Cytokine Signatures in Psoriatic Arthritis Patients Indicate Different Phenotypic Traits Comparing Responders and Non-Responders of IL-17A and TNFα Inhibitors

This study aimed to explore the dynamic interactions between 32 cytokines and biomarkers in Psoriatic Arthritis (PsA) patients to compare cytokine signatures of treatment responders and non-responders. Biomarkers were measured before and after four months of treatment in 39 PsA patients initiating either Tumor Necrosis Factor alpha inhibitor (TNFi) or Interleukin-17A inhibitor (IL-17Ai). Response to treatment was defined by the composite measure, Disease Activity in Psoriatic Arthritis (DAPSA). A two-component principal component analysis (PCA) was implemented to describe cytokine signatures comparing DAPSA50 responders and non-responders. The cytokine signature of TNFi responders was driven by the correlated cytokines interferon γ (IFNγ) and IL-6, additionally associated with IL-12/IL-23p40, TNFα, and CRP, while the cytokine signature of TNFi non-responders was driven by the correlated cytokines IL-15, IL-8, and IFNγ. IL-17Ai responders were characterized by contributions of strongly correlated Th17 inflammatory cytokines, IL-17A, IL-12/IL-23p40, IL-22 to the cytokine signature, whereas IL-17A and IL-12/IL-23p40 did not demonstrate significant contribution in IL-17Ai non-responders. Based on PCA results it was possible to differentiate DAPSA50 responders and non-responders to treatment, endorsing additional examination of cytokine interaction models in PsA patients and supporting further PsA patient immune stratification to improve individualized treatment of PsA patients.

Existing knowledge shows that stimulation of the innate immune response induces secretion of pro-inflammatory cytokines, including tumor necrosis factors-α (TNFα), Interleukin (IL)-6 IL-12, IL-22, and IL-23 from neutrophils, dendritic cells, monocytes, and macrophages in inflamed tissues [5]. Tissue inflammation leads to an upregulation of tissue-homing chemokines promoting the recruitment of additional inflammatory immune cell subtypes [6,7] activated by antigen-presenting cells [8]. The increased production and release of pro-inflammatory cytokines such as TNFα, interferon-γ (IFNγ), and IL-12/IL-23 [9] stimulate the activation and differentiation of the inflammatory T cell subset; T helper cell type 1 (Th1) and T helper cell type 17 (Th17) [10]. Th1 cells and Th17 cells are both considered to have a pathological effect in PsA through the secretion of IFNγ, IL-17, IL-22, etc. amplifying and sustaining inflammation in PsA [5]. The interactions between cells and mediators of the innate and the adaptive immune system inducing, amplifying, and sustaining inflammation in PsA are a complex and poorly understood process. Thus, researchers have been able to target treatment towards the immune systems to relieve PsA patients of many debilitating symptoms [11].
Medical therapies targeting the inflammatory pathway by the inhibitory effect on the inflammatory pathway of PsA include TNFα inhibitors (TNFi), IL-17A inhibitors (IL-17Ai) and IL-17A/F inhibitors, IL-12/IL-23p40 inhibitors, IL-23 inhibitors, and Janus kinase inhibitors (JAKi) [12]. The therapies have shown a tremendous impact treating PsA [13,14] significantly improving the degree of arthritis, enthesitis, and psoriasis, etc. [15]. However, cytokine inhibition still exhibits an incomplete effect in some patients [16], including side-effects such as TNFi-induced psoriasis in a subgroup of patients [17]. The diversity in clinical effects of available medical therapies demonstrates the need for a better understanding of the PsA immunopathogenesis and improved stratification of PsA patients prior to initiation of treatment.
Here we present follow-up data examining cytokine signatures of the immune response and inflammation associated biomarkers and cytokines in PsA patients initiating cytokine inhibition therapies, either TNFi or IL-17Ai due to active PsA. The primary objective of the study was to examine the interplay between cytokines associated with PsA immunopathogenesis with a two-component principal component analysis (PCA), i.e., describe cytokine signatures at baseline in responders versus non-responders of TNFi or IL-17Ai, aiming to explore differences in the immune response, that might explain the various effects of cytokine inhibition therapies. The secondary objective was to evaluate change to the cytokine signatures after four months of treatment.

