Anti-Tumor Necrosis Factor Drug Responses and Skin-Blood DNA Methylation Age: Relationships in Moderate-to-Severe Psoriasis

Abstract

Studies have examined the utility of DNA methylation as a biomarker of psoriasis treatment responses, but investigations of treatment responses with Skin-Blood DNA methylation age (SkinBloodAge) – a methylation-based measure of health designed using skin tissues – are lacking. Using a HumanMethylation450 BeadChip blood DNA methylation dataset from 70 white patients who presented with moderate-to-severe plaque psoriasis and were treated with anti-tumor necrosis factor (TNF) agents in Madrid, Spain, we examined the cross-sectional relationships of SkinBloodAge with anti-TNF treatment responses. Partial responders had a 7.2-year higher mean SkinBloodAge than excellent responders (p=0.03). In linear regression models adjusted for chronological age, sex, and anti-TNF agents, on average, partial responders had a 2.65-year higher SkinBloodAge than excellent responders (95%CI: 0.44, 4.86, p=0.02). This relationship was attenuated in a sensitivity analysis adjusting for white blood cells including known T cell mediators of psoriasis pathophysiology (β=1.91-years, 95%CI: −0.50, 4.32, p=0.12). Overall, our study suggests that partial responders to anti-TNF therapy have higher SkinBloodAges when compared to excellent responders. Although these findings still need to be confirmed more broadly, they further suggest that SkinBloodAge may be a useful treatment response biomarker that can be incorporated with other blood tests before anti-TNF therapy initiation in moderate-to-severe psoriasis patients.

1 |. Background

Of the 125 million people living with psoriasis worldwide, 20–32% will likely experience a loss of response to biologic therapy (1). The marked probability of treatment failure in psoriasis further exacerbates the physical/psychosocial toll that the disease has on patients and is linked to increased direct medical costs (2). Further compounding the problem, existing biomarkers of treatment response have been non-specific and suboptimal in directing therapies (3). DNA methylation is a biological process that impacts DNA expression via the transfer of methyl groups (-CH₃) to cytosine nucleotides on DNA, and has been implicated in the pathogenesis of psoriasis including the activation and proliferation of keratinocytes (4,5). Additional DNA methylation studies uncovering inflammatory pathways known to be involved in psoriasis and demonstrating DNA methylation sex differences in psoriasis physiology contribute to the proposition that DNA methylation can also serve as a novel psoriasis treatment response biomarker (6–8).

More recently, DNA methylation/epigenetic measures of age (also called clocks) have provided new opportunities to examine DNA methylation relationships (9). Depending on the specific metric, DNA methylation age measures can be used as predictors of chronological age, predictors of disease risk, and/or predictors of mortality risk (10,11). Although many human population studies have examined the relationships of these measures with a host of disease processes, studies examining psoriasis relationships remain scarce. One prior study reported a null association of psoriasis with the 353-CpG DNA methylation age metric (DNAmAge), which is primarily considered to be a predictor of chronological age (12). Nevertheless, to the best of our knowledge, no studies have evaluated psoriasis relationships with the methylation/epigenetic clock for blood and skin (SkinBloodAge) (13). SkinBloodAge was specifically created after studies demonstrated suboptimal performance of DNAmAge in fibroblasts and other tissues that are readily available from skin biopsies. Studies have since shown that SkinBloodAge can detect age acceleration, that was previously undetectable with other DNA methylation age measures, in skin samples from patients with diseases like Hutchinson Gilford Progeria Syndrome (13). Importantly, SkinBloodAge maintains high accuracy in blood cells and provides an opportunity to perform blood studies that may have increased relevance to skin pathologies (13). Given that a number of screening blood tests (e.g. complete blood counts, metabolic panels, hepatitis serologies) already must be performed to first determine if psoriasis patients are appropriate candidates for biologic therapies like anti-tumor necrosis factor (TNF) agents, a blood biomarker able to predict treatment response would be of great utility and could be incorporated into existing physician workflows (14).

1.1 |. Questions Addressed

In this study, we examined the relationship of SkinBloodAge with anti-TNF treatment responses in moderate-to-severe psoriasis patients (13). We also performed sensitivity analyses examining anti-TNF treatment response relationships with other DNA methylation-based aging markers not specifically created using skin tissues.

