Abstract
Background
Atopic dermatitis (AD) is characterized by Staphylococcus aureus (S. aureus) colonization. Longitudinal early life data delineating relationships of S. aureus colonization, barrier function, and AD outcomes are lacking. We define longitudinal S. aureus endotypes and AD pathogenesis in early life.
Methods
We defined longitudinal S. aureus skin colonization phenotypes across two annual visits (non-colonized: V1−V2−, early transient: V1+V2−, late onset: V1−V2+, persistent: V1+V2+) in the Mechanisms of Progression of Atopic Dermatitis to Asthma in Children cohort. We analyzed AD severity, sensitization, and skin barrier function across phenotypes, and performed mediation analyses between colonization and FLG expression.
Results
Persistent S. aureus colonization was associated with increased SCORAD at V1 (33.5 vs. 19.0, p = 0.004) and V2 (40.1 vs.16.9, p < 0.001), and lower non-lesional (NL) FLG at V2 (1.77 vs 4.09, p = 0.029) compared to the non-colonized phenotype, with early transient and late onset colonization as intermediate phenotypes. Children colonized at V2 demonstrated a decrease in NL-FLG expression from V1 to V2 compared to those non-colonized at V2 (p=0.0012), who maintained expression. This effect remained significant even after adjusting for V1 colonization and SCORAD (p = 0.011).
Conclusions
Our findings are the first to present longitudinal quantitative FLG expression and S. aureus skin colonization in early life and suggest that a decrease in NL-FLG drives later colonization. Hence, therapies to maintain NL-FLG expression may prevent S. aureus colonization. Further, a longitudinal AD endotype of persistent colonization is characterized by increased AD severity, sensitization, and decreasing NL-FLG.
Keywords: atopic dermatitis, filaggrin expression, longitudinal colonization, mediation analysis, Staphylococcus aureus
Graphical Abstract
We assessed S. aureus skin colonization and AD outcomes across two annual visits in the MPAACH cohort.
Persistent S. aureus colonization is characterized by increased severity, allergic sensitization, and decreasing never-lesional filaggrin expression.
Low never-lesional filaggrin expression positively mediates future S.aureus colonization. Maintaining filaggrin expression may be therapeutically important to halt colonization and worse AD outcomes.
Abbreviations: AD, atopic dermatitis; MPAACH, Mechanisms of Progression of Atopic Dermatitis to Asthma in Children; SA, Staphylococcus aureus; SCORAD, SCORing Atopic Dermatitis
INTRODUCTION
Atopic dermatitis (AD) is a chronic, relapsing inflammatory skin disease that affects 15–30% of children and 2–10% of adults, and often precedes the development of allergic comorbidities.1, 2 AD is characterized by a dysfunctional skin barrier, which may be associated with deficiencies in the skin structural protein filaggrin (FLG), Th2 inflammation, skin microbial dysbiosis and exposure to exogenous irritants and/or allergens.3
Dysbiosis of the skin microflora with increased colonization by the gram-positive opportunistic bacterium Staphylococcus aureus (S. aureus) and the loss of commensal skin microbes is a hallmark of AD.4 S. aureus exacerbates skin inflammation and dominates the microbiome during AD flares, thereby worsening AD severity.5 Defects in the epidermal barrier worsen inflammation and further enhance entry and penetrance of S. aureus into the dermis.6 Presence of FLG breakdown products, essential components of the natural moisturizing factor (NMF), has been shown to reduce S. aureus growth rate and cell density in vitro.7 Further, FLG mutations are associated with reduced barrier integrity and increased S. aureus colonization on lesional skin in vivo.6, 8 While this link between AD severity, FLG deficiency, and S. aureus colonization is increasingly recognized, studies to date have not examined S. aureus colonization and FLG expression over time in early life. 5, 9–11 A recently convened workgroup of thought leaders in the field recognized that defining longitudinal phenotypic and endotypic subgroups of AD is a critical need.12 Herein, we directly address this by leveraging the MPAACH (Mechanisms of Progression from AD to Asthma in Children) early-life cohort (n=326), which includes extensive longitudinal collection of clinical information and biospecimens13. We quantified skin FLG expression levels over time, defined longitudinal S. aureus skin colonization phenotypes and the impact on AD clinical outcomes, and then integrated these to explain the interplay between skin FLG and S. aureus colonization. The results from this study provide novel insights into key temporal relationships and drivers of S. aureus colonization in early life, a critical gap in current knowledge.
METHODS
Subjects
MPAACH is a prospective longitudinal early-life cohort of children with AD that has been previously described.13,14 Eligible children were aged ≤3 years upon enrollment, had a gestation of ≥36 weeks, and either a diagnosis of AD (based on the Hanifin and Rajka Criteria for Atopic Dermatitis) or the parent(s)/legal authorized representative indicated a positive response to each of the 3 questions from the Children’s Eczema Questionnaire (CEQ)15. Samples and specimens were collected at each annual visit. Children are instructed to stop using lotions, creams, and ointments on the skin the night prior to collection. This study was approved by the Institutional Review Board at CCHMC, and all subjects signed informed consent/parental permission prior to participation.
Keratinocyte Sampling and Identification of S. aureus
Lesional skin was defined by the Hanifin and Rajka diagnostic criteria for atopic dermatitis.16 Non-lesional (NL) skin was defined as at least 10 cm from lesional (L) sites and had no history of lesions at that site according to parental report. SmartSolve skin tape strips measuring 1 inch x 1 inch were taken from L and NL-skin at each visit. Briefly, skin cells were sampled by placing the tape strip on the skin, gently massaging the tape for 15–20 seconds, removing the tape, and storing it in ice-cold in BL buffer supplemented with 2% thio-glycerol (Promega, Madison, WI). This process was repeated to provide a total of 12 tape strips for each sampled skin site at each visit. RNA was extracted from tapes 8–9 using a two-stage process. The tapes were first extracted with a phenol-chloroform extraction and then the resulting RNA was purified using the ReliaPrep RNA cell mini prep kit following manufacturers guidelines (Promega). Complementary DNA was made with SuperScript IV VILO (ThermoFisher) and RT-qPCR reactions were carried out with the following Taqman gene expression assays: 18S (Hs03003631_g1), FLG (Hs00856927_g1). FLG expression levels were normalized to 18S levels. The maximum gene expression value from either tape strip 8 or 9 from each subject was used to denote FLG expression.