Baseline Characteristics
Paired baseline and four-months follow-up data, including clinical and biomarker data, were available from 39 PsA patients initiating biological disease modifying anti-rheumatic drugs (bDMARDs); 11 TNFi responders, 9 TNFi non-responders, 7 IL-17Ai responders, and 12 IL-17Ai non-responders. Additionally, data on 10 PsA patients initiating the conventional synthetic disease modifying the anti-rheumatic drug (csDMARD) methotrexate (MTX) were available for ancillary analysis to compare baseline cytokine signatures of the bDMARD initiators with a baseline cytokine signature of a patient group that was close to treatment naïve (Table 1), i.e., describing the inflammatory response of PsA.

Biomarker Change after Four Months in TNFi DAPSA50 Responders and Non-Responders
In the PCA summarizing biomarker contribution to the immune mechanisms of TNFi responders, C1 explained an impressive 36.8%, while C2 explained 16.0% of the variance. Continuous contribution was seen from inflammatory cytokines IFNγ (C1L 0.704), IL-27 (C1L 0.662), and IL-6 (C1L 0.610), indicating the ongoing importance of these cytokines to the cytokine signature of TNFi responders ( Table 2). This is supported by the addition of TNFα (C1L 0.651) as an important contributing cytokine to C1. Thus, with contribution < 5% to the component, implied an increased significance to the immune response after TNFi. Comparison of baseline and four month follow up further implied the increasing importance of anti-inflammatory mechanisms possibly driven by IL-10 (CL1 −0.542) or IL-RA (CL1 0.704). Though, only IL-10 was inversely correlated to the remainder. Overall, IFNγ, IL-27, IL-6, IL-1RA fell within the same quadrant closely associated to TNFα and CRP (Figure 1). In TNFi non-responders, C1 explained 31.7% of the variance, while C2 explained 19.2%. As for TNFi responders, the continuous importance of IL-7 (C1L 0.817), TNFα (C1L 0.570), IL-22 (C1L 0.718), and IL-27 (C1L 0.581) ( Table 2) as a part of PsA immunopathogenesis in TNFi non-responders was evident in C1 and with the contribution of IFNγ (C2L 0.758) to C2. Additionally, an important inversely correlated contribution of the inflammatory cytokine IL-17A (CC −0.839) was demonstrated and IL-17A had evidently become a significant contribution to the immune signature of TNFi non-responders compared to the baseline contribution. Further, significant contribution from IL-1α (C1L −0.761) was added to C1, and from IL-12/IL-23p40 (C2L 0.597) to C2 (Table 2). However, IL-12/IL-23p40 loadings were inversely correlated to IL-17A in both components, possibly suggesting alternative pathways from the classical IL-23/IL-17 pathway associated with increasing importance of IL-17A. Overall, IL-17A and IL-1α were inversely correlated to most other cytokines of significance-TNFα, IL-27, and IFNγ-whereas both clusters were associated with IL-22 and IL-7 ( Figure 1).

Biomarker Change after Four Months in IL17Ai DAPSA50 Responders and Non-Responders
The PCA evaluating the immune signature in IL-17Ai responders at four month follow-up revealed that C1 explained 42.7% of the variance, while C2 explained 20.1% of the variance. Main drivers of C1 included Th1 cytokine IFNγ (C1L 0.861) and the induced cytokines IP-10/CXCL10 (C1L 0.935); the latter did not contribute significantly to the components at baseline, and neither did IL-8 (C1L 0.841). With C1 explaining twice as much as C2, IL-12/IL-23p40 (C2L: 0.589) and IL-17i (C2L: 0.755) are considered less prominent cytokines within the immune signature at four month follow-up compared to baseline suggesting a switch in the immune signature in IL-17Ai responders. The PCA evaluating biomarkers at four month follow-up in non-responders to IL-17Ai revealed C1 explaining 22.4% of the variance, while C2 explained 19.4% of the variance. Significant contribution to the C1 was provided by correlated inflammatory cytokines IFNγ (C1L −0.534), IL-12/IL-23p40 (C1L −0.712), and IL-8 (C1L −0.632). All three were not considered significant cytokines of the components at baseline. IL-12/IL-23p40 and IFNγ were additionally strongly correlated with TNFα (C1L −0.616) and within the same quadrant of the correlation plot ( Figure 2). Results imply a possible increasing importance of a Th1 and Th17 inflammatory response to the immune signature in DAPSA50 non-responders to IL17Ai. Additional changes to the biomarker signature at follow up included less significant contribution from inflammatory products IP-10 (C1L 0.143/C2L 0.003) and CRP (C1L 0.068/C2L 0.000) and the anti-inflammatory cytokine IL-1RA (C1L 0.273/C2L −0.209).

