Abstract

   Epigenetic age acceleration has previously been observed in
   inflammatory skin disease; however, less is known regarding recently
   described age-related gene expression patterns (“transcriptional
   clocks”). We investigated the role of transcriptional clocks in
   patients with hidradenitis suppurativa (n = 37), those with atopic
   dermatitis (n = 27), those with plaque psoriasis (n = 28), and healthy
   subjects (n = 38) using 7 clock algorithms, to improve the
   understanding of underlying pathophysiology and disease trajectory.
   Five of 7 transcriptional clocks demonstrated moderate-to-strong
   accuracy in predicting age across groups (patients with atopic
   dermatitis: ρ = 0.40–0.86, those with hidradenitis suppurativa: ρ =
   0.46–0.74, those with plaque psoriasis: ρ = 0.50–0.80, healthy
   subjects: ρ = 0.32–0.60; P < .05). Age acceleration was observed in
   lesional versus healthy (patients with atopic dermatitis: +3.9∼9.8y, t
   = 2.8∼5.9; those with hidradenitis suppurativa: +5.0∼6.1y, t = 2.5∼4.1;
   those with plaque psoriasis: +6.5∼12.5y, t = 5.1∼8.0; P < .05) and in
   lesional versus nonlesional skin in all diseases and less frequently
   observed in nonlesional versus healthy skin. In atopic dermatitis,
   loss-of-function sequence variants in the FLG gene were associated with
   transcriptional age acceleration, including FLGR244X/2282del4 dual
   carrier status (t = 2.3, P < .05) and FLGR501X carrier status (t = 2.6,
   P < .05). Pathway enrichment analyses revealed that clock genes are
   enriched in signatures related to aging, inflammation, and metabolism.
   Our study provides evidence for transcriptional age acceleration in
   inflammatory skin disease and sets a foundation for further
   investigation into the role of age-related transcriptional changes in
   the pathophysiology of these diseases.

   Keywords: Aging, Atopic dermatitis, Hidradenitis suppurativa,
   Psoriasis, RNA-seq

Introduction

   Hidradenitis suppurativa (HS), atopic dermatitis (AD), and plaque
   psoriasis (PP) are common inflammatory skin diseases that collectively
   impact over 5% of all individuals; are associated with significantly
   reduced patient QOL; and have variable interindividual presentation
   across the lifespan ([39]Bieber, 2023; [40]Frew et al, 2021). Although
   each disease has distinct risk factors, mechanisms, and clinical
   presentation, all 3 are chronic and have a poorly understood temporal
   dimension associated with disease trajectory ([41]Frew et al, 2021;
   [42]Griffiths et al, 2021; [43]Langan et al, 2020). Thus, investigating
   the age-related dynamics of inflammatory skin diseases may improve the
   understanding of underlying pathophysiology and improve prediction of
   disease trajectory on an individual patient basis.

   Computational “clock” algorithms trained using large human datasets can
   accurately predict chronological and biological age and health span on
   the basis of genomic data ([44]Horvath and Raj, 2018). Deviation of
   predicted age from true age (“age acceleration”) based on DNA
   methylation profiles has been associated with numerous inflammatory
   diseases, including HS, asthma, and psoriasis ([45]Jeremian et al,
   2024; [46]Lukac et al, 2023; [47]Peng et al, 2019). Gene expression,
   such as DNA methylation, also changes with age; however, few studies to
   date have assessed age acceleration on the basis of the more recent
   “transcriptional clocks.” Such algorithms use various signatures of
   gene expression, including genes highly correlated with or
   differentially expressed with older age as well as those found to be
   related to the aging process, to predict age. Seven transcriptional
   clock signatures have recently been defined and aggregated into a
   multisignature transcriptional age estimator RNAAgeCalc ([48]Ren and
   Kuan, 2020). Exploring transcriptional clock deviation in inflammatory
   skin disease may shed light on whether these conditions are associated
   with accelerated aging. To this end, we investigated the
   transcriptional aging signatures in existing datasets of HS, AD, and
   PP.

