follicle PRL mRNA and protein and, while elevating PRLR mRNA, reduced PRLR protein in the ORS, a divergence consistent with post-transcriptional control (from, e.g., ligand-triggered receptor internalization, lysosomal turnover, or altered translation) [7]. By contrast, in the same organ-culture system, 17β-estradiol increased PRL and PRLR protein in defined epithelial compartments, particularly the ORS, and, in parallel ORS-keratinocyte cultures, upregulated PRL and PRLR transcripts; unlike TRH, it did not produce the selective ORS PRLR downshift and instead tended to increase epithelial PRLR [7]. Functionally, TRH thus couples a high PRL ligand and limited PRLR receptor state in the ORS, tempering overstimulation while preserving paracrine prolactin tone, whereas 17β-estradiol tends to globally sensitize the follicle’s epithelium to PRL. This axis is also of translational interest, given anti-PRLR therapeutics in development for androgenetic alopecia (AGA) (e.g., HMI-115, a monoclonal PRLR-blocking antibody by Hope Medicine), which aim to blunt PRL-driven epithelial responses [8]. Beyond follicular regeneration programs, TRH accelerates re-epithelialization and enhances keratinocyte proliferation and activation in human skin wound models, marking it as an active epithelial morphogen in skin and appendages [9]. Together, these findings define a keratinocyte-centric, autocrine, and paracrine TRH module within the HF and skin epithelium that coordinates metabolic, pigmentary, cytokine, and reparative programs, with clear translational promise [4–9], and with supportive preclinical safety data for TRH delivery to mammals [10].

Having established the epithelial TRH module, we next consider TSH, whose receptor’s (TSH-R) distribution in the human scalp is mesenchyme-biased and whose acute effects in intact follicles are biochemically evident yet morphologically modest. The mesenchyme, composed of the dermal papilla (DP) and connective tissue sheath (CTS), shows strong TSH-R expression. By contrast, the follicular epithelium, composed of the hair matrix, ORS, hair bulge, IRS, and hair shaft, is minimal or undetectable for TSH-R [11–13]. In human HFs ex vivo, TSH engaged the follicular axis biochemically, with increased cAMP in the culture medium by ~3-fold compared to control follicles. Additionally, a shift in a restricted set of TSH-responsive transcripts was found, despite no change in classical or macroscopic hair-regrowth markers compared to control. Classical thyroid genes – thyroglobulin (TG) and thyroid transcription-factor 1 (TTF1) – were upregulated at the whole-follicle level, along with a slight increase in connective tissue growth factor (CTGF). Further downstream effects were noticed locally, such as mesenchymal contractility in the CTS (via α-SMA expression), enhanced mitochondrial respiratory capacity in the matrix (via MTCO1 expression), and epithelial cytoskeletal programming (via KRT5 expression) – concrete pathway engagement even when macroscopic growth and pigmentation endpoints remain largely unchanged [11].

Complementary, compartment-specific studies underscore that TSH-R coupling in skin is strongly context-dependent. In primary interfollicular dermal fibroblasts, TSH increased proliferation (BrdU ≈1.5× at 72 h) without a 2 h intracellular cAMP rise and without induction of classical TSH-dependent Tg/TPO/NIS/TSH-R mRNAs or thyroglobulin secretion, indicating non-canonical (non-cAMP-dominant) TSH-R coupling, likely via MAPK - ERK, PI3K - AKT, or β-arrestin-scaffolded routes [12]. In contrast, in isolated primary keratinocytes, TSH-R was confirmed by Western blot, and TSH elicited direct receptor-coupled responses (intracellular cAMP↑ at 2 h; BrdU↑ at 72 h), suggesting a route of canonical action in the epidermis of the skin [12]. In full-thickness skin organ culture, TSH upregulated epidermal differentiation markers (involucrin, loricrin, K5/14) without mesenchymal read-outs reports, despite epidermal TSH-R protein being below IHC detection-consistent with paracrine or low-abundance receptor signalling [13]. Taken together, TSH clearly activates HF and skin thyroid signalling, yet its modulation of intact human HF growth, apoptosis, and pigmentation appears highly niche-dependent and comparatively muted compared to other elements of the HPT axis [11–13].

