A systematic review and critical evaluation of immunohistochemical associations in hidradenitis suppurativa

Background: Hidradenitis suppurativa (HS) is a chronic inflammatory disease with significant morbidity and impact on quality of life. Our understanding of the pathophysiology is incomplete, impairing efforts to develop novel therapeutic targets. Immunohistochemistry studies have produced conflicting results and no systematic evaluation of study methods and results has been undertaken to date. Methods: This systematic review aimed to collate and describe all reports of immunohistochemical staining in HS. This systematic review was registered with PROSPERO and conducted in line with the PRISMA reporting guidelines. Potential bias was assessed using the NIH Criteria and antibodies used across various studies were tabulated and compared. Results: A total of 22 articles were identified describing results from 494 HS patients and 168 controls. 87 unique immunohistochemical targets were identified. The overall quality of studies was sub-optimal with staining intensity confounded by active treatment. Conflicting data was identified and able to be reconciled through critical evaluation of the study methodology. Conclusions: Keratinocyte hyperplasia with loss of cytokeratin markers co-localizes with inflammation comprising of dendritic Cells, T-lymphocytes and macrophages, which are known to play central roles in inflammation in HS. Primary follicular occlusion as a pathogenic paradigm and the principal driver of HS is unclear based upon the findings of this review. Inflammation as a primary driver of disease with secondary hyperkeratosis and follicular occlusion is more consistent with the current published data.


Introduction
Hidradenitis suppurativa (HS) is a chronic inflammatory disease, the exact pathophysiology of which remains incompletely defined 1 . Numerous inflammatory mediators including TNF-α 2 , IL-17 2,3 , IL-32 4 and IL-36 subtypes 5,6 have been implicated in the disease. However, there is an incomplete understanding of the source and triggers of these mediators and how they sustain the chronic inflammation that characterizes this disease 1,2 . The pathogenic paradigm of HS has evolved dramatically since the first description by Velpau in 1839 7 . First thought of as an apocrinitis of infectious aetiology, it is now considered a disorder of follicular occlusion and more recently an inflammatory disease characterised by a keratinocyte mediated inflammatory response 6 . However, the variable response to topical, systemic and biologic therapies in HS 8 indicate our understanding of disease pathophysiology is incomplete when compared to other cutaneous inflammatory diseases such as psoriasis 9 and atopic dermatitis 10 . Existing studies examining the histology and immunohistochemical profiling of HS tissues represent conflicting results, for example in the degree of dermal dendritic cell infiltration 11,12 and the production of TNF-alpha in the follicular unit 13,14 . These results may be influenced by heterogeneous sampling methods, laboratory processing methods and data analysis 15 . An additional complicating factor is that clinical comorbidities which are strongly associated with disease activity in HS, such as obesity 16 , diabetes 17 , inflammatory bowel disease 18 , and smoking 19 also impact inflammatory cell activity in the skin 18,[20][21][22] . Hence it remains unclear whether the presence or absence of these conditions may confound the findings of immunohistochemical studies in HS 15, 23 and whether clinical stratification of patients is required to identify distinct pathogenic pathways, which may be amenable to pharmacological intervention. This variability across studies makes comparing data problematic. To date no systematic analysis of immunohistochemical studies has been undertaken to compare results, methodology and analytical techniques.

Objectives
The objectives of this systematic review are: 1) To collate and describe all published reports of immunohistochemical studies in HS 2) To critically evaluate the sampling, laboratory and analysis techniques used in each study to determine if comparisons can be made across studies.

Methods
This systematic review was registered with PROSPERO 24 (Registration number CRD42018104763) and was conducted in line with the PRISMA 25 . The STROBE statement 26 was used to assess the observational studies included in this study.
Data sources Information Sources for this review encompassed Pubmed (1946-July 1 2018), Scopus (2004-July 1 2018) and Web of Science (1990-July 1 2018) as shown in Figure 1. Search strategy is presented as Table 1.