Sheep blood agar contact plates (Hardy Diagnostics) were used to collect host skin biome samples from L and NL-skin at each visit. Bacterial DNA was subjected to quantitative polymerase chain reaction (qPCR) targeting femA-S. aureus and Universal 16S rRNA to confirm S. aureus at either L or NL skin. Presence and absence of femA was determined based on the negative control at a Ct cutoff value of 30 on the ThermoFisher Cloud Connect online software. Phenotypes were determined by presence or absence of S. aureus by femA positive qPCR at either L or NL site at a visit. Detailed description of all methods and materials can be found in the Supporting Information.
RESULTS
S. aureus presence is associated with a more severe AD phenotype
There were 326 MPAACH subjects who had completed both V1 and V2 at the time of analysis. Bacterial isolates positive with the agglutination assay across V1 and V2 were further verified as S. aureus (SA) with femA qPCR to identify subjects with S. aureus presence or absence at each visit.
The subset of 326 subjects was initially categorized into two groups: non-colonized (SA-) and colonized (SA+) (Table 1). The SA+ group (19%) had S. aureus presence on L or NL skin at either V1 or V2 whereas the SA- group (81%) did not show S. aureus presence at V1 nor V2 on L or NL skin. At V1 and V2, the SA+ group had significantly higher SCORAD (p = 0.008, p < 0.001) and L-TEWL (p < 0.001, p = 0.004) and was more likely to have moderate/severe AD (p = 0.007, p < 0.001) and to be allergen-sensitized (p = 0.035, p = 0.027) compared to the SA- group. In addition, both L-FLG (p = 0.022) and NL-FLG (p < 0.001) expression at V2, but not V1 L or NL-FLG expression, were significantly lower in colonized subjects (Table 1). There were no baseline differences in age, sex, race, parental reported antibiotic usage in the past 12 months, steroid cream usage, early onset AD and environmental exposures between the two groups.
Table 1.
Characteristics of MPAACH cohort by S. aureus colonization†
| S. aureus colonization | Sample size for each variable | Non-colonized | Colonized | p-value‡ |
|---|---|---|---|---|
| n = 264 (81.0) | n = 62 (19.0) | |||
| Demographic | ||||
| Age (y) | 326 | 2.1 (1.5–2.4) | 1.9 (1.5–2.4) | 0.42 |
| Male sex | 326 | 137 (52) | 35 (56) | 0.52 |
| Black race | 326 | 180 (68) | 35 (56) | 0.08 |
| Early onset AD (<3 mos of age) | 326 | 111 (42) | 31 (50) | 0.25 |
| AD Severity at Visit 1 | ||||
| SCORAD (score) | 326 | 19.0 (13.0–29.1) | 27.9 (12.7–41.5) | 0.008 |
| Severity group | 326 | 0.007 | ||
| Mild (SCORAD < 25) | 172 (65) | 29 (47) | ||
| Moderate/Severe (SCORAD ≥ 25) | 92 (35) | 33 (53) | ||
| AD Severity at Visit 2 | ||||
| SCORAD (score) | 324 | 16.9 (9.6–24.7) | 26.2 (11.2–42.9) | < 0.001 |
| Severity group | 324 | < 0.001 | ||
| Mild (SCORAD < 25) | 198 (75) | 27 (43) | ||
| Moderate/Severe (SCORAD ≥ 25) | 64 (24) | 35 (57) | ||
| Skin barrier function at Visit 1 | ||||
| Lesional skin | ||||
| Keratinocyte FLG expression§ | 267 | 2.36 (0.68–6.05) | 1.95 (0.54–5.14) | 0.56 |
| TEWL (g/m2/h) | 324 | 11.5 (8.4–18.3) | 16.7 (11.3–28.9) | < 0.001 |
| Non-lesional skin | ||||
| Keratinocyte FLG expression§ | 268 | 3.97 (1.21–8.83) | 3.31 (1.72–7.79) | 0.76 |
| TEWL (g/m2/h) | 323 | 9.4 (7.1–13.2) | 9.6 (8.0–12.9) | 0.32 |
| Skin barrier function at Visit 2 | ||||
| Lesional skin | ||||
| Keratinocyte FLG expression§ | 241 | 2.89 (1.12–8.37) | 1.13 (0.30–4.45) | 0.022 |
| TEWL (g/m2/h) | 314 | 11.7 (8.2–16.2) | 17.9 (8.8–24.2) | 0.004 |
| Non-lesional skin | ||||
| Keratinocyte FLG expression§ | 242 | 4.09 (1.48–9.00) | 1.65 (0.48–4.48) | < 0.001 |
| TEWL (g/m2/h) | 312 | 9.2 (7.1–11.5) | 9.7 (7.6–13.6) | 0.17 |
| Sensitization at Visit 1 | ||||
| Sensitization group | 326 | 0.035 | ||
| Nonsensitized | 154 (58) | 27 (43) | ||
| Sensitized | 110 (42) | 35 (57) | ||
| Total serum IgE | 236 | 32 (10–93) | 46 (13–228) | 0.13 |
| Sensitization at Visit 2 | ||||
| Sensitization group | 323 | 0.027 | ||
| Nonsensitized | 150 (57) | 26 (42) | ||
| Sensitized | 111 (42) | 36 (58) | ||
| Total serum IgE | 204 | 70 (18–142) | 118 (30–408) | 0.053 |
Data are presented in median (IQR) or n (%).
Comparison between S. aureus colonization groups were done using Wilcoxon rank-sum test for continuous variables, Chi-square tests for categorical variables.
Season-adjusted FLG expression values were used in the comparison.