Baseline Characteristics of MTX Initiators
Ancillary analysis evaluating the baseline cytokine signature of PsA patients initiating treatment with MTX (Table S2 and Figure S1) was conducted to provide an analysis of cytokine signatures in a population that was close to treatment and bDMARD naïve, i.e., a baseline immune signature not subjected to strong immune modulatory therapies, as the study aimed to analyze inflammatory markers in blood plasma obtained before starting the medical therapy. C1 and C2 explained 28.9% and 17.5% of the total variance. C1 was driven by the strongest contribution from correlated cytokines TNFα (C1L 00849), IL-8 (0.928), IL-7 (C1L 0.975), and additionally with strong contribution from CRP (CL1 0.946). C2 was driven by correlated cytokines IL-27 (C2L 0.781) and IL-22 (C2L 0.604). While the contribution and close association between cytokines IL-22, IL-27, and IL-7 was a continuous pattern in all described cytokine signatures, the contribution and relationship between additional cytokines was that TNFα and IL-8 were most similar to TNFi initiators. However, with only little influence of IFNγ on both C1 and C2 of MTX initiators. Considering clinical parameters, the PsA patients constituting these subgroups were also comparable, with the majority being bDMARD naïve and with a tendency of shorter disease duration.

Discussion
The study aimed to characterize baseline cytokine signatures and interactions in PsA patients, evaluating possible phenotypic differences between responders and nonresponders to bDMARD therapies, TNFi, and IL-17Ai, and assess changes to the signatures over time. The PCA was implemented as an established tool to examine phenotypic traits found in the variance of retrieved data of the patient subgroups [27,28]. Homogeneous clinical phenotypes (Table 1) justified the comparison of patients on biomarker level.
Cytokines associated with PsA immunopathogenesis [29] demonstrated a significant contribution to individual PsA patient subgroups. Thus, phenotypic traits were revealed comparing DAPSA50 responders and non-responders of TNFi and IL-17Ai. It applied to all PsA patient subgroups, TNFi responders, TNFi non-responders, IL-17Ai responders, IL-17Ai non-responders, and MTX initiators that cytokines, IL-7 and IL-27, demonstrated a significant contribution to the cytokine signatures. The same pattern applied to IL-22. Though, IL-22 was not a significant contributor to the cytokine signature of TNFi responders. IL-7 plays a central part in T cell maturation and survival [30], providing a foundation for an increased number of specific T cell subtypes known to PsA. IL-7 has been found to effectively stimulate the production of IFNγ and TNFα in rheumatoid arthritis [23] and IL-17 in spondyloarthritis [31]. In the majority of PsA patient subgroups, IL-22 was strongly correlated with IL-27. IL-22 is considered a pro-inflammatory cytokine in the immunopathogenesis of PsA [9,32], while IL-27 is considered a polyfunctional cytokine with inflammatory properties affecting a wide range of cells, and anti-inflammatory properties promoting IL-10 [22]. Interestingly, IL-27 in both TNFi non-responders, IL-17Ai initiators, and MTX initiators was associated and correlated with IL-22, IL-17A, and/or TNFα at baseline suggesting an inflammatory influence of IL-27 in PsA patients supported by a previous review on IL-27 in a PsA patient cohort [33].
Examining phenotypic differences indicated the importance of the Th17 inflammatory pathway, including IL-12/IL-23/IL-17 signaling. However, the revealed fascinating traits possibly explain why IL-17Ai non-responders did not achieve the DAPSA50 treatment response. Mechanisms including TGFβ and IL-6 [34] in association with IL-23 [35] are wellknown inducers of Th17 cells, in which research indicates that IL-23 might be responsible for the pathological effect of Th17 leading to IL-17A production and inflammation [36]. Both IL-12/IL23p40, IL-17A, and IL-6 did not contribute significantly to the immune signature of IL-17Ai non-responders (Table 3). On the contrary, the study demonstrated the close association between IL-12/IL-23p40 and IL-17A in TNFi and IL-17Ai responders; although, only with strong contribution to the components in IL-17Ai responders. These results suggest differences in cytokine phenotypes and possibly the need for different medicinal therapies in individual PsA patient subgroups.
Despite this, Th17 associated inflammatory mechanisms are the main postulated drivers of PsA. Results from this study further implied the contribution from Th1 associated inflammation as well with closely correlated biomarkers TNFα and CRP, and IFNγ together with IL-12/IL-23p40 [37,38] in MTX initiators, TNFi and IL-17Ai responders. IFNγ is known to promote Th1 cell differentiation and inhibit Th17 cell differentiation [39,40], indicating the continued importance of the Th1 inflammatory pathway in PsA immunopathogenesis. This is supported by recent evidence from our group implicating different immune cellular phenotypes in PsA, which additionally suggested that the Th1 driven phenotype was associated with shorter disease duration [41]. This is in line with the current study revealing a strong contribution of TNFα and/or IFNγ, together with only little contribution from IL-12/IL-23p40 and IL-17A in MTX and TNFi initiators, i.e., PsA patient subgroups with a tendency of having shorter disease duration (Table 1) and treated with less bDMARD prior to the current treatment initiation. However, the significant contribution from IFNγ to all cytokine signatures (except IL-17i non-responders) may be attributed to T cell plasticity in which Th17 cells developed IFNγ-secreting properties [42]. These results suggest the importance of dynamic T cell functionality in PsA, which has also been proposed in previous research retrieving both Th1 and Th17 cells from the arthritic joint [43].