Study eligibility criteria
Eligibility criteria for this review included cohort studies, casecontrol studies and other observational studies with no restrictions of patient age, sex, ethnicity or language of publication. Eligible studies included those reporting the results of immunohistochemical findings in HS. Studies deemed not eligible included articles which provided no new data, only a review or summary of previously published data.
Appraisal and synthesis methods Data collection was performed independently by 2 authors (JWF & JEH), with any disagreements regarding inclusion of citations being referred to a third author (JGK) for mediation. Information was collected using a standardized data collection form (available as Extended data 27 ) with the principal outcomes of interest being the immunohistochemical stain of interest, the site and rated intensity of staining (as described by authors), and comparison with perilesional/ unaffected/ control tissue. If data from individual patients was not available then the aggregate data was collected.
Potential sources of bias in the identified studies are acknowledged including the small size of patient cohorts, the variability in sampling and laboratory techniques, antibodies published and reactants used. Therefore these variables (where available) were collated to assess the heterogeneity of studies. Bias was also assessed using the NIH quality assessment tool for observational studies 28 .

Descriptive analysis
The demographics of the patients of the included studies are presented in Table 2 (Table 4, Table 5 and Table 6).

Immunohistochemistry results
Epidermis. The epidermis of HS lesional tissue expressed the normal array of keratins (K) in the basal (K5, K14) and suprabasal (K1, K2e, K10) layers. K6, K16 and K17 staining were increased compared to healthy controls in the suprabasal epidermis in one study 30 , however, K6 and K17 staining was not increased in the epidermis (only in non-keratinized portions of sinus tracts) in a second study 38 . Where K6 and K17 were positive in suprabasal epidermis, K17 staining was more pronounced than K6 staining 30 . K19 was weakly positive in acanthotic epidermis 37 . Ki67 staining was elevated in basal and suprabasal epidermis. Normal staining patterns of desmoplakin, plakophilin and plakoglobin were seen 38 . Cells staining positive for CD1a, CD206, CD207 and CD209 were seen throughout the epidermis 40 . CD3, CD4, CD8 and to a lesser degree CD68 positive cells demonstrated epidermotropism in sites of epidermal acanthosis 30,33 . CD29 and cholera toxin (double positive)            staining cells were seen on the slopes of papillae of the epidermis 35 . hBD2 (human beta defensin) staining was decreased throughout the epidermis in two studies 13,41 whilst hBD3 staining was increased throughout the suprabasal epidermis 14,42 , however only significantly in Hurley Stage 1 and 2 patients (p=0.045) 42 . hBD4 was decreased in suprabasal epidermis compared to healthy controls (p=0.001) 41 . Contradictory findings were seen in toll like receptor (TLR) 2 staining with an increase in the epidermis co-localizing with dendritic cells and macrophages in one study 40 but suppressed in a second study 41 . Levels of TLR3, TLR4, TLR7, TLR9, ICAM-1, TGF-Beta and IGF-1 were only assessed by one study and all were suppressed throughout the epidermis compared with controls 41 . RNAase7 was increased in expression compared to healthy controls (p<0.05) 42 . MMP2 was positively expressed in keratinocytes throughout the epidermis 13 and MMP8 in neutrophils within the epidermis 43 . TNF-α was highly expressed in macrophages and lymphocytes present in the epidermis, particular in the basal layers 13,14 and NLRP3, MIF, S100A7, LL37/Cathelicidin and α-MSH all positive in suprabasal keratinocytes 30,41 . IL-6 and IL-10 were reported as suppressed compared to healthy control skin 41 , however, IL-36 subtypes were highly expressed in epidermal keratinocytes (more suprabasal than basal) 5,6 with IL-32 also positive in the stratum granulosum 4 .
Dermis. CD1a, CD11c, CD206, CD207, CD209 and Factor XIIIa positive cells were identified in the dermis in three separate studies 12,30,40 , however the degree of infiltration varied. Dermal infiltrates of CD3, CD4, and CD8 positive cells, continuous with the epidermal infiltrates were a consistent feature of lesional HS dermis and were increased over controls 30,33 . The distribution of these cells was most pronounced in the interfollicular dermis (ie. towards the papillary slopes) and perifollicularly (ie. peri-infundibularly) 30,33 . CD56, CD68 and CD138 positive cells were diffusely seen throughout the dermis 40 . CD19 and CD20 positive pseudolymphoid follicles have been noted in other studies 30,40 . Single keratinocytes have also been identified in the dermis which stain with pancytokeratin markers (AE1/AE3/PKC26) 36  Apocrine gland nuclei stained weakly positive for estrogen receptor 39 and androgen receptor 39 , and these results were reported as no different from control specimens 39 . Lysozyme staining of apocrine glands was seen in cases of vulval HS only 34 .
Immunohistochemistry methods. The list of antibodies used for IHC staining is presented in Table 6. Consistent antibodies were used for CD1a; CD20 and tryptase staining, whilst different antibodies were used for other staining targets. Antibodies used were not described in two studies 34,36 .