The most downstream component of the HPT axis, thyroid hormones themselves act directly on the human follicle. Nuclear thyroid hormone receptors (TRs) are compartmentally expressed, and exogenous T3 and T4 independently reshape anagen, pigmentation, and epithelial programs [14,15]. Nuclear TRs, specifically thyroid receptor TRβ1 (TRα has not been found in skin or the HF), show strong nuclear signal in the ORS and hair matrix, moderate signal in the IRS, and weak-to-moderate signal in the DP and CTS, as mapped in human hair follicular units [14]. In human scalp HFs ex vivo, both T3 and T4 prolonged the anagen phase compared to control, increased proliferation and reduced apoptosis in the hair matrix, stimulated intrafollicular pigmentation, reduced catagen promoter TGF-β2, and remodelled epithelial keratins (CK6↑, CK14↓); hair-shaft elongation itself showed no consistent gain [15]. In full-thickness human scalp skin, topical T3 and T4 increased the percentage of follicles in the anagen phase, but without significant changes in matrix Ki-67 expression or melanogenesis by Warthin–Starry, while regrowth markers K15 (bulge), IGF-1 + FGF-7 (proximal ORS), and CD31+ endothelial cells increased and catagen promoter TGF-β2 decreased, indicating growth-supportive reprogramming with dose and vehicle-specific nuances [2]. As complementary evidence from primary genetics, mice lacking both TRα1 and TRβ (both TRβ1 and TRβ2 isoforms) display impaired hair cycling and reduced follicular keratinocyte proliferation, whereas single-receptor knockouts show milder or compensable phenotypes – indicating functional requirement with partial isoform redundancy [16,17].

Local TH signalling is shaped by the intrafollicular deiodinase system, DIO2 and DIO3. DIO2 converts T4 into T3 – the active form of thyroid hormone – while DIO3 inactivates thyroid hormone activity via the conversion of T3 into T2 and T4 into rT3. Microdissected human scalp follicles transcribe both enzymes, and DIO2 activity is demonstrated in primary human epidermal keratinocytes in monolayer culture (neonatal foreskin and adult epidermis), converting T4 to T3 in vitro [15,18,19]. In immortalised human DP cells, DIO2 was detected in non-balding DP cells and below readable expression in balding DP cells – posing the question whether balding DP cells have lower DIO2 and consequently T3 activity; DIO3 was not detected in either group [20]. In addition, DIO2 transcription has been reported in human bulge epithelial stem cells isolated by navigated laser-capture microdissection and profiled by microarray, supporting local T4 to T3 activation within the stem-cell niche [21]. Lastly, in mouse skin in vivo, DIO3 is enriched in the anagen hair-matrix epithelium and absent from the DP, indicating local inactivation of TH in the proliferative matrix and a relative lack of TH quenching in the DP, consistent with a compartmental thyroid-hormone gradient [22]. Compartment-resolved localisation of DIO2 and DIO3 within in-tact human HF remains unsettled, and leaves an opening for a future study [15]. Together, historical research positions THs as a locally regulated, druggable axis in the follicle, one that supports anagen maintenance and epithelial remodeling via TRβ1, and is tunable through deiodinase activity and delivery context [2,13–15,22].