2 |. Experimental Design

2.1 |. Study Sample

This cross-sectional analysis was conducted using a NCBI GEO HumanMethylation450 BeadChip dataset (Series GSE151278, uploaded to “https://www.ncbi.nlm.nih.gov/geo” on May 27, 2020) initially collected for a previously published epigenetic-wide association study (15). The dataset is comprised of peripheral blood DNA methylation samples from 70 White patients who presented with moderate-to-severe plaque psoriasis and were treated with an anti-TNF agent (Adalimumab, Etanercept, or Infliximab) at the Hospital Universitario de La Princesa in Madrid, Spain. Drug responses were categorized based on the Psoriasis Area and Severity Index (PASI). Patients were categorized as “excellent responders” if they achieved at least a 90% improvement in PASI at 3 and 6 months of therapy, while “partial responders” achieved less than a 70% PASI improvement in the same period. Demographic characteristics of each participant included chronological age (continuous [years]), chronological age of the first drug therapy for psoriasis (continuous [years]), and sex (dichotomous [male or female]) was also provided for each participant. Informed consent and ethical conduct of study information regarding the collection of this data for all participants can be found in the initial publication (15).

2.2 |. DNA Methylation Data and DNA Methylation Age Measures

Methylation idat files were downloaded from NCBI. Full descriptions of DNA processing for the downloaded dataset can be found in the initial publication (15). SWAN normalization was performed using the “minfi” R package. DNA methylation data was uploaded to Horvath’s publicly available online calculator (http://dnamage.genetics.ucla.edu). The online calculator provided downloadable output which included values for SkinBloodAge, DNAmAge, PhenoAge, and GrimAge. Pace of Aging (PoA) was calculated from the same DNA methylation dataset using R code available at https://github.com/danbelsky/DunedinPoAm38.git. DNAmAge was evaluated because it was used in a prior psoriasis study (12). PhenoAge was selected because it is the most robust general methylation predictor of health status described in the literature (16). GrimAge was selected because it is the most robust methylation predictor of mortality described in the literature (11). Finally, PoA was selected because it is the most novel methylation measure of the pace of biological aging (17).

2.3 |. Statistical Analysis

We used linear regression models adjusted for chronological age, chronological age of first psoriasis drug initiation, current anti-TNF agent, and sex to determine associations of anti-TNF treatment responses with SkinBloodAge. Given the known role of the immune response and white blood cells as a mediator in psoriasis pathology, we did not adjust for white blood cell proportions in our main analysis models (18). All covariates were determined a priori based on available data and the previous study using this dataset (15). We next performed sensitivity analyses, using all of the same covariates as the primary analysis, evaluating the associations of treatment response with four other methylation-based aging markers: DNAmAge, PhenoAge, GrimAge, and PoA. We performed an additional sensitivity analysis repeating all analyses with additional adjustments for white blood cell proportions (CD8 naïve T cells, CD8pCD28nCD45Ra- T cells, plasma blasts, CD4T cells, NK cells, monocytes, and granulocytes) (19,20).

All statistical analyses were performed using R Version 3.6.3 (R Core Team, Vienna, Austria). We considered a P-value < 0.05 to be statistically significant.

3 |. Results

Descriptive statistics for the study participants are presented in Table 1. The majority of participants were male (61%) and were taking Etanercept (38%). Participants had a mean (SD) chronological age, SkinBloodAge, DNAmAge, PhenoAge, GrimAge, and PoA of 47.1 (14.6), 45.4 (13.0), 41.3 (13.6), 41.3 (15.0), 56.6 (12.5), and 0.89 (0.09) years, respectively. On average, partial responders had a statistically significant higher SkinBloodAge (+7.2 years), PhenoAge (+9.5 years), and Pace of Aging (0.06 years). Chronological age (+5.6 years), age of first drug initiation (+1.3 years) and DNAmAge (+6.6 years) for partial responders trended similarly. GrimAge for partial responders trended in the opposite direction (−4 years) (Table 1). Chronological age and SkinBloodAge were highly correlated with DNAmAge, PhenoAge, and GrimAge, but not PoA (Figure 1).

Table 1.