Assessment of Bias.
The result of bias assessment using NIH criteria is presented in Table 7. All 22 articles clearly stated the research question of interest with well-defined study populations. The application of inclusion and exclusion criteria, or the calculation of sample size, or effect estimates were not described in any study. Exposures (ie. the presence of disease) were established and measured in all studies prior to the outcome measures (IHC staining) being assessed and the disease was established for such a time that a relationship between exposure and outcome would be identified if one existed. Different levels of exposure (severity of disease) was taken into account in only two studies 4,12 and was consistently measured using Hurley staging across all studies. No articles accounted for all possible confounding variables such as obesity, diabetes, family history or smoking status (Table 3).

Quality of data and risk of bias
The overall quality of data in this systematic review was suboptimal with poor correction for potential confounding factors with only two of the 22 studies using objective measurement systems for IHC staining intensity 4,12 . The proportion of smokers was elevated (94%) compared to the rates of smoking in the HS population at large (70-89%) 45 . A number of studies (17/22) did not stratify results by treatment therefore there is a risk that staining intensity of pro-inflammatory mediators may be reduced due to concomitant treatment at the time of biopsy. The use of de-paraffinized tissue in retrospective studies 30,33,34 can lead to false negatives in IHC dependent upon the preparation method of the original sample and the de-paraffinization process 15 . Hence there are factors in the population studied in this review which may bring into question the reliability of staining quantification. However, the presence or absence of IHC staining, particularly when confirmed in multiple studies is still considered reliable despite the risks of bias.

Conflicting results
Conflicting results were identified in dermal CD1a staining 12,30,40 , dermal CK19 staining [36][37][38] , Epidermal TLR2 staining 40,41 and TNF alpha staining in the follicular infundibulum 13,14 . Regarding    CD1a staining, two of the studies reported only a mild dermal infiltrate of CD1a positive cells 30,40 , with a third study demonstrating a significant infiltration of these cells 12 . This third study clearly documented all treatment was withheld 3 weeks prior to the biopsies being taken 12 , whereas there is no description in the other two articles regarding the discontinuation or ongoing use of treatments 30,40 . Therefore, with the possibility of partially treated disease, an artificial reduction in the number of dermal dendritic cells is a possibility as treatment for HS (such as adalimumab) has been demonstrated to effectively reduce the infiltration of dendritic cells in vivo 12 . Similarly, studies examining TNF-alpha staining also differed in their stratification of patient based upon active treatment 13,14 . Significant reductions in TNF alpha staining were seen in the study with no documentation of treatment cessation 14 when compared to the one study with clear documentation that all patients had treatment ceased prior to biopsy 13 . K19 staining was reported negative in all areas of the sinus tracts in one study 37 , whereas two additional studies 36,38 described positive K19 staining in sinus tracts (one study non-specifically 36 and the second in the deep inflamed, non-keratinized epithelium of the tract 38 ). The difference between these staining patterns may be explained by the presence of inflammation. Kurzen et al. 38 described the presence of K19 staining in non-keratinized epithelium of the deep sinus tracts only when associated with inflammation (Type 3 epithelia), staining was negative when no inflammation was present (Type 2 epithelia) 38 . Kurokawa et al. did not differentiate between inflamed and non-inflamed non-keratinized epithelium in their study 37 , and noted that the lesser degree of inflammation seen histologically may explain their differing results in comparison to Kurzen's study 37 .
Localization of production of inflammatory mediators IHC staining, in particular co-staining with cellular markers and cytokines has enabled the localization of inflammatory mediators in order to ascertain the functional aspects of infiltrating inflammatory cells in HS, particularly highlighting the strong T h 17 polarity of inflammation in HS 3 . A schematic representation of the pathogenesis of HS based upon the findings of this review is presented in Figure 2. This highlights the interrelationship between inflammation and hyperkeratinization. Localization of TNF-α 13 , IL-12 32 , IL-23 32 and IL-32 4 , TLR2 40 , MMP2 13 , MMP8 43 and LL-37/cathelicidin 14,29 production to infiltrating dermal macrophages and lymphocytes as well as localization of IL-36 subtypes 5 , LL-37/cathelicidin 14,29 , IL-1β 32 and IL-22 32 to keratinocytes illustrate the feed forward mechanisms similar to those seen in psoriasis 9 and atopic dermatitis 10 which likely contribute to persistent inflammation in HS. Rather than keratinocytes being innocent bystanders, these IHC findings demonstrate the central role keratinocytes play as producers of key inflammatory mediators as well as mediators of products (such as TGF-β and ICAM) 41 that may contribute to fibroblast dysregulation and hypertrophic scarring 46 . A remaining unanswered question includes the temporal relationship between keratinocyte hyperproliferation and the activation of inflammatory cells infiltrating the dermis and epidermis in HS.