Results

Transcriptional clocks algorithms reveal moderate-to-strong correlations
between chronological and transcriptional age in inflammatory skin disease

   In this study, we investigated the transcriptional signatures of aging
   in skin transcriptomes from patients with AD, those with HS, and those
   with PP, and healthy subjects using 7 transcriptional aging “clock”
   algorithms. We observed moderate-to-strong correlation between
   chronological and transcriptional age for 5 of 7 clocks (DESeq2,
   Pearson, Dev, GTExAge, and Peters) in lesional and nonlesional skin of
   patients with AD (ρ = 0.40–0.86, P < .05), those with HS (2 datasets, ρ
   = 0.46–0.74, P < .05), those with PP (ρ = 0.50–0.80, P < .05), and
   healthy subjects (ρ = 0.32–0.60, P < .05). Despite pan-clock
   significant correlation among the AD, PP, and healthy subject groups,
   age prediction was notably less accurate in lesional (7 of 10
   nonsignificant comparisons [5 clocks × 2 datasets]) and, to a lesser
   degree, nonlesional (1 of 10 nonsignificant comparisons) skin of
   patients with HS. This effect was predominantly observed in lesional
   skin of patients with HS in both datasets and highlights the clinical
   and transcriptional variability among patients with HS. Nevertheless,
   these findings (summarized in [49]Table 1 and [50]Figure 1 and
   [51]Supplementary Sheet S1) broadly demonstrate the accuracy (strength
   of correlation) of transcriptional clock algorithms in predicting age
   in healthy and inflamed skin. No significant correlations were found in
   any groups per the deMagalhaes and GenAge algorithms, and these
   signatures were omitted from further analyses.

Table 1.

   Transcriptional Age Shows Moderate-to-Strong Correlation with
   Chronological Age across 3 Inflammatory Skin Diseases and 5
   Transcriptional Age Estimators
   RNAge Clocks DESeq2 Pearson Dev GTExAge Peters
   Correlation of chronological age with transcriptional age
   Healthy subjects 0.53∗∗ 0.48∗∗ 0.53∗∗ 0.60∗∗ 0.32∗
   Atopic dermatitis 0.41∗ | 0.78∗∗∗ 0.75∗∗∗ | 0.86∗∗∗ 0.40∗ | 0.57∗∗
   0.54∗∗ | 0.75∗∗∗ 0.51∗ | 0.42∗
   Hidradenitis suppurativa (I) 0.21 | 0.74∗∗ 0.16 | 0.69∗∗ 0.16 | 0.49∗
   0.075 | 0.54∗ 0.31 | 0.52∗
   Hidradenitis suppurativa (II) 0.62∗ | 0.68∗∗∗ 0.47∗ | 0.73∗∗∗ 0.42 |
   0,29 0.36 | 0.49∗ 0.50∗ | 0.46∗
   Plaque psoriasis 0.70∗∗∗ | 0.80∗∗∗ 0.75∗∗∗ | 0.69∗∗∗ 0.50∗∗ | 0.51∗∗
   0.59∗∗ | 0.70∗∗ 0.57∗∗ | 0.50∗∗
   Transcriptional age acceleration in inflammatory skin disease
   Disease versus healthy subjects
   Atopic dermatitis +9.8∗∗∗ (6.2–13.3)
   t = 5.7∗∗∗ (1.75) +9.4∗∗∗ (6.4–12.4)
   t = 5.9∗∗∗ (1.6) +3.9∗ (1.0–6.7)
   t = 2.8∗∗ (1.4) −0.5 (−3.6 to 2.6)
   t = −0.13 (1.6) 0.0 (−3.0 to 3.0)
   t = −0.22 (1.4)
   Hidradenitis suppurativa (I) +2.7 (−0.6 to 5.9)
   t = 1.8 (1.9) +5.7∗∗ (2.5–8.9)
   t = 2.8∗∗ (1.9) −0.5 (−4.2 to 3.1)
   t = −0.6 (1.8) +0.9 (−2.9 to 4.8)
   t = −0.08 (2.0) +5.6 (0.3–11.0)
   t = 1.1 (2.2)
   Hidradenitis suppurativa (II) +3.4 (0.2–7.0)
   t = 2.5∗ (2.0) +6.1∗∗ (2.7–9.4)
   t = 4.1∗∗ (1.9) +5.0∗ (1.7 to 8.4)
   t = 3.1∗∗ (1.8) +1.5 (1.7–4.8)
   t = 1.3 (1.8) +2.2 (0.6–5.0)
   t = 1.9 (1.5)
   Plaque psoriasis +2.5 (0.7–5.6)
   t = 1.5 (1.7) +7.4∗∗∗ (4.4–10.4)
   t = 5.1∗∗∗ (1.6) +12.5∗∗∗ (9.2 to 15.7)
   t = 8.0∗∗∗ (1.6) −2.8 (−6.0 to 0.4)
   t = −1.4 (1.6) +6.5∗∗∗ (3.8–9.2)
   t = 5.6∗∗∗ (1.3)
   Lesional versus nonlesional skin
   Atopic dermatitis +3.6 (−0.2 to 7.4)
   t = 2.2∗ (2.1) +3.2 (−0.2 to 6.6)
   t = 2.1∗ (1.7) +2.6 (−0.6 to 5.8)
   t = 1.5 (1.7) −1.7 (−5.2 to 1.8)
   t = −1.2 (1.7) −0.2 (−3.4 to 3.0)
   t = −0.7 (1.6)
   Hidradenitis suppurativa (I) −2.2 (−6.4 to 2.0)
   t = −1.2 (1.9) +2.4 (−1.7 to 6.5)
   t = 1.3 (1.9) −8.7∗∗ (−13.4 to −3.9)
   t = −4.0∗∗ (2.2) +3.5 (−0.4 to 7.4)
   t = 2.0 (1.8) +8.4∗∗ (3.3–13.5)
   t = 3.6∗∗ (2.3)
   Hidradenitis suppurativa (II) −0.7 (−4.8 to 3.3)
   t = −0.5 (1.0) +2.7 (−1.0 to 6.4)
   t = 3.2∗∗ (0.9) +2.9 (−1.2 to 7.0)
   t = 2.2∗ (1.4) +3.1 (−0.6 to 6.8)
   t = 2.9∗∗ (1.2) +0.6 (−3.0 to 4.2)
   t = 0.2 (1.3)
   Plaque psoriasis +0.4 (−2.7 to 3.5)
   t = 0.08 (1.5) +5.0∗∗ (2.0–8.0)
   t = 3.5∗∗ (1.4) +10.5∗∗∗ (7.5–13.5)
   t = 6.9 (1.5) −4.9∗ (−1.6 to −8.2)
   t = −3.1∗∗ (1.6) +3.0∗ (0.9–5.1)
   t = 3.0∗∗ (1.0)
   [52]Open in a new tab