Pharmaacokinetics

TRH is short-lived in blood (~7-minute half-life), as it is rapidly cleaved by the TRH-degrading ectoenzyme (TRH-DE; pyroglutamyl aminopeptidase II), producing brief systemic peaks and dose-limited tolerability - hence the rationale for protected delivery and non-invasive routes [23–25]. In primates and rats, TRH nanoparticles administered intranasally were well tolerated, with no drug-attributed systemic or local toxicity, supporting translational delivery to epithelia and appendages [10]. Taltirelin, a TRH analogue approved for human use in Japan, speeds re-epithelialization in mouse wounds and drives fibroblast migration and proliferation in vitro, proof that TRH analogues can work in skin, even if human HF results have not yet been reported [26]. At the human TRH-R, taltirelin shows lower binding affinity and lower IP₃ - Ca²⁺ pathway potency than TRH, but higher intrinsic efficacy in selected assays, a profile consistent with durable signalling and resistance to rapid breakdown [27]. Whether those receptor-level differences translate in human skin is ultimately gated by access: only the ligand that traverses the follicular or epidermal barrier and survives peripheral cleavage can engage TRH-R in situ [23–27]. The TRH-degrading ectoenzyme operates outside the central nervous system (CNS), with a soluble liver form shaping peripheral clearance. Mapping this enzyme within human HF and scalp skin compartments remains an open task [23–25].

In parallel with TRH pharmacokinetics and delivery, the active T3 merits brief pharmacokinetic context. In human plasma, T3 circulates at ~2 nM (vs. ~90 nM for T4) and is almost entirely protein-bound; only ~0.3% of T3 is freely circulating (vs. ~0.03% for T4) [21,28]. For both T3 and T4, binding to TBG, TTR, and albumin shrinks the free pool and slows clearance; because T4 binds these carriers more tightly than T3 (free T4 ~0.03% vs. free T3 ~0.3%), it is cleared more slowly, with a longer half-life (~7 days) than T3’s (~1 day in euthyroid adults; shorter in hyperthyroidism, longer in hypothyroidism) [21,28]. In practice, immediate-release liothyronine (LT3) has a relatively short clinical effective half-life of ~5–6 h, while the terminal elimination half-life of total serum T3 is ~19 h; T3 falls fast at first but has a slower terminal phase [29,30]. Once unbound, free T3 traverses dedicated transporters (e.g., MCT8/10) to engage nuclear TRs [31]. For topical T3, protein binding depends on compartment: it is not protein-bound at the skin surface; after penetrating viable tissue, it may bind interstitial albumin or enter cells via transporters, and any fraction reaching the bloodstream becomes > 99% carrier-bound, which, with deiodination and hepatic clearance, then governs systemic pharmacokinetics. As a complementary strategy, TRβ-selective thyromimetics (e.g., Resmetirom, VK2809, TDM-105795) favor TRβ and liver partitioning, aiming to retain treatment benefits while minimizing TRα-mediated off-targets [22]. Early topical TRβ data are encouraging, with TDM-105795 showing positive Phase 2 clinical trial results, but not yet peer-reviewed for human HF ex vivo [33].

Organ culture of microdissected human scalp HFs provides a unique, physiologically relevant model for interrogating both specific HF and general human biology [34]. By treating intact follicular units surrounded by scalp skin, one can capture paracrine signalling between the epithelial and mesenchymal compartments of humankind’s smallest mini-organ, including cells of the perifollicular microenvironment [34]. In this pilot study, we exploited this model to examine how combined T3 + TRH therapy alters the human HF transcriptome, the first test of a T3 + TRH combination not only in hair follicle biology, but in all biology. We hypothesised that topical co‑administration of T3 and TRH at physiologically relevant nanomolar to micromolar doses would elicit transcriptional changes important for hair growth.

Follicular unit procurement and culture

Ten terminal follicular units – each containing one or more full-length HFs with their perifollicular microenvironment – were harvested by follicular unit transplantation (FUT) from the occipital scalp of a single male donor with AGA. Explants were equilibrated overnight in serum-free Williams’ E medium supplemented with insulin (10 µg/mL, Sigma), L-glutamine (2 mM, Sigma), hydrocortisone (10 ng/mL, Sigma), and 1% penicillin-streptomycin (Sigma), at 37 °C/5 % CO₂, in line with established human HF organ-culture protocols. On the following day, medium was renewed and units received either supplemented Williams’ E Medium only (Control) or supplemented Williams’ E Medium with escalating TRH doses (2.548 µM, 25.48 nM, 254.8 nM, Sigma) with a constant T3 dose (10 nM, Sigma), with concentrations informed by prior work demonstrating that each ligand alone prolongs anagen in human hair-follicle organ culture [1,2]. Treatments (1 mL per unit) were renewed every other day for seven days.