Comparing differences between baseline levels of biomarkers revealed only small differences and only between IL-17Ai responders and IL-17Ai non-responders. Besides the cytokines, IL-7, IL-10, and IL-8, already discussed, increased levels of VEGF-A, VEGF-C, MIP-1β, TARC, MIP-3α, and CRP in IL-17Ai responders were reported at baseline possibly mirroring increased inflammation in IL-17Ai responders. However, it is important to note that both angiogenesis associated biomarkers, VEGF-A and VEGF-C, the chemokine TARC, and CRP were all significant contributors to the inflammatory signature of IL-17Ai non-responders as well, indicating an overall influence of the biomarkers to the signatures of PsA patients initiating IL-17Ai.
Data from the four-month follow-up visits confirmed changes to the cytokine signatures, which differed depending on the response to treatment. The continued substantial contribution of inflammatory cytokines might be explained by the DAPSA50 stratification as it is most likely that the subgroup of PsA patients improving~50% in DAPSA still have an on-going inflammatory response. Studies examining changes in cytokines, TNFα, IFNγ, IL-21, IL-6, IL-8, IL-10, and IL-17 during TNFi demonstrated an initial increase in cytokine levels [44][45][46] further, supporting the sustained contribution to immune signature revealed in the current study. Studies further imply the importance of considering whether four months is sufficient time to conduct a follow-up examination of true changes to cytokine levels due to varying levels of cytokines [44].
Limitations of the study include the definition of responders and non-responders based on DAPSA50 corresponding to minor improvement in disease activity [47]. It is assumed that the subgroups of PsA patients that did not reach minimal or low disease activity presented with an ongoing inflammatory process that is reflected in the cytokines measured. It is believed that inclusion of PsA patients reaching DAPSA70 or DAPSA80 would make phenotypic differences clearer than presented in the current study. Fourmonths follow up might be acknowledged as a limitation considering if four months are sufficient to evaluate changes in the clinical manifestations and the immune response. Four-month follow up was decided in accordance with international recommendations endorsing follow up after 3-6 months as this is the time where a clinical response would be expected or change of treatment would be considered [48].
An additional limitation is that the PsA patients in the study were compared independently of whether receiving bDMARD as monotherapy or in combination with additional csDMARD, and only 31.6% of included IL-17Ai initiators were bDMARD naïve. Previous bDMARD treatment and concomitant csDMARD treatment most likely had an immunemodulatory effect compared to the bio-naïve patients and the patients initiating bDMARD monotherapy, respectively. However, considering individualized treatment of PsA, results are still considered of great interest as patients initiating IL-17Ai mirror real-life difficult-totreat patients with high PsA disease activity no different to MTX and TNFi initiators, and the PsA patients in combination therapy still exhibit an inflammatory response associated with high disease activity. Additional examination of the influence of csDMARDs on the inflammatory immune response might be needed.
Moreover, the study only includes an examination of PsA patients initiating TNFi and IL-17Ai. It is important to acknowledge that several new medical drugs targeting other parts of the inflammatory immune pathway of PsA haven been approved for treatment of PsA, including IL-17A/F inhibitors, JAK inhibitors, IL-23 inhibitors, etc. We believe that it is necessary to study the influence of the individual drugs with different targets as it is expected that they will modulate the immune response differently. Immune modulation induced by the individual drugs is important to understand in order to provide individualized treatment of PsA. Lastly, the PCA is limited in that it does not include any value change in cytokine levels, but only describes cytokine interactions. Further, it was decided to use the two-component PCA not considering additional statistical approaches defining the number of components, which might require inclusion of additional components describing cytokine signatures of individual PsA patient subgroups. However, including the two components explaining the greatest degree of variance secured the comparability between subgroups.