   Abbreviations: CI, confidence interval; ns, not significant.

   Within disease groups, lesional skin is represented on the left, and
   nonlesional skin is on the right, separated by a vertical bar.
   Transcriptional age is accelerated in lesional versus in healthy skin
   and in lesional versus nonlesional skin within each disease group.
   Correlation is represented by Spearman’s rho. Age acceleration is
   represented in years (+/−), with significance determined by groupwise
   t-tests (95% CI) and multiple linear regression testing (t-value,
   [standard error]) of AgeAccelResid from each algorithm, adjusted for
   multiple comparisons using the false discovery rate method. Significant
   values are shown in bold. Paired analyses were implemented to assess
   lesional versus nonlesional skin within each disease group. Analyses of
   the hidradenitis suppurativa group were conducted separately on
   datasets [53]GSE151243 (I) and [54]GSE213761 (II).

Figure 1.

   [55]Figure 1
   [56]Open in a new tab

   Spearman correlation of chronological and transcriptional age. We
   observed moderate-to-strong association between chronological and
   transcriptional age in lesional and nonlesional skin of patients with
   AD (red), HS (yellow), and PP (green) as well as in skin from healthy
   subjects (gray), calculated using 5 transcriptional clock algorithms
   (DESeq2, Pearson, Dev, GTExAge, and Peters). Correlations are
   represented by Spearman’s rho (95% confidence intervals are in shaded
   color). AD, atopic dermatitis; HS, hidradenitis suppurative; PP, plaque
   psoriasis.