Insights into pathophysiology of HS
The current pathophysiological paradigm of HS is one of follicular infundibular occlusion leading to follicle rupture and a resultant inflammatory cascade 1 . This paradigm was based on the pivotal work of Shelley and Cahn in 1955 47 , whom demonstrated the induction of HS after application of belladonna impregnated tape to manually epilated axillae of 12 men. Only 3 of the 12 men developed the lesions described, and infection from the manual epilation procedure could not be excluded as a cause of the lesions, but this study enabled the paradigm to slowly shift away from one of apocrinitis, which had been in place since the original descriptions of the disease 7 . Detailed descriptions of infundibular hyperkeratosis (also termed poral occlusion) were made by Jemec et al. 7 and demonstrated the secondary involvement of apocrinitis in HS lesions. Jemec noted that poral occlusion was seen to occur alongside inflammation, but there was no suggestion of causation in one direction or another 7 .
Although individual cases of epidermal hyperkeratosis in the absence of inflammation are noted 7,33 , these cases are established or chronic lesions associated with significant fibrosis which is documented to be associated with reduce inflammatory infiltrate 7,33 . A consistent finding in all studies of this review is the co-localization of infundibular ORS keratinocyte hyperplasia with CD3, CD4, CD8 and CD68 positive inflammatory cells expressing TNF-α, IL-12, IL-23 and IL-32 4,13,32,40,43 . K19 is also documented as positive in the infundibulum suggesting keratinocyte hyperplasia [36][37][38] . However, it remains unclear whether keratinocyte hyperplasia induces the inflammatory cascade or if the inflammatory cascade induces the keratinocyte hyperplasia. The presence of inflammation in clinically normal, peri-lesional HS skin is well documented 4,30,33 implying the existence of a pre-clinical inflammation preceding symptoms of follicular occlusion. This is consistent with recent findings in acne pathogenesis that suggest that inflammation precede follicular hyperkeratosis and development of microcomedones 48 and is also pivotal in the ongoing development of nodulocystic acne and acne scars 49 . This pre-clinical inflammation is also consistent with the pathogenic paradigm in psoriasis and atopic dermatitis 9,10 with inflammation driving epidermal hyperkeratosis and alterations in keratinocyte maturation, consistent with the spongiform infundibulfolliculitis seen in established lesions of HS 50 . Our disparate findings in K19 staining in deep non-keratinized sinus tract epithelia with and without inflammation 37,38 also fit with this paradigm. In contrast, findings which would hold consistency with the current follicular occlusion paradigm would include infundibular occlusion preceding the development of inflammation, as well as alterations to desmosomal and hemidesmosomal proteins which would allow for rupture of the occluded follicles in order to drive the development of dermal inflammation and sinus tract formation. Although Danby et al. 51 reports reduced PAS positivity in the basement membrane zone at the sebo-follicular junction associated with inflammation in HS, it is likely that the reduced basement membrane integrity is secondary to inflammation and release of TGF-β and MMP2 52 (cytokines known to be altered in HS lesional skin and consistent with an abnormal wound healing response) rather than the follicular rupture being the primary driver of inflammation.