Treatment groups

The four experimental groups were defined by the concentrations of T3 and TRH applied to the follicular units (Table 1). Each group contained two or three replicates.

Table 1: Treatment Groups

Culture timeline

Media were changed and treatments reapplied on alternate days. At each medium change, macroscopic images of each follicular unit were recorded with an Echo Revolve microscope with settings as shown (Table 2).

Table 2: Echo Revolve Microscope Settings for Day 1 and Day 7 Collected Images

On day 7, eight hours after the final treatment, follicular units were snap-frozen in liquid nitrogen and transported at –80 °C before RNA extraction (Table 3).

Table 3: Treatment Timeline

RNA-sequencing and analysis

Total RNA was isolated from each follicular unit. Integrity was assessed with an Agilent TapeStation; RNA integrity numbers ranged from ≈ 5.9–7.3. mRNA was enriched by poly‑A selection, spike‑in controls were added and libraries were prepared with unique molecular identifiers. Sequencing was performed on an Illumina platform (paired-end, ≈50 million reads per sample). Reads were aligned to the human reference genome using STAR, and raw gene counts were imported into DESeq2. Library size normalisation and dispersion estimation were performed within DESeq2, with a design matrix specifying group assignments. Differential expression was assessed using Wald tests and Benjamini-Hochberg adjustment (α = 0.05). Functional enrichment was conducted using KEGG pathway analysis.

Macroscopic morphology

Serial imaging at each medium change provided macroscopic length readouts for individual follicles over Days 1–7 in our human HF organ-culture system. Lengths were taken directly from unit images, and per-follicle Day 1–7 percent changes were computed. Follicular units were labelled as follows, according to their unique well plates: TRH-High (Group A: A1; A2; A3); TRH-Med (Group B: A4; B1); TRH-Low (Group C: B2; B3; B4); and Control (Control: C1; C2). To avoid pseudo-replication, we treated unique shafts from the same follicle as within-follicle biological replicates for visualization (e.g., A4a–b; C2a–c) and interpreted central tendencies at the group level (Fig. 1–4). Two prespecified exclusions were applied in length analysis: one Group C follicle (B3) exhibited shaft breakage due to skin curvature near the hair bulb, and one control follicle (C1) entered catagen during the experiment (Fig. 9). Both were omitted from our hair-shaft elongation analysis (Fig. 10–13).

Hair-shaft elongation

All included follicles elongated measurably over 7 days in organ culture (Fig. 1–8), with every Day-7 value exceeding its Day-1 baseline. Across shafts, percent length change ranged from ~12.1% to ~29.8% (Fig. 10). The largest response was observed in Group C (TRH-Low; B2, 29.8%), while the most modest response occurred in Group B (TRH-Med; A4a, 12.1%). Median percent elongation by group was: Group A (TRH-High) ~17.96% (mean ~19.94%), Group B (TRH-Med) ~18.05% (mean ~16.33%), Group C (TRH-Low) ~24.36% (mean ~24.36%), and Control (C2) ~20.54% (mean ~20.44%) (Fig. 11–13). Thus, within this exploratory series, all TRH doses preserved anagen-like macroscopic output, with a descriptive trend toward greater elongation at Group C (TRH-Low) relative to control, and broadly comparable outputs for Group A (TRH-High) and Group B (TRH-Med). These macroscopic outcomes cohere with the mechanistic readouts presented in our RNA-sequencing analysis, and are directionally compatible with pro-anagen TRH and T3 application in intact human HFs.