Persistent S. aureus colonization is associated with a more severe AD phenotype
Next, we further classified the colonization phenotypes to elucidate the temporal relationship between colonization and skin barrier function, sensitization, and clinical outcomes. We defined four phenotypes, and their characteristics are summarized in Table 2: non-colonized: SA- at V1 and V2; early transient: SA+ at V1, but SA- at V2; late onset: SA- at V1, but SA+ at V2; persistent: SA+ at V1 and V2. There were no baseline differences in demographics, parental reported antibiotic usage in the past 12 months, steroid cream usage, early onset AD or environmental exposures between the phenotypes.
Table 2.
Characteristics of MPAACH cohort by S. aureus phenotypes†
| S. aureus phenotype | Sample size for each variable | Non-colonized n = 264 (81.0) |
Early Transient n = 38 (11.6) |
Late Onset n = 15 (4.6) |
Persistent n= 9 (2.8) |
p-value‡ |
|---|---|---|---|---|---|---|
| Demographic | ||||||
| Age (y) | 326 | 2.1 (1.5–2.4) | 1.9 (1.5–2.4) | 1.8 (1.4–2.4) | 2.1 (1.9–2.4) | 0.65 |
| Male sex | 326 | 137 (52) | 21 (55) | 10 (67) | 4 (44) | 0.66 |
| Black race | 326 | 180 (68) | 22 (58) | 8 (53) | 5 (56) | 0.36 |
| Early onset AD (<3 mos of age) | 326 | 111 (42) | 16 (42) | 10 (67) | 5 (56) | 0.26 |
| AD Severity at Visit 1 | ||||||
| SCORAD (score) | 326 | 19.0 (13.0–29.1) | 25.3 (12.5–42.9) | 27.3 (11.5–37.4) | 33.5 (22.8–44.4) | 0.012 |
| Severity group | 326 | 0.045 | ||||
| Mild (SCORAD < 25) | 172 (65) | 19 (50) | 7 (47) | 3 (33) | ||
| Moderate/Severe (SCORAD ≥ 25) | 92 (35) | 19 (50) | 8 (53) | 6 (67) | ||
| AD Severity at Visit 2 | ||||||
| SCORAD (score) | 324 | 16.9 (9.6–24.7) | 26.0 (10.2–36.6) | 25.1 (12.0–37.7) | 40.1 (28.0–68.8) | <0.001 |
| Severity group | 324 | <0.001 | ||||
| Mild (SCORAD < 25) | 198 (75) | 18 (47) | 7 (47) | 2 (22) | ||
| Moderate/Severe (SCORAD ≥ 25) | 64 (24) | 20 (53) | 8 (53) | 7 (78) | ||
| Skin barrier function at Visit 1 | ||||||
| Lesional skin | ||||||
| Keratinocyte FLG expression§ | 267 | 2.36 (0.68–6.05) | 1.23 (0.22–5.10) | 3.50 (1.29–8.28) | 2.26 (1.87–3.18) | 0.35 |
| TEWL (g/m2/h) | 324 | 11.5 (8.4–18.3) | 19.6 (11.3–32.1) | 14.7 (12.5–19.5) | 15.3 (11.4–19.7) | 0.002 |
| Non-lesional skin | ||||||
| Keratinocyte FLG expression§ | 268 | 3.97 (1.21–8.83) | 2.27 (1.22–6.52) | 11.49 (4.36–15.09) | 4.06 (2.28–9.35) | 0.10 |
| TEWL (g/m2/h) | 323 | 9.4 (7.1–13.2) | 9.4 (8.0–12.4) | 9.6 (7.7–13.7) | 11.4 (9.2–14.5) | 0.62 |
| Skin barrier function at Visit 2 | ||||||
| Lesional skin | ||||||
| Keratinocyte FLG expression§ | 241 | 2.89 (1.12–8.37) | 2.09 (0.68–4.68) | 0.81 (0.20–2.89) | 0.26 (0.15–2.64) | 0.06 |
| TEWL (g/m2/h) | 314 | 11.7 (8.2–16.2) | 16 (8.5–22.1) | 18.3 (11.7–24.2) | 20.5 (8.7–45.1) | 0.022 |
| Non-lesional skin | ||||||
| Keratinocyte FLG expression§ | 242 | 4.09 (1.48–9.00) | 1.65 (0.47–4.60) | 2.13 (0.49–3.61) | 1.77 (0.46–2.79) | 0.004 |
| TEWL (g/m2/h) | 312 | 9.2 (7.1–11.5) | 9.7 (7.6–13.4) | 10.2 (7.8–13.1) | 9.6 (7.7–14.2) | 0.59 |
Data are presented in median (IQR) or n (%).
Comparison across S. aureus phenotypes are done using Kruskal-Wallis test for continuous variables, Chi-square tests for categorical variables. P-values of variables with small sample size (i.e. cell size of 0) are not reported.
Season-adjusted FLG expression values were used in the comparison.
There were significant differences in SCORAD across all phenotypes (V1: p = 0.012, V2: p < 0.001, Table 2). Children with persistent S. aureus colonization had a higher median SCORAD at V1 compared to non-colonized children (33.5 vs. 19.0, p = 0.004), and this difference was more pronounced at V2 (40.1 vs.16.9, p < 0.001). Children with early transient and late onset colonization had intermediate SCORAD values between non-colonized and persistent groups at both V1 and V2 (Table 2).
We next evaluated the proportion of subjects with mild vs moderate/severe AD between the S. aureus phenotypes at V1 and V2. Compared to the non-colonized phenotype, there was a higher proportion of children with moderate/severe AD at V2 in the early transient (53% vs. 24%, p < 0.001), late onset (53% vs. 24%, p = 0.028) and persistent (78% vs. 24%, p = 0.001) phenotypes (Fig 1A). Although there were differences in the proportion of subjects with mild vs moderate/severe AD among the S. aureus phenotypes at V1 (p = 0.045, Table 2), there were no significant differences across pairwise comparisons at V1 (Fig 1A).