Transcriptional age is accelerated in diseased versus healthy skin and in
lesional versus nonlesional skin

   We next assessed for transcriptional age acceleration using groupwise
   multiple t-tests assessing differences in the mean age acceleration
   residual (AgeAccelResid) between both lesional and nonlesional patient
   skin and healthy skin as well as between lesional and nonlesional skin
   within each disease group. This was supplemented with multiple linear
   regression analyses using AgeAccelResid as the dependent variable,
   adjusted for age, sex, and all available patient demographic factors.
   Compared with that in healthy subjects, transcriptional age was
   accelerated in the lesional skin of patients with AD (+3.9∼9.8, 95%
   confidence interval [CI] = 1.0–13.3 years [3 clocks], t = 2.8∼5.7 [3
   clocks], P < .05), those with HS (+5.0∼6.1, 95% CI = 1.7–9.4 years [3
   clocks], t = 2.5∼4.1 [2 clocks], P < .05), and those with PP
   (+6.5∼12.5, 95% CI = 3.8–15.7 years [3 clocks], t = 5.1∼8.0 [3 clocks],
   P < .01) ([57]Table 1).

   Comparing nonlesional skin of patients with healthy skin revealed mixed
   findings and fewer significant differences in transcriptional aging. In
   patients with AD, linear regression revealed age acceleration in the
   lesional versus nonlesional skin (t = 2.1∼2.2 [2 clocks], P < .05). In
   the nonadalimumab-treated HS dataset ([58]GSE151243), nonlesional skin
   showed accelerated aging (+8.1, 95% CI = 4.1–12.1 years, P < .05) per
   the Dev algorithm and decelerated aging (−8.4, 95% CI = 3.3–13.5 years)
   per the Peters algorithm. By contrast, nonlesional skin in the
   adalimumab-treated HS group ([59]GSE213761) showed age deceleration per
   regression modeling (t = −2.2∼3.2 [3 clocks], P < .05) but not per the
   t-test method ([60]Table 1).

   We observed mixed findings in comparing lesional with nonlesional skin
   within each disease group. In AD, lesional skin was correlated with
   accelerated transcriptional age, per regression modeling but not the
   t-test method (t = 2.1∼2.2 [2 clocks], P < .05). In the untreated HS
   group, transcriptional age was accelerated in lesional skin per the
   Peters (+8.4, 95% CI = 3.3–13.5 years, P < .01) and Pearson (t = 2.8, P
   < .01) signatures and decelerated per the Dev signature (−8.7, 95% CI =
   −13.4 to −3.9 years, P < .01), whereas in the treated HS group,
   lesional skin showed transcriptional acceleration per regression
   modeling only (t = 2.2∼3.2 [3 clocks], P < .05). Similarly, lesional
   skin in patients with PP showed accelerated transcriptional age per
   both the t-test method (+3.0∼10.5, 95% CI = 0.9–13.5 years [3 clocks],
   P < .05) and regression modeling (t = 3.0∼6.9 [4 clocks], P < .05) and
   decelerated age per the GTExAge signature (−4.9, 95% CI = −1.6 to −8.2
   years, P < .05) ([61]Table 1).

Transcriptional age acceleration is associated with clinical and demographic
patient factors

   We next assessed for the effects of clinical and demographic patient
   factors on age acceleration by conducting multiple paired linear
   regression analyses of lesional versus nonlesional skin within each
   disease group, incorporating all available patient data (described in
   Materials and Methods). Among patients with AD, 3 loss-of-function
   sequence variants in the FLG gene were associated with transcriptional
   age acceleration per the Peters clock, including FLGR244X/2282del4 dual
   carrier status (t = 2.3, P < .05) and FLGR501X carrier status (t = 2.6,
   P < .05), whereas the absence of keratosis pilaris among patients was
   associated with decelerated transcriptional age per the same signature
   (t = −2.2, P < .05). No significant correlations were observed when
   comparing lesional with nonlesional skin among untreated patients with
   HS; however, in comparing both lesional and nonlesional skin from this
   cohort with healthy skin, White European race was identified as a
   factor associated with decelerated transcriptional age per the Peters
   clock (t = −2.2 to −3.2, P < .05). Among the treated HS group,
   accelerated age was negatively correlated with total number of nodules
   (t = −2.1 to −2.7 [2 clocks], P < .05) and fistulae (t = −2.4 to −2.6
   [2 clocks], P < .05) and positively correlated with male sex (t = 2.6
   [DESeq2], P < .05) and, surprisingly, nonsmoking status (t = 2.9
   [DESeq2], P < .05), whereas decelerated age was associated with the
   “low” (vs “medium” and “high”) abscess group (t = −2.5 [GTExAge], P <
   .05). Among patients with HS, no significant differences before and
   after adalimumab treatment were observed. Moreover, no significant
   associations between transcriptional age and patient factors were
   observed in PP.