Dose-dependent global transcriptional shifts

Principal-component analysis (PCA) revealed clear separation of controls from treated follicles, with dose-stratified clustering among T3 + TRH groups. Replicates grouped by dose, indicating a robust treatment effect with clear separation by dose (Fig. 14). Differential expression analysis quantified these shifts. Relative to control, Group C (TRH-Low) induced ~532 DEGs (predominantly upregulated), Group A (TRH-High) induced ~237 DEGs (also up-skewed), and Group B (TRH-Med) induced ~216 DEGs (skewed toward downregulation) (Fig. 15). Thus, Groups C and A doses elicit broad transcriptional activation, whereas the intermediate dose applied in Group B appears to engage homeostatic (negative-feedback) programs. One would not expect a triphasic dose-response from TRH application a priori, and so Group B’s deviation from Groups C and A may be due to its smaller sample size (2 units versus 3), phase differences in the hair cycle (e.g., a catagen shift), or natural variance.

Activation of hypoxic and lipid-metabolism pathways

In the KEGG analysis, hypoxia-inducible factor (HIF) signaling is noticeably enriched in follicular units treated with T3 + TRH. Within the HIF family, the enrichment reflects the canonical HIF-1α program – glycolysis and angiogenesis – with VEGFA, SERPINE1/PAI-1, and PDK1-mediated glycolytic routing evident in our dataset (Fig. 16). In human follicles, both epithelial and mesenchymal compartments mount clear HIF-1 programs: hypoxic preconditioning of human HFSCs reduced oxidative stress and cell death while increasing VEGF transcription [35]; in a complementary model, hypoxia induced HIF-1α - VEGF signalling in adipose-derived mesenchymal stem cells, whose secretome enhanced human DP proliferation via ERK1/2, an effect blunted by HIF-1α knockdown [36]. In DP cells, HIF-1 outputs have been readily inducible: prolyl-hydroxylase inhibitors (hypoxia mimetics) raise DP proliferation to minoxidil-like levels in vitro [37]. Consistent with this biology, our data show dose-responsive activation of downstream HIF-1α targets in Groups C and A, supporting the view that combined treatment counteracts hypoxia-related growth repression and favors anagen maintenance, in line with effects from the individual treatment of TRH and THs in human HFs [1,2].

In a contrasting KEGG analysis from a past study, it was reported that HIF-1 signaling enrichment was driven chiefly by elevation of the oxygen-sensing prolyl-hydroxylases (EGLN1/3) (Table 4) [38]. These components of HIF-1’s negative-feedback arm indicate altered hypoxia feedback rather than the broad activation of downstream targets, as observed in our dataset. In the same past study, EGLN1/3 were found upregulated in vertex follicles by RNA-seq and corroborated by qPCR and immunofluorescence, with staining most evident in ORS and connective-tissue sheath [38]. Because EGLN1/3 enzymes are induced by HIF signalling and then dampen the pathway’s activity, their elevation may reflect compensatory activation or suppression of the pathway. Taken together, our findings point to therapeutic activation of downstream HIF-1α effectors under T3 and TRH combination, whereas this past AGA dataset reflects altered hypoxia sensing. We hypothesize that HIF-linked angiogenesis may be beneficial by improving vascular support and that HIF-driven glycolytic routing may be beneficial by meeting the energy demands of dividing stem cells. Both hypotheses warrant direct, compartment-resolved testing in humans clinically over a multi-month time period.

Comparison: T3 + TRH (treated vs. control) vs. AGA (vertex vs. occipital):

Table 4: Summary of Data Across Our and Related Studies

Peroxisome proliferator-activated receptor (PPAR) signalling was also enriched in our KEGG analysis (Fig. 16). Within the PPAR family (PPARα, PPARβ/δ, PPARγ), the enrichment maps to the PPARα and PPARγ programs. Both predominantly expressed in the HF epithelium to run complementary programs, PPARα is the ‘burn’ switch, coordinating fatty-acid uptake and catabolism via peroxisomal and mitochondrial β-oxidation (e.g., ACOX1, CPT1A) to fuel oxidative phosphorylation; PPARγ is the ‘organize’ switch, orchestrating a lipid-handling program (e.g., CD36, FABPs, ACSLs, PLIN1, SCD1) that maintains a non-toxic lipid milieu, trans-represses NF-κB - AP-1 to dampen TNFα - IL-6 signalling, and preserves ORS stem-cell competence [39–42]. Genetic disruption of PPARγ in mouse skin provokes cicatricial alopecia, and broader PPAR pathway disruption is seen in human AGA progression. Paired analyses of vertex and occipital follicles show that, as hair bulbs miniaturize, PPARGC1A (encoding the co-activator PGC-1α) and broader PPAR pathway programs increase, consistent with activation of a PPAR-centric metabolic program during disease progression [40]. In parallel, bald scalps show higher sebum triglycerides and palmitic acid, suggesting a lipid-rich milieu that could secondarily recruit PPAR programs as compensation [43].