Fig 1. AD severity, sensitization status and FLG expression across S. aureus phenotypes in MPAACH subset.
(A) AD severity (B) sensitization pattern (C,D) non-lesional (NL) and (E,F) lesional (L) FLG at V1 and V2 across four S. aureus phenotypes. Differences across phenotypes were evaluated using Kruskal-Wallis test, followed by Dunn’s test where significance was observed. Comparisons of mild vs. moderate-to-severe (Mod/Sev) AD and non-sensitized vs. sensitized groups between phenotypes were done using Chi-square test. Outliers were removed in the figures for better visualization.
We next evaluated allergen sensitization amongst the colonization phenotypes. At V1, there was a higher proportion of subjects who were allergen-sensitized in the persistent phenotype compared to the early transient phenotype (89% vs. 47%, p = 0.030) and non-colonized phenotype (88% vs. 42%, p = 0.012) (Fig 1B). This higher prevalence of sensitization continued at V2 in children with persistent S. aureus colonization compared to the non-colonized children (78% vs. 42.0%, p = 0.045). However, the relationships between sensitization patterns and SA phenotype may be dependent on AD severity as their association was not significant after controlling for AD severity.
Decreased non-lesional skin barrier integrity is associated with S. aureus colonization
We next evaluated FLG RNA expression and TEWL across the S. aureus phenotypes. At V2, NL-FLG expression was lower in the early transient (p = 0.017), late onset (p = 0.034) and persistent (p = 0.029) colonization phenotypes compared to the non-colonized group (Fig 1D). Further, the data showed a trend of decreasing NL-FLG expression from non-colonized to early transient, early transient to late onset, and late onset to persistent groups, suggesting a dose effect. At V1, the median NL-FLG expression was lower in the early transient compared to the late onset colonization phenotype (p = 0.012, Fig 1C), but no other differences were noted between the groups. At V1, L-FLG expression did not differ between the phenotypes whereas at V2, the early transient colonization group had higher L-FLG expression compared to the late onset colonization group (p = 0.013, Fig 1E–F). There was a higher L-TEWL at V1 in the early transient phenotype (p < 0.001, Fig S1A) and a higher NL-TEWL at V2 in the late onset phenotype (p = 0.042, Fig S1D) compared to the non-colonized phenotype. We next evaluated AD severity and skin barrier function at V1 and V2 within each phenotype (Table 3). The non-colonized group showed improvement in SCORAD (p<0.001) and NL-TEWL (p=0.013) over time. The V1 colonized only (early transient) showed no change in these parameters over time, but the V2 colonized only (late onset) showed a dramatic decrease in NL-FLG (p = 0.031), and the persistent colonized group showed low barrier function and high SCORAD throughout both timepoints. These data validated our key findings across all groups.
Table 3.
Comparison of V1 and V2 SCORAD, sensitization and skin barrier function in different S. aureus colonization/phenotype group†
| S. aureus colonization/phenotype | Non-colonized (n = 264) |
Early Transient (n = 38) |
||||
|---|---|---|---|---|---|---|
| V1 | V2 | p ‡ | V1 | V2 | p ‡ | |
| AD severity | ||||||
| SCORAD (score) | 19.0 (13.0–29.1) | 16.9 (9.6–24.7) | < 0.001 | 25.3 (12.5–42.9) | 26.0 (10.2–36.6) | 0.12 |
| Moderate/Severe AD (SCORAD ≥ 25) | 92 (35%) | 64 (24%) | 0.002 | 19 (50%) | 20 (53%) | 0.99 |
| Sensitization | ||||||
| Sensitized | 110 (42%) | 111 (42%) | 0.99 | 18 (47%) | 21 (55%) | 0.45 |
| Total serum IgE | 32 (10–93) | 70 (18–142) | < 0.001 | 44 (11–195) | 62 (14–282) | 0.20 |
| Skin barrier - Lesional skin | ||||||
| Keratinocyte FLG expression§ | 2.36 (0.69–6.05) | 2.89 (1.12–8.37) | 0.33 | 1.23 (0.22–5.10) | 2.09 (0.68–4.68) | 0.92 |
| TEWL (g/m2/h) | 11.5 (8.4–18.3) | 11.7 (8.2–16.2) | 0.25 | 19.6 (11.3–32.1) | 16.0 (8.5–22.1) | 0.15 |
| Skin barrier - Non-lesional skin | ||||||
| Keratinocyte FLG expression§ | 3.97 (1.21–8.83) | 4.09 (1.48–9.00) | 0.67 | 2.27 (1.22–6.52) | 1.65 (0.47–4.60) | 0.58 |
| TEWL (g/m2/h) | 9.4 (7.1–13.2) | 9.2 (7.1–11.5) | 0.013 | 9.4 (8.0–12.4) | 9.7 (7.6–13.4) | 0.78 |
| S. aureus colonization/phenotype | Late Onset (n = 15) |
Persistent (n = 9) |
||||
| V1 | V2 | p ‡ | V1 | V2 | p ‡ | |
|
| ||||||
| AD severity | ||||||
| SCORAD (score) | 27.3 (11.5–37.4) | 25.1 (12.0–37.7) | 0.60 | 33.5 (22.8–44.4) | 40.1 (28.0–68.8) | 0.42 |
| Moderate/Severe AD (SCORAD ≥ 25) | 8 (53%) | 8 (53%) | 0.99 | 6 (67%) | 7 (78%) | 0.99 |
| Sensitization | ||||||
| Sensitized | 9 (60%) | 8 (53%) | 0.99 | 8 (89%) | 7 (78%) | 0.99 |
| Total serum IgE | 64 (14–170) | 144 (72–430) | 0.56 | 176 (14–250) | 93 (55–335) | 0.50 |
| Skin barrier - Lesional skin | ||||||
| Keratinocyte FLG expression§ | 3.50 (1.29–8.28) | 0.81 (0.20–2.89) | 0.31 | 2.26 (1.87–3.18) | 0.26 (0.15–2.64) | 0.69 |
| TEWL (g/m2/h) | 14.7 (12.5–19.5) | 18.3 (11.7–24.2) | 0.79 | 15.3 (11.4–19.7) | 20.5 (8.7–45.1) | 0.36 |
| Skin barrier - Non-lesional skin | ||||||
| Keratinocyte FLG expression§ | 11.49 (4.36–15.09) | 2.13 (0.49–3.61) | 0.031 | 4.06 (2.29–9.36) | 1.77 (0.46–2.79) | 0.15 |
| TEWL (g/m2/h) | 9.6 (7.7–13.7) | 10.2 (7.8–13.1) | 0.79 | 11.4 (9.2–14.5) | 9.6 (7.7–14.2) | 0.65 |
Data are presented in median (IQR) or n (%).