Descriptive Statistics of Study Sample

	All Subjects (N = 70)	Excellent Responders (N = 49)	Partial Responders (N = 21)	Pa
Age Variables
Chronological Age (years), mean (SD)	47.1 (14.6)	45.4 (14.2)	51.0 (15.0)	0.15
Chronological Age at First Drug Initiation (years), mean (SD)	25.8 (11.6)	25.4 (11.0)	26.7 (13.1)	0.70
SkinBloodAge (years), mean (SD)	45.4 (13.0)	43.4 (12.8)	50.6 (12.3)	0.03
DNAmAge (years), mean (SD)	41.3 (13.6)	39.3 (13.5)	45.9 (13.1)	0.07
PhenoAge (years), mean (SD)	41.3 (15.0)	38.5 (14.6)	48.0 (14.0)	0.01
GrimAge (years), mean (SD)	56.6 (12.5)	54.8 (13.0)	50.8 (10.5)	0.05
Pace of Aging (years), mean (SD)	0.89 (0.09)	0.87 (0.09)	0.93 (0.08)	0.02
Health Variables
Sex, N (%)				0.83
Female	27 (39)	18 (37)	9 (43)
Male	43 (61)	31 (63)	12 (57)
Drug, N (%)				0.14
Adalimumab	25 (36)	21 (43)	4 (19)
Etanercept	27 (38)	16 (33)	11 (52)
Infliximab	18 (26)	12 (24)	6 (29)

^aThe t-test and χ²-test were performed for continuous and categorical variables, respectively, comparing excellent and partial responders.

P values < 0.05 are italicized and considered statistically significant.

Figure 1 |. Pearson Correlation Matrix.

Pearson correlation matrix of age variables used in the analyses. The figure depicts correlation coefficients, scatter plots, and variable distributions. Asterisks denote statistical significance of correlations: * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

Table 2 summarizes the results of fully-adjusted linear models where participant’s anti-TNF agent treatment response was modeled as a predictor of each methylation age measure. Compared to excellent responders, partial responders had a 2.65-year higher SkinBloodAge (95%CI: 0.44, 4.86, p = 0.02). Similar trends of increased age in partial responders were observed for DNAmAge (β = 2.12-years, 95%CI: −1.04, 5.29, p = 0.19), PhenoAge (β = 4.19-years, 95%CI: −0.28, 8.66, p = 0.07), GrimAge (β = 1.55-years, 95%CI: −1.31, 4.40, p = 0.28), and PoA (β = 0.04-years, 95%CI: −0.001, 0.09, p = 0.06), but these findings did not reach statistical significance. Among the model covariates, chronological age was significantly associated with SkinBloodAge (β = 0.83-years, 95%CI: 0.74, 0.92, p < 0.001), DNAmAge (β = 0.81-years, 95%CI: 0.68, 0.94, p < 0.001), PhenoAge (β = 0.86-years, 95%CI: 0.68, 1.05, p < 0.001), and GrimAge (β = 0.78-years, 95%CI: 0.66, 0.90, p < 0.001). Chronological age was not associated with PoA. Moreover, sex, chronological age when any drug therapy for psoriasis was started, and current anti-TNF agent were not associated with any methylation age outcome. Furthermore, no covariate – including sex – significantly modified the association of treatment response with SkinBloodAge. All anti-TNF therapy relationships were attenuated and were no longer statistically significant after adjustments for white blood cell proportions (Table S1). Nevertheless, among all DNA methylation age measures, the effect estimate for SkinBloodAge remained the largest. Compared to excellent responders, partial responders had a 1.91-year higher SkinBloodAge (95%CI: −0.50, 4.32, p = 0.12) after adjusting for white blood cell proportions (Table S1).

Table 2.

Anti-TNF Drug Response and DNA Methylation Age Relationships

Modelsa	Difference in DNA Methylation Age (95% CI)	P
SkinBloodAge
Treatment Response
Excellent Responders	reference	-
Partial Responders	2.65 (0.44, 4.86)	0.02
Chronological Age	0.83 (0.74, 0.92)	<0.001
Age of First Drug	0.003 (−0.11, 0.12)	0.95
Sex
Male	reference	-
Female	0.20 (−1.85, 2.25)	0.85
Drug
Adalimumab	reference	-
Etanercept	−0.68 (−3.09, 1.73)	0.58
Infliximab	0.58 (−2.03, 3.18)	0.66
DNAmAge
Treatment Response
Excellent Responders	reference	-
Partial Responders	2.12 (−1.04, 5.29)	0.19
Chronological Age	0.81 (0.68, 0.94)	<0.001
Age of First Drug	0.03 (−0.13, 0.20)	0.69
Sex
Male	reference	-
Female	−1.06 (−4.00, 1.88)	0.47
Drug
Adalimumab	reference	-
Etanercept	−1.83 (−5.28, 1.62)	0.29
Infliximab	0.80 (−2.93, 4.53)	0.67
PhenoAge
Treatment Response
Excellent Responders	reference	-
Partial Responders	4.19 (−0.28, 8.66)	0.07
Chronological Age	0.86 (0.68, 1.05)	<0.001
Age of First Drug	−0.06 (−0.30, 0.17)	0.59
Sex
Male	reference	-
Female	−0.41 (−4.56, 3.73)	0.84
Drug
Adalimumab	reference	-
Etanercept	1.54 (−3.33, 6.41)	0.53
Infliximab	2.56 (−2.71, 7.83)	0.34
GrimAge
Treatment Response
Excellent Responders	reference	-
Partial Responders	1.55 (−1.31, 4.40)	0.28
Chronological Age	0.78 (0.66, 0.90)	<0.001
Age of First Drug	−0.01 (−0.16, 0.14)	0.90
Sex
Male	reference	-
Female	2.36 (−0.29, 5.00)	0.08
Drug
Adalimumab	reference	-
Etanercept	0.26 (−2.85, 3.37)	0.87
Infliximab	1.47 (−1.90, 4.83)	0.39
Pace of Aging
Treatment Response
Excellent Responders	reference	-
Partial Responders	0.04 (−0.001, 0.09)	0.06
Chronological Age	0.002 (−0.0002, 0.004)	0.08
Age of First Drug	−0.002 (−0.005, 0.0001)	0.06
Sex
Male	reference	-
Female	−0.01 (−0.05, 0.03)	0.69
Drug
Adalimumab	reference	-
Etanercept	−0.001 (−0.05, 0.05)	0.96
Infliximab	0.04 (−0.01, 0.09)	0.13