Transcriptional clock algorithms are enriched for signatures associated with
inflammatory skin diseases, aging, and metabolism

   Our findings were reinforced by gene–disease analysis, which revealed
   128 and 220 genes across the GTEx and Peters clocks, respectively, to
   be associated with at least 1 disease. Among these, 21 genes were found
   in both algorithms, and 321 were associated with either AD or PP.
   Moreover, several clock genes were associated with all 3 diseases,
   including IFNG, TGFB1, IL12RB1, NLRP3, NOD2, and TNF. Among those
   associated with HS are 2 genes implicated in its pathophysiology
   (NCSTN, PSEN1, APH1B) ([62]van Straalen et al, 2022). Moreover, several
   disease-associated clock genes are known to be key regulators of
   metabolism (IGF2, MTOR, MAPK14) and aging (PARP1, FOXO1), summarized in
   [63]Figure 2 and [64]Supplementary Sheets S2 and [65]S3.

Figure 2.

   [66]Figure 2
   [67]Open in a new tab

   Transcriptional age acceleration is associated with expression changes
   to inflammatory skin disease–associated genes and biological pathways
   related to proinflammatory signaling, metabolism, and aging.
   Association between genes (left hand side) underlying 2 transcriptional
   clocks and HS, AD, and PP was analyzed. Color denotes strength of GDA
   score for the GTEx (blue) and clocks Peters (red). Asterisks (∗)
   indicate that the gene is also found in the opposing clock. AD, atopic
   dermatitis; GDA, gene–disease association; HS, hidradenitis
   suppurative; PP, plaque psoriasis.

   Pathway enrichment analysis of clock-associated genes showed
   significant enrichment for pathways relevant to the pathophysiology of
   all 3 diseases. These include AD (T helper 1 and T helper 2 cell
   differentiation [hsa04658] and IL-4–mediated signaling events
   [AgingReG]), PP (IL-23–mediated signaling events [AgingReG] and
   IL-22–induced cell proliferation in psoriasis [Elsevier Pathway
   Collection]), and HS (T helper 17 cell differentiation [hsa04659] and
   B-cell receptor signaling [hsa04662] and activation [Gene Ontology
   [GO]: 0042180]) as well as pathways relevant to all 3 diseases (T-cell
   differentiation [GO: 0030217] and activation [GO: 0042110], NF-kB
   signaling [hsa04064, GO :0038061], ceramide signaling pathway
   [AgingReG], and TNF receptor signaling pathway [AgingReG]). This was
   supported by significant association of clock genes with inflammatory
   diseases from the Elsevier database, including AD, vitiligo, psoriatic
   arthritis, and systemic lupus erythematosus ([68]Figures 3 and [69]4
   and [70]Supplementary Sheets S4 and [71]S5). Both clocks also
   incorporate the transcriptional state of several druggable targets,
   including members of Jak–signal transducer and activator of
   transcription signaling pathway (Jak1, JakMIP1, TYK2, STAT1, STAT2,
   STAT3); ILs and their receptors associated with HS and psoriasis
   (IL17D, IL12RB1), AD and psoriasis (IL15, ITK), among others (IL1R2,
   IL7R, IL10RA, IL10RB, IL18BP, IL27RA); several C-X-C, C-C, and C-X3-C
   motif chemokine receptors and ligands (CXCR4, CXCR5, CXCR6, CXCR7,
   CXCL9, CXCL12, CXCL14, CXCL16, CCL5, CX3CR1); PP-associated major
   histocompatibility complex genes (HLA-B, HLA-C); key cytokine gene TNF;
   and several TGFβ-associated genes (TGFB1, TGFBR2, TGFBR3, TGFBRAP1)
   ([72]Tsai and Tsai, 2022; [73]van Straalen et al, 2022). In addition to
   inflammation-associated processes, clock genes were represented among
   pathways related to metabolism and cell signaling (protein kinase
   B–mTOR signaling [AgingReG], TP53-regulated metabolic genes [AgingReG],
   androgen receptor [AgingReG], epidermal GF/EGFR signaling pathway
   [WikiPathways], energy metabolism [WikiPathways]) and aging (NAD
   metabolism sirtuins and aging [WP3630], SIRT7 signaling in aging
   [Elsevier Pathway Collection], senescence-associated secretory
   phenotype [AgingReG], regulation of telomerase [AgingReG], and
   sphingolipid metabolism in senescence [WikiPathways]). Moreover, 41
   genes from the GTEx and Peters clocks overlapped with those implicated
   in the “unhealthy skin signature.” These findings further demonstrate
   that dysregulation of processes related to inflammation, cell
   signaling, metabolism, and loss of homeostasis as captured by
   transcriptional age acceleration may be linked to inflammatory skin
   disease ([74]Figure 4 and [75]Supplementary Sheets S2 and [76]S4–6).