Comparison between V1 and V2 within each population were done using Wilcoxon signed rank test for continuous variables, McNemar’s tests for categorical variables.
Season-adjusted FLG expressions were used in the comparison.
V2 S. aureus colonization is associated with a decrease in non-lesional filaggrin expression over time
To further study temporal associations between S. aureus colonization and FLG expression, subjects were dichotomized into V2 SA+ (orange line) and V2 SA- (blue line) groups. The L-FLG and NL-FLG expression values at each visit were plotted, and linear mixed-effect model equations were calculated for each group (Fig 2). There was a significant difference in the change or delta in NL-FLG expression level from V1 to V2 based on V2 S. aureus colonization (p = 0.0012, Fig 2A). Specifically, children who were colonized with S. aureus at V2 (V2 SA+) demonstrated a 0.98 unit decrease in NL-FLG expression over time on average, while those non-colonized at V2 (V2 SA-) had a 0.03 unit increase in their NL-FLG over time. To determine the impact of colonization status at V1 on these observations, we further stratified by V1 SA status (V1 SA+ and V1 SA-) (Fig 2B–C). The decrease in NL-FLG expression from V1 to V2 in those colonized at V2 was observed in both V1 SA+ and SA- groups, although the V1 SA+ group reached only nominal significance. Similarly, the increase in NL-FLG expression from V1 to V2 in those non-colonized at V2 was observed in both V1 SA+ and SA- groups. These results suggested that the NL-FLG expression over time determines colonization at V2 and that this is independent of colonization at V1. To determine whether this was the case, we directly examined whether the change in NL-FLG expression (V2 – V1 NL-FLG, denoted as delta NL-FLG) was associated with S. aureus colonization status. V2, but not V1 colonization, was significantly associated with decreased delta NL-FLG (V1 colonization: p = 0.56; V2 colonization: p = 0.010). This relationship between V2 colonization and delta NL-FLG remained significant after controlling for V1 S. aureus colonization status and V1 SCORAD (p = 0.011). There were no significant differences in L-FLG expression over time irrespective of colonization status (Fig 2D–F).
Fig 2. FLG expression level and S. aureus colonization across V1 and V2.
(A) Decreasing trend of NL-FLG expression in participants with V2 S. aureus colonization, (B,C) stratified by V1 S. aureus colonization. (D) L-FLG expression from V1 to V2 in participants with V2 S. aureus colonization, (E,F) stratified by V1 S. aureus colonization. Model equations were constructed using linear mixed-effect model to account for the random subject effect, fixed effects of V2 S. aureus colonization and time (i.e., visit), and an interaction term between colonization and time. P-values represent the significance of interaction terms. Season-adjusted FLG expression values are presented. Dashed lines represent data on the individual scale; solid lines represent the mean for each group.
Non-lesional filaggrin expression mediates S. aureus colonization
We developed a mediation model to examine the causal relationship between S. aureus status and FLG expression. In the model, V1 S. aureus status was the independent variable, V2 S. aureus status was the dependent variable, and V2 NL-FLG expression was the mediator. Our sample size was too small to evaluate delta NL-FLG as the mediator. The direct effect of V1 S. aureus on V2 S. aureus status was significant (ADE average = 0.144, p = 0.027). V1 S. aureus status was also marginally associated with lower V2 NL-FLG (β = −0.30, p = 0.09). We observed a mediating effect of V2 NL-FLG expression levels on V2 S. aureus colonization (ACME average = 0.046, p = 0.046).
This mediation model focused on a unit-change in the expression levels. As differences in FLG expression among children often did not exceed 1 unit, interpretation of the effect sizes can be challenging. Thus, we compared the higher and lower FLG expression tertiles (n = 162). The mediation effect of V2 NL-FLG expression on V2 S. aureus status was significant (ACME average = 0.05, p = 0.018) but the direct effect was not (ADE average = 0.073, p = 0.28). When evaluating the components of the mediation, we found that lower V2 NL-FLG is associated with 11.04 times the odds of being colonized by S. aureus at V2 (p = 0.024) compared to higher FLG expression. We also found that V1 SA+ children had 3.32 times the odds of having lower V2 NL-FLG (p = 0.017) compared to V1 SA- children, which is consistent with the patterns of NL-FLG seen in Figure 1. However, the presence of a significant mediation term supports that V2 NL-FLG expression contributes to V2 colonization risk above the direct risk of V1 colonization status. These data show that while NL-FLG expression mediates S. aureus colonization at V2 independent of V1 colonization, this effect is likely modified by V1 colonization status. Indeed, when we evaluated the effect size of the ACME model conditional on V1 S. aureus status, we observed that S. aureus positive group had approximately twice the effect size for our mediation model. Formal interaction between V1 S. aureus colonization status and V2 NL-FLG expression was not significant (p = 0.26). The main effect of V2 NL-FLG on V2 S. aureus colonization remained significant even after adjustment for V1 and V2 SCORAD, race, and age, supporting the robustness of our findings (data not shown).