^aModels adjusted for chronological age, age of first drug initiation, sex, and anti-TNF drug.

P values < 0.05 are italicized and considered statistically significant in these analyses.

4 |. Conclusions and Perspectives

Our findings demonstrate significantly higher SkinBloodAges in partial responders when compared to excellent responders. Importantly, these findings are independent of chronological age, sex, and anti-TNF agents. Similar increased trends in other methylation-based metrics further suggests that there may be inherent epigenetic differences between partial and excellent responders. Such a premise is supported by existing literature reporting differential DNA methylation in psoriatic tissues as compared to normal tissues as well as research demonstrating the ability to subclassify types of psoriasis by epigenetic signatures (21,22).

We identified one prior study demonstrating no significant DNAmAge differences between psoriatic tissues from patients with psoriasis, normal tissues from patients with psoriasis, and normal tissues from healthy control subjects (12). Given that we replicate this null finding in the present analysis, but report significant findings with SkinBloodAge suggests that methylation metrics intentionally crafted with skin tissues are most sensitive for detecting methylation differences in psoriatic processes. As previously mentioned, SkinBloodAge was developed due to the suboptimal performance of other methylation-age metrics in fibroblasts (13). Using human blood cells, fibroblasts, keratinocytes, and other skin relevant tissues, the authors were able to develop a biological age metric with good performance in skin and blood tissues. For this reason, SkinBloodAge likely has a sensitivity to psoriasis even though we are measuring it in blood cells. Utilizing a cohort with moderate-to-severe disease as well as the known role of fibroblasts in psoriasis pathology (23,24), may also contribute to why our strongest observed associations are with SkinBloodAge and none of the other metrics. Still, additional studies are needed to better characterize the relationship of SkinBloodAge with other skin pathologies.

Furthermore, it is important to address the results of our models adjusting for white blood cell proportions. Given the known role of white blood cells – especially T cells – in mediating psoriasis pathophysiology, we expected adjusting for white blood cells to attenuate our results (18). As expected, the effect estimates from these analyses were attenuated and no longer demonstrated statistically significant differences between partial and excellent responders. However, of all DNA methylation age measures examined, the effect estimate for the relationship of anti-TNF treatment response with SkinBloodAge remained the largest in magnitude and further suggests a unique sensitivity of SkinBloodAge for psoriasis relationships. Again, additional studies remain necessary to further interrogate these relationships.

The strengths of the present study include the incorporation of novel methylation-based markers of aging, and the utilization of a previously published dataset. Nevertheless, this study does have some limitations. Specifically, we had limited information on important covariates including demographic, clinical (e.g. body mass index, PASI at beginning of treatment), lifestyle, and drug (e.g dosage) factors. Thus, we cannot rule out the impact of residual confounding and the aforementioned unknowns in our analyses. Furthermore, our findings are derived from a study sample of middle-age Spanish patients, and additional analyses are necessary to establish their generalizability.

In sum, our study suggests that SkinBloodAge may be a useful treatment response biomarker for moderate-to-severe psoriasis patients prior to the initiation of anti-TNF therapy. If these findings prove to be generalizable, SkinBloodAge may help facilitate a risk-adapted approach to treatment and improve patient outcomes in moderate-to-severe psoriasis.