Figure 3.

   [77]Figure 3
   [78]Open in a new tab

   Pathway and disease enrichment analysis based on the GTEx and Peters
   transcriptional age algorithm genes. Pathway databases include the
   Elsevier Pathway Collection (top), GO (biological process) (middle),
   and KEGG (bottom), showing enrichment for pathways associated with
   common inflammatory skin diseases; broad-scale immune signaling; and
   immune cell activation, regulation, and function. Pathways and
   functions are ranked by their fold enrichment, with color denoting
   statistical significance and the size of each dot denoting the number
   of genes associated with that pathway/function. FDR, false discovery
   rate; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Figure 4.

   [79]Figure 4
   [80]Open in a new tab

   Extended molecular pathway enrichment analysis based on the GTEx and
   Peters transcriptional age algorithm genes. Additional pathway analysis
   based on the GTEx and Peters genes using the AgingReG (top, does not
   capture fold enrichment) and WikiPathways (bottom) repositories
   demonstrates enrichment for pathways linked to inflammation, cell
   signaling, metabolic function, and aging. Pathways and functions are
   ranked by their fold enrichment, with color denoting statistical
   significance and the size of each dot denoting the number of genes
   associated with that pathway/function. Asterisks indicate that the
   enriched pathway is based on analysis from the Peters (∗) and GTEx (∗∗)
   algorithms or both (∗∗∗). Complete list of identified genes and
   pathways is provided in Supplementary Sheets S2, S4, and S5.
   FDR-adjusted p-values are presented. FDR, false discovery rate.

Discussion

   This study provides early evidence of transcriptional age acceleration
   in HS, AD, and PP on the basis of case–control and lesional–healthy
   skin comparisons and highlights the potential usefulness of
   transcriptional aging estimation algorithms in investigating the
   dynamic presentation of inflammatory skin diseases.

   We found that 5 of 7 algorithms contained in the RNAAgeCalc package
   reliably correlated transcriptional to chronological age, demonstrating
   their overall accuracy in predicting age in lesional, nonlesional, and
   healthy skin. These 5 clocks (DESeq2, Pearson, Dev, GTExAge, and
   Peters) were those that included genes that either correlated with age
   or were differentially expressed with age but did not include specific
   age-related genes. This suggests a possible indirect effect of these
   diseases on aging. We found that all signatures, except Dev, were
   reasonably accurate in predicting age among all tested groups. As
   expected, nonlesional skin of patients typically showed stronger
   correlation with age than lesional skin. This was particularly evident
   in the 2 HS datasets, which demonstrated the poorest correlation with
   age in both lesional and nonlesional skin, likely owing to confounding
   comorbid conditions, including other inflammatory and metabolic
   disorders often found in HS.