To evaluate the specificity of the mediation effect, we evaluated additional mediation models. We considered V1 NL-FLG and V1/V2 L-FLG expression as mediators, but these were not significant (data not shown). Lastly, to evaluate whether changes in FLG expression were mediated by S. aureus colonization, we developed another mediation model to study the relationship between V1 NL-FLG expression as the independent variable, V2 NL-FLG expression as the dependent variable, and V2 S. aureus status as the mediator. The mediation effect was not significant (Table S1).
DISCUSSION
This is the first study to report longitudinal quantitative skin FLG expression levels and S. aureus skin colonization data in children with AD over time in early life. By leveraging the MPAACH prospective early-life cohort of children with AD, we defined longitudinal S. aureus skin colonization phenotypes, determined skin L- and NL-FLG expression trajectories over time, and then integrated these to understand the interplay of skin FLG levels and S. aureus colonization and their impact on sensitization patterns and AD clinical outcomes. Our data revealed that V2 NL-FLG, but not L- FLG, expression levels mediated V2 S. aureus colonization independent of V1 and V2 AD severity, underscoring the critical role that the NL skin plays in AD pathogenesis. The effect was modified by V1 S. aureus colonization status such that children colonized at V1 had increased odds of having lower V2 NL-FLG compared to children without S. aureus colonization at V1. Children colonized with S. aureus at V2 showed a decrease in NL-FLG expression over time while those who were V2 SA- maintained or slightly increased NL-FLG expression. Persistent colonization was associated with increased sensitization and worsened AD severity as assessed by SCORAD. Our findings reveal a longitudinal AD endotype of persistent S. aureus colonization characterized by increased AD severity, allergic sensitization, and decrease in NL-FLG expression. Collectively, these data strongly suggest that intervention designed to interrupt and change the decrease in NL-FLG expression early in life may be critical in blocking later S. aureus colonization and the clinical sequalae.
V2 S. aureus colonization was directly mediated by V2 NL-FLG expression demonstrating the contribution of NL-skin barrier disruption to late onset and persistent colonization. FLG breakdown products have been reported to lead to increased acidification of media, decreased density of S. aureus and decreased production of bacterial colonization and immune evasion proteins in an in vitro model7, so a decrease in FLG expression could impact future colonization. Another study showed increased S. aureus adherence with either FLG-knock down or supplementation of IL-31, a cytokine which has been shown to downregulate FLG expression in epidermal skin models.17,18 An additive effect was observed with both FLG-knock down and IL-31 supplementation.18 Therefore, lower FLG may create a permissive environment in the context of attenuated immunity for further colonization. In a recent pediatric study, skin S. aureus load in children with difficult-to-treat AD was not associated with the FLG mutation status,19 but they did not examine the expression level of FLG, which is affected by more than genetics. This underscores a major strength of our study, where we quantified FLG RNA expression levels in NL and L skin over time. We did not observe significance when using V1 or V2 L-FLG expression as mediators in our analyses, indicating that the link between FLG expression and colonization may be restricted to NL-skin. This reinforces previous findings from our group and others that NL-skin is driving the AD and sensitization phenotype.13, 20 This mediation, however, does not prove causality and does not rule out the contribution of other potential factors such as medication and inflammation in S. aureus colonization. While we didn’t see a mediating effect by S. aureus colonization on NL-FLG expression, it remains possible and even likely that S. aureus colonization also acts upstream leading to further decreased FLG, setting up a detrimental feed forward loop. For example, S. aureus superantigen B (SEB) has been shown to increase IL-31 expression and suppress FLG expression.21 Medications could also impact our findings. While we did not observe differences in the use of parental reported use of antibiotics in the past 12 months and steroid cream usage at baseline between the colonization phenotypes, consistent use of medications including antibiotics and biologics may affect colonization rates and therefore, FLG expression.
We observed that V2 SA+ children had a decrease in NL-FLG expression from V1 to V2 compared to those who were V2 SA-, who maintained similar FLG levels over time. This decreae in NL-FLG expression was not conditional on V1 colonization status but likely driven by the V1 SA- group, as the magnitude of the decrease is more pronounced in this group. Our linear regression models reinforced that changes in NL-FLG are associated with V2, but not V1, S. aureus colonization, suggesting that the change is important for determining future colonization for the late onset and persistent phenotypes in our sample. A previous study demonstrated sustained change in FLG expression within weeks in response to treatment22. Maintenance of NL-FLG expression levels in V2 SA- children may be due to numerous factors including better compliance with skin care regimens for eczema treatment which may lead to improved barrier function and/or decreased inflammation. Th2 inflammation has been shown to lead to reduced skin FLG expression.23 As such, a decrease in local inflammation would be predicted to increase FLG expression and overall skin barrier health.24 In contrast, persistent inflammation and/or poor skin care may lead to sustained decreased FLG expression over time.
Persistent colonization over time was associated with a higher SCORAD and increased AD severity and proportion of children with sensitization. Studies have shown that S. aureus is correlated with increased expression of proinflammatory cytokines including TSLP, IL-8, IL-1β from keratinocytes, mast cell degranulation and increased Th2-mediated inflammation.25 As such, differences in an individual’s ability to clear S. aureus colonization likely also plays a role in AD outcomes.
To our knowledge, this study is the first to examine the relationship between longitudinal FLG expression and S. aureus colonization. Two related studies from 2012 and 2017 performed 16S rRNA sequencing and shotgun sequencing on skin tapes on 422 pediatric samples across baseline, flare and post-flare periods but neither investigated skin barrier function.5, 11 Other studies have examined strain variation over time and association with AD severity.9, 26, 27 Another recent study of 49 AD and control adults identified distinct microbial configurations (“dermotypes”) identified by shotgun metagenomics across two visits separated by 4 weeks. Although samples were collected to examine components of the NMF over time, their analysis was restricted to visit 1, where they found no difference between the NMF protein quantities between the dermotypes.28 While we also showed an association between colonization and AD severity, the mediating effect of NL-FLG on colonization remained significant independent of V1 and V2 SCORAD.