A more consistent hypothesis which accounts for the observed results of this review would be that of subclinical inflammation (due to a variety of triggers and immunological primers as illustrated in Figure 2) driving keratinocyte proliferation in the interfollicular epidermis and the follicular ORS, with follicular occlusion being a secondary phenomenon (mediated by TLR2 and IL-1α as documented in the development of comedones) 53 . The development of sinus tracts and hypertrophic scarring may also be mediated by the keratinocyte inflammatory response given the alterations in important wound healing mediators including TGF-β, ICAM-1 and comparisons by other authors of an altered wound healing response 8 in HS. This comparison would be appropriate given the high levels of dermal MMP2 13 and MMP8 43 ; the loss of keratinocyte maturation markers (K2e, K10, K19) 36-38 adhesion molecules (DG1 and DCN2) 38 in the non keratinized inflamed epithelium of the deep dermis; suppressed levels of ICAM-1 41 (seen impaired wound healing 54 ) and TGF-β 41 which leads to the dysregulation of TGF-β receptor ratio on fibroblasts which is linked with the development of hypertrophic scarring 46,54 seen in HS. These alterations to keratinocyte maturation are reminiscent of epithelial mesenchymal transition (EMT) 52 which may also explain the presence of free keratinocytes in the dermis in established lesions of HS 7, 36 . Indeed, as ICAM-1 is up-regulated Figure 2. Schematic Representation of Immunohistochemical findings in hidradenitis suppurativa. Immunological 'priming' occurs due to the contribution of adipose tissue, genetic susceptibility, smoking-related inflammatory mediators and obesity related pro-inflammatory signals and the composition of the microbiome. Increased activity of cDC1, cDC2 and T cells lead to both keratinocyte hyperplasia via the actions of IL-12 and IL-23, as well as a T h 17 predominant immune response. Alterations of antimicrobial peptides (AMP's) also occur throughout the epidermis. IHC staining localize Langerhan cells and activated dendritic cells to the epidermis and the dermo-epidermal junction. A population of epidermotropic CD8 T cells are also present. IHC staining indicates a mixed inflammatory infiltrate in the dermis, with contributions from Dendritic cells, B cells, T cells and plasma cells. Within sinus tracts, adhesion molecules are preserved, but inflammation in associated with non-keratinised sinus tracts leads to a loss of K19. The development of scarring and sinus tracts is associated with MMP2, ICAM-1 and TGF-Beta, with possible augmentation of ICAM-1 and TGF-B signaling via specific components of the microbiome. TNF-a, PGE2 and CXCL2 then lead to additional feed forward mechanisms perpetuating the inflammatory cycle. by pro-inflammatory mediators 54 , the low level of ICAM-1 noted appears paradoxical, however specific bacteria (including Porphyromonas species) which have been associated with HS 44,55 can suppress ICAM-1 production as an immune evasion strategy 56 . This implies that exogenous triggers (possibly including bacterial stimuli) can be a common cause for the initial inflammatory cascade as well as the development of tunneling and hypertrophic scarring in HS.

Conclusions
This systematic review of immunohistochemical staining of lesions in HS has highlighted the heterogeneity of studies and the methodological issues, which bring into question some of the results of IHC staining in HS lesions. The design of studies and variable reporting of potential confounding factors (such as ongoing or previous treatments) makes it impossible to compare staining intensity across studies. The results of existing studies suggest a florid inflammatory reaction comprising of T-lymphocytes, macrophages and dendritic cells with a strong Th-17 signature along with a keratinocyte mediated IL-36 inflammatory loop associated with keratinocyte hyperproliferation. The follicular occlusion paradigm as a primary driver of HS is unclear given the findings of this review and other histological and cytokine studies and inflammation as a primary driver of disease with secondary hyperkeratosis and occlusion is a plausible hypothesis.

Data availability
All data underlying the results are available as part of the article and no additional source data are required