In our dataset, the leading-edge genes resolve family-level PPAR engagement into an α-skewed oxidative and peroxisomal module (ABCD2, ACSL1, ACSL5, PECR, IDH2) with a γ component (PPARG, CD36). Functionally, PPARα agonism has prolonged anagen in human HFs ex vivo (within a Goldilocks dose window), whereas PPARγ activation reduces hair-shaft production even as it dampens inflammatory cues (↓TNFα, ↓IL-6) and enhances bulge stem-cell markers (↑K15) – benefits for the ORS niche that nevertheless came with a net anti-trichogenic effect at the follicle level, likely reflecting too drastic of a lipid diversion in other cell compartments [41,42]. Notably, PPARGC1A was not differentially expressed in any treatment group in the present study, indicating that PPAR-pathway engagement under T3 + TRH can proceed via receptor - target induction (PPARG, CD36, ABCD2, PECR, ACSL1/5) without a detectable rise in the co-activator PPARGC1A (PGC-1α). [41,42]. Practically, a PPARα-biased and γ-calibrated approach, without concurrent PPARGC1A up-shift, may sustain lipid homeostasis and inflammation control without the anti-trichogenic liabilities of a full PPARγ - PGC-1α program.

Metabolic reprogramming of carbon and fatty-acid pathways

Distinct from the PPAR-focused lipid findings above, a broad carbohydrate-metabolism signature was observed. In Groups C and A, gluconeogenesis, glucagon signalling, fructose - mannose metabolism, and pyruvate metabolism were enriched, indicating increased carbon flux under treatment. Fatty-acid-linked modules, including adipocytokine signalling, fatty-acid metabolism, regulation of lipolysis, and glycerolipid metabolism, were likewise enriched across doses (Fig. 16). These lipid modules are engaged not only via PPAR transcriptional control but also via hormonal - AMPK crosstalk (adiponectin, leptin, insulin → AMPK), which post-translationally tunes fuel choice: promoting fatty-acid oxidation, limiting lipotoxic synthesis (via ACC - SREBP axes), and coordinating autophagy and membrane remodelling [44–50]. This systems-level coupling complements the carbohydrate shifts: while glycolysis supplies fast ATP and anabolic precursors, AMPK-directed lipid routing prevents overload and supports high-throughput turnover in the ORS and matrix [39, 50–53]. Together, these signatures point to a coordinated shift in carbon and lipid metabolism under T3 + TRH, compatible with epithelial turnover in the ORS and hair matrix and prolonged anagen [2].