The strengths of this study lie in the cohort size and extensive biospecimen and data collection capabilities of MPAACH, but it is not without limitations. By design, all participants in this MPAACH cohort have atopic dermatitis, so we cannot make conclusions regarding AD development. However, we did not observe any differences when examining early onset AD (AD appearance at or before 3 months of age) vs late onset (AD appearance after 3 months of age) between the non-colonized vs colonized groups and between the phenotypes. The study may also be limited by inherent variability in direct sampling of skin via agar plates versus other sampling methodologies. Approximately 20% of the subjects in this study were colonized at either visit. This is in line with a previous study utilizing skin swabs which identified S. aureus positive axillar swabs in approximately 15% of infants at the time of AD onset.29 Importantly, variability in sampling methodologies, locations, downstream analysis of biospecimen, S. aureus biofilm production, and ages of participants yield varying results on colonization rates.30 Sequencing may overrepresent microbes that are present in very low quantities whereas contact plates methodology utilized herein may be useful in directly sampling S. aureus isolates that are most relevant to disease progression. We also utilized a multi-layered approach to ensure that our results are specific to S. aureus, which could lead to lower colonization rates compared to other studies. Skin barrier function may also be determined by lipids and other structural proteins such as loricrin and claudin, which we intend to investigate in future analyses. Lastly, the prevalence of commensal microbes that may affect S. aureus colonization was not assessed. However, metagenomic studies on skin tape collected from L- and NL-skin at each MPAACH visit will allow us to determine the taxonomic, strain and functional profiles of the microbiome over time in the cohort.31
Our findings regarding the temporal relationship between FLG expression and S. aureus colonization in early life have significant clinical implications. Given the relationship between persistent colonization and worse clinical outcomes, preventing or disrupting colonization is an objective of therapies. Broad-range antimicrobial therapies exhibit minimal benefits and unintended side effects, highlighting the need for other personalized therapeutic strategies.32 Current AD therapies mostly target L-skin and are aimed at decreasing inflammation or increasing skin moisture, but there are no strategies that specifically target FLG expression in early life in NL-skin. Our findings suggest that it may be clinically useful to monitor skin FLG levels over time and optimize therapies aimed at inhibiting future and persistent S. aureus colonization in early life to prevent worsening AD and progression to other comorbidities.
Supplementary Material
Acknowledgements:
We thank the participating families and staff on the MPAACH study.
Funding:
This work was supported by NIH grant U19AI070235 (G.K.K.H.), R01 AI127392 (G.K.K.H.), and T32 GM063483 (A.D.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Abbreviations:
- ACME
average causal mediation effect
- AD
atopic dermatitis
- ADE
average direct effect
- CCHMC
Cincinnati Children’s Hospital Medical Center
- FLG
filaggrin
- L/NL
lesional/non-lesional
- MPAACH
Mechanisms of Progression of Atopic Dermatitis (AD) to Asthma in Children
- MSA
Mannitol Salt Agar
- OR
odds ratio
- PCR
polymerase chain reaction
- SA
Staphylococcus aureus
- SCORAD
SCORing Atopic Dermatitis index
- SPT
skin prick test
- TEWL
Trans-epidermal water loss
- V1
Visit 1
- V2
Visit 2
Footnotes
Conflict of interest: The authors have no conflicts of interest.
References
- 1.Spergel JM. From atopic dermatitis to asthma: the atopic march. Ann Allergy Asthma Immunol 2010;105(2):99–106; quiz 7–9, 17. [DOI] [PubMed] [Google Scholar]
- 2.Fishbein AB, Silverberg JI, Wilson EJ, Ong PY. Update on Atopic Dermatitis: Diagnosis, Severity Assessment, and Treatment Selection. J Allergy Clin Immunol Pract 2020;8(1):91–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Eyerich K, Eyerich S, Biedermann T. The Multi-Modal Immune Pathogenesis of Atopic Eczema. Trends Immunol 2015;36(12):788–801. [DOI] [PubMed] [Google Scholar]
- 4.Paller AS, Kong HH, Seed P, Naik S, Scharschmidt TC, Gallo RL, et al. The microbiome in patients with atopic dermatitis. J Allergy Clin Immunol 2019;143(1):26–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Byrd AL, Deming C, Cassidy SKB, Harrison OJ, Ng WI, Conlan S, et al. Staphylococcus aureus and Staphylococcus epidermidis strain diversity underlying pediatric atopic dermatitis. Sci Transl Med 2017;9(397). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Nakatsuji T, Chen TH, Two AM, Chun KA, Narala S, Geha RS, et al. Staphylococcus aureus Exploits Epidermal Barrier Defects in Atopic Dermatitis to Trigger Cytokine Expression. J Invest Dermatol 2016;136(11):2192–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Miajlovic H, Fallon PG, Irvine AD, Foster TJ. Effect of filaggrin breakdown products on growth of and protein expression by Staphylococcus aureus. J Allergy Clin Immunol 2010;126(6):1184–90 e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Clausen ML, Edslev SM, Andersen PS, Clemmensen K, Krogfelt KA, Agner T. Staphylococcus aureus colonization in atopic eczema and its association with filaggrin gene mutations. Br J Dermatol 2017;177(5):1394–400. [DOI] [PubMed] [Google Scholar]
- 9.Clausen ML, Edslev SM, Norreslet LB, Sorensen JA, Andersen PS, Agner T. Temporal variation of Staphylococcus aureus clonal complexes in atopic dermatitis: a follow-up study. Br J Dermatol 2019;180(1):181–6. [DOI] [PubMed] [Google Scholar]