   Importantly, we found evidence of age acceleration in both disease and
   healthy skin comparisons as well as lesional and nonlesional
   comparisons within each disease group ([81]Table 1), although
   acceleration was not uniform across all algorithms. Furthermore, both
   Pearson and Dev signatures predicted age acceleration across the
   greatest number of group comparisons and in all 3 diseases studied.
   Interestingly, comparing nonlesional with healthy skin revealed far
   fewer significant differences in age acceleration than analyses of
   lesional skin, with nonlesional skin showing decelerated age in both AD
   and 1 cohort of HS. Notably, we observed that the carrier status of
   patients with AD for 3 FLG loss-of-function sequence variants that
   collectively confer >3-fold risk of AD (FLGR244X/2282del4 and FLGR501X)
   was positively associated with age acceleration, whereas absence of
   keratosis pilaris showed a negative association, per the Peters
   algorithm. This is particularly compelling because this aging signature
   does not contain FLG in its estimation, suggesting that the deleterious
   effects of FLG downregulation and skin barrier loss on disease severity
   are reflected in the dysregulated expression of aging- and
   immune-related genes ([82]Gao et al, 2009; [83]Greisenegger et al,
   2010; [84]Traidl et al, 2021).

   Examination of pathways enriched within clock-associated genes
   highlighted pathways associated with aging, cell signaling, metabolism,
   and inflammation ([85]Figure 3, [86]Figure 4). These interestingly also
   included druggable targets across diseases. Indeed, work by other
   groups have demonstrated a similar dysregulation in the metabolic and
   inflammatory milieu of AD ([87]Quaade et al, 2025; [88]Sääf et al,
   2008), PP ([89]Gniadecki et al, 2023), and HS ([90]de Oliveira et al,
   2022; [91]Suh-Yun Joh et al, 2023).

   Taken together, our analyses strengthen the hypothesis that
   transcriptional age acceleration may be a marker of the inflammatory
   and metabolic burden of disease. This is further supported by our
   findings of less prevalent age acceleration effects and even age
   deceleration in one analysis of AD when comparing nonlesional with
   healthy skin as well as positive association between major genetic risk
   factors for AD and age acceleration. These findings are associative and
   not causative and necessitate further investigation using multiple
   validation cohorts from diverse populations to determine what, if any,
   relationship transcriptional age acceleration has with disease; whether
   it is simply a reflection of disease-associated transcriptional state
   or correlates with pathophysiology; and to what degree this value is
   influenced by environmental factors and other confounders. Further
   studies should also assess the potential for these algorithms to
   capture multiple dimensions of each disorder on an intraindividual
   level; including progression, symptom severity, and treatment response
   would be beneficial to truly appreciate the effect and directionality
   of disease dynamics with changes in transcriptional age.

   Several limitations to our study exist, including the use of public
   datasets, which may be confounded by intercohort differences in data
   generation and processing, and unknown patient comorbidities, which we
   have done our best to minimize by conducting most analyses in a within
   dataset fashion. Our analyses capture a single time point in disease,
   and it would be important to determine whether these age-related
   changes persist or change over time or, indeed, whether they might be
   reversible with treatment or disease remission. Finally, although the
   RNAAgeCalc-based algorithms were initially trained on largely White
   American and White European patient samples, its authors note
   variations in age prediction with different ancestry groups, even when
   using precalculated signatures trained on multiracial participant data
   ([92]Ren and Kuan, 2020). Although most patients investigated in this
   study were of White European and White American origin, mixed
   populations in both the design of the aging algorithms and in one of
   our HS disease cohorts may skew age estimates.

   A major point of consideration remains the interpretation of whether
   these differentially expressed genes represent accelerated aging or
   rather also reflect alternative regulatory programming in response to
   disease state because much of our understanding of molecular clocks
   (predominantly DNA methylation–based clocks) to date relies on observed
   correlations. Ongoing work in the field continues to resolve the
   question of aging- versus disease-related signatures. [93]Kabacik et al
   (2022) found that epigenetic aging was associated with several
   hallmarks of aging, including nutrient sensing, mitochondrial activity,
   and stem cell composition, but was distinct from other hallmarks,
   including telomere attrition, cellular senescence, and genomic
   instability. They suggest that aging results from pathways that include
   both those associated with epigenetic aging and those likely resulting
   from “wear and tear.” [94]Teschendorff and Horvath (2025) recently
   reviewed our current understanding of epigenetic clocks and highlighted
   the need to use emerging technologies such as single-cell RNA
   sequencing to build tissue type–specific and cell type–specific DNA
   methylation references to improve existing epigenetic (and, by