HIF - Glycolysis Connection

In review of these signatures, metabolic control of follicular compartments is closely coupled to the connection between hypoxia and glycolysis. Low O₂ stabilises HIF-1α, which transcriptionally routes carbon toward anaerobic glycolysis (GLUT1, HK, PFK, LDHA) and away from mitochondrial oxidation via PDK1-mediated PDH inhibition [54]. The human epithelial HF stem-cell niche is physiologically hypoxic, and glycolytic metabolism regulates HFSC and ORS-progenitor proliferation [39]. In mice, keratin-15-positive (K15⁺) HFSCs preferentially use glycolysis, whereas their rapidly proliferating Ki67⁺ progeny rely more on oxidative phosphorylation [55]. Concordantly, murine HFSCs in vivo are enriched for glycolytic programs, consistent with a glycolysis-biased stem-cell state [55,56]. Also in vivo, cAMP - CREB signalling is found to increase murine HFSC’s glycolytic flux and prime them for activation and anagen entry. Moreover, topical adrenergic agonists or direct cAMP delivery are sufficient to activate murine HFSCs, with K15⁺ HFSC transcriptomics showing upregulation of Slc2a1 (GLUT1), Gapdh, and Ldha and skin metabolomics demonstrating increased lactate production [57]. Human intervention studies may align with this model. Topical mitochondrial-pyruvate-carrier (MPC) inhibition may drive early anagen in vivo: MPC inhibitors were found to trigger visible pigmentation and hair growth within 6–9 days, approximately half the time to anagen seen in vehicle-treated controls [58,59]. Derivatives of this mechanism are now in clinical testing [58,60,61]. Additionally, topical inhibition of the electron transport chain shortens telogen and triggers anagen onset by favouring glycolytic flux [62]. Together with our KEGG enrichment (glycolysis, gluconeogenesis, pyruvate and lipid metabolism), these data support a model in which T3 + TRH shifts follicular bioenergetics toward HIF-1-stabilized glycolysis and coordinated lipid handling to sustain anagen.

Synergistic actions of TRH and T3

TRH and T3 engage complementary axes within the human HF. TRH acts through epithelial TRH-R to trigger Ca²⁺- IP₃ and cAMP - CREB signalling, boost mitochondrial programs, and promote shaft elongation with anti-apoptotic effects [1,5,6], whereas T3 binds nuclear TRs (predominantly TRβ1 in the hair matrix and ORS) to reprogram transcription and metabolism, with topical THs alone prolonging anagen and enhancing mitochondrial function ex vivo [2,14,15]. In combination, these inputs converged on broader pathway activation than expected from either alone in our dataset: HIF-1–linked angiogenesis and glycolytic routing, α-skewed PPAR and peroxisomal lipid handling with selected γ components, and suppression of catagen cues (e.g., TGFB2) co-occurred, features consistent with a microenvironment that simultaneously increases vascular support, shifts bioenergetics toward glycolysis, and maintains niche lipid and inflammatory homeostasis [1,2]. Conceptually, TRH’s rapid epithelial signalling and mitochondrial priming [5,6] can potentiate T3-dependent transcriptional remodeling [14,15], yielding a coordinated pro-anagen state that neither ligand alone achieves as comprehensively.

Limitations

This pilot study analysed a small number of ex vivo follicular units from a single donor. Although organ culture is a physiologically relevant model [34], future studies should include more donors and in vivo validation. Differential expression analyses were based on bulk RNA‑seq and therefore cannot resolve cell‑type‑specific responses. Despite these limitations, the dose‑dependent transcriptional changes and pathway enrichments observed provide mechanistic insights and lay the groundwork for further discovery.

Conclusion

Combined topical T3 + TRH treatment reprograms the human hair-follicle transcriptome in a clear, dose-dependent manner, with PCA and DEG profiles showing robust separation from controls. Mechanistically, the co-therapy converges on a pro-anagen state, characterized by activation of canonical HIF-1α readouts (e.g., VEGFA, SERPINE1/PAI-1; PDK1-linked glycolytic routing), engagement of PPAR-centric lipid programs (α-skewed with selective γ components), and suppression of catagen pressure (TGFB2↓), a pattern that aligns with and extends prior single-agent TRH and TH effects in human HF systems [1,2]. Together, these changes imply improved vascular support, glycolysis-biased bioenergetics for rapidly cycling epithelia, and calibrated lipid and inflammatory homeostasis within the niche, all compatible with sustained anagen and follicle integrity [1,2]. While confirmation in multi-donor cohorts and in vivo is warranted, these data nominate intermittent topical T3 + TRH co-therapy as a rational, mechanism-guided strategy for hair-loss disorders, and motivate next-step studies that are compartment-resolved (single-cell or space-transcriptomics), protein-level (e.g., HIF-1α, PDK1, PPARα, PPAR γ), and outcome-anchored (anagen duration, shaft output), with careful attention to dose scheduling and systemic thyroid safety.

References