- 10.Barnes CJ, Clausen ML, Asplund M, Rasmussen L, Olesen CM, Yusel YT, et al. Temporal and Spatial Variation of the Skin-Associated Bacteria from Healthy Participants and Atopic Dermatitis Patients. mSphere 2022;7(1):e0091721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kong HH, Oh J, Deming C, Conlan S, Grice EA, Beatson MA, et al. Temporal shifts in the skin microbiome associated with disease flares and treatment in children with atopic dermatitis. Genome Res 2012;22(5):850–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Davidson WF, Leung DYM, Beck LA, Berin CM, Boguniewicz M, Busse WW, et al. Report from the National Institute of Allergy and Infectious Diseases workshop on “Atopic dermatitis and the atopic march: Mechanisms and interventions”. J Allergy Clin Immunol 2019;143(3):894–913. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Biagini Myers JM, Sherenian MG, Baatyrbek Kyzy A, Alarcon R, An A, Flege Z, et al. Events in Normal Skin Promote Early-Life Atopic Dermatitis-The MPAACH Cohort. J Allergy Clin Immunol Pract 2020;8(7):2285–93 e6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Biagini JM, Kroner JW, Baatyrbek Kyzy A, Gonzales A, He H, Stevens M, et al. Longitudinal atopic dermatitis endotypes: An atopic march paradigm that includes Black children. J Allergy Clin Immunol 2022;149(5):1702–10 e4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.von Kobyletzki LB, Berner A, Carlstedt F, Hasselgren M, Bornehag CG, Svensson A. Validation of a parental questionnaire to identify atopic dermatitis in a population-based sample of children up to 2 years of age. Dermatology. 2013;226(3):222–6. [DOI] [PubMed] [Google Scholar]
- 16.Hanifin JM, Rajka G. Diagnostic Features of Atopic Dermatitis. Acta Dermatovener (Stockholm) Suppl 1980:42–7. [Google Scholar]
- 17.Cornelissen C, Marquardt Y, Czaja K, Wenzel J, Frank J, Luscher-Firzlaff J, et al. IL-31 regulates differentiation and filaggrin expression in human organotypic skin models. J Allergy Clin Immunol 2012;129(2):426–33, 33 e1–8. [DOI] [PubMed] [Google Scholar]
- 18.van Drongelen V, Haisma EM, Out-Luiting JJ, Nibbering PH, El Ghalbzouri A. Reduced filaggrin expression is accompanied by increased Staphylococcus aureus colonization of epidermal skin models. Clin Exp Allergy. 2014;44(12):1515–24. [DOI] [PubMed] [Google Scholar]
- 19.van Mierlo MMF, Pardo LM, Fieten KB, van den Broek TJ, Schuren FHJ, van Geel M, et al. The Skin and Nose Microbiome and Its Association with Filaggrin Gene Mutations in Pediatric Atopic Dermatitis. Dermatology. 2022;238(5):928–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Leung DYM, Calatroni A, Zaramela LS, LeBeau PK, Dyjack N, Brar K, et al. The nonlesional skin surface distinguishes atopic dermatitis with food allergy as a unique endotype. Sci Transl Med 2019;11(480). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Sonkoly E, Muller A, Lauerma AI, Pivarcsi A, Soto H, Kemeny L, et al. IL-31: a new link between T cells and pruritus in atopic skin inflammation. J Allergy Clin Immunol 2006;117(2):411–7. [DOI] [PubMed] [Google Scholar]
- 22.Guttman-Yassky E, Bissonnette R, Ungar B, Suarez-Farinas M, Ardeleanu M, Esaki H, et al. Dupilumab progressively improves systemic and cutaneous abnormalities in patients with atopic dermatitis. J Allergy Clin Immunol 2019;143(1):155–72. [DOI] [PubMed] [Google Scholar]
- 23.Howell MD, Kim BE, Gao P, Grant AV, Boguniewicz M, Debenedetto A, et al. Cytokine modulation of atopic dermatitis filaggrin skin expression. J Allergy Clin Immunol 2007;120(1):150–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kim J, Kim BE, Ahn K, Leung DYM. Interactions Between Atopic Dermatitis and Staphylococcus aureus Infection: Clinical Implications. Allergy Asthma Immunol Res 2019;11(5):593–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Park KD, Pak SC, Park KK. The Pathogenetic Effect of Natural and Bacterial Toxins on Atopic Dermatitis. Toxins (Basel) 2016;9(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Alsterholm M, Strombeck L, Ljung A, Karami N, Widjestam J, Gillstedt M, et al. Variation in Staphylococcus aureus Colonization in Relation to Disease Severity in Adults with Atopic Dermatitis during a Five-month Follow-up. Acta Derm Venereol 2017;97(7):802–7. [DOI] [PubMed] [Google Scholar]
- 27.Lomholt H, Andersen KE, Kilian M. Staphylococcus aureus clonal dynamics and virulence factors in children with atopic dermatitis. J Invest Dermatol 2005;125(5):977–82. [DOI] [PubMed] [Google Scholar]
- 28.Tay ASL, Li C, Nandi T, Chng KR, Andiappan AK, Mettu VS, et al. Atopic dermatitis microbiomes stratify into ecologic dermotypes enabling microbial virulence and disease severity. J Allergy Clin Immunol 2021;147(4):1329–40. [DOI] [PubMed] [Google Scholar]
- 29.Meylan P, Lang C, Mermoud S, Johannsen A, Norrenberg S, Hohl D, et al. Skin Colonization by Staphylococcus aureus Precedes the Clinical Diagnosis of Atopic Dermatitis in Infancy. J Invest Dermatol 2017;137(12):2497–504. [DOI] [PubMed] [Google Scholar]
- 30.Totte JE, van der Feltz WT, Hennekam M, van Belkum A, van Zuuren EJ, Pasmans SG. Prevalence and odds of Staphylococcus aureus carriage in atopic dermatitis: a systematic review and meta-analysis. Br J Dermatol 2016;175(4):687–95. [DOI] [PubMed] [Google Scholar]
- 31.Stevens ML, Gonzalez T, Schauberger E, Baatyrbek Kyzy A, Andersen H, Spagna D, et al. Simultaneous skin biome and keratinocyte genomic capture reveals microbiome differences by depth of sampling. J Allergy Clin Immunol 2020;146(6):1442–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.George SM, Karanovic S, Harrison DA, Rani A, Birnie AJ, Bath-Hextall FJ, et al. Interventions to reduce Staphylococcus aureus in the management of eczema. Cochrane Database Syst Rev 2019;2019(10). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.