Study Setup and Participants
Plasma samples were included from a consecutive clinical PsA patient cohort at the Parker Institute, Frederiksberg, Denmark [49], registered on clinicaltrials.gov (NCT02572700). Plasma samples and clinical data were obtained from 49 PsA patients with planned initiation of either TNFi (n = 20, 19 starting Adalimumab, 1 starting Certolizumab pegol), IL-17Ai (n = 19, 15 starting Ixekizumab, 4 starting Secukinumab), and MTX (n = 10). An additional inclusion criterium was established for the current study requiring paired biomarker data, including baseline and four-month follow-up data with a maximum of 10% of missing data for the individual biomarker to be included in the PCA.

Clinical Examination and Preparation of Blood
PsA patients participated in a baseline visit at treatment initiation and after four months of treatment. Clinical data collected comprised SJC, TJC, SPARCC enthesitis score, US assessing synovial hypertrophy, and PASI. Additionally, patient-reported visual analogue scale (VAS) global health was obtained to retrieve composite measure DAPSA quantifying disease activity. Peripheral blood was collected in EDTA vacutainer tubes (Greiner bio-one, Kremsmünster, Germany) and centrifuged to obtain plasma. Plasma was transferred and stored at −80 • C until analysis.

Analysis of Plasma Markers
The MSD V-Plex 54-plex kit (Mesoscale Discovery (MSD), Meso Scale Diagnostic, LCC, Rocksville, DC, USA) (Table S3). Thawed plasma samples were diluted in-plate in line with manufacturer's protocol except for panels, Chemokine panel 1, Cytokine panel 2, and the Th17 panel, which were diluted 2× established in an initial pilot. Plates were read on the MSD QuickPlex SQ 120 with software Discovery Workbench version 4.0.12. All samples were loaded in technical duplicates and measurements with intra-assay coefficient of variation (CV) ≥ 30 were excluded prior to data analysis. Biomarker absolute values were given in pg/mL.

Statistical Analysis
Based on exclusion criteria presented in Sections 4.1 and 4.3, 32 biomarkers (Tables 1 and 2) were included for the primary analysis. PsA patients were grouped based on the DAPSA50 response (yes/no), i.e., if patients reached a 50% improvement in DAPSA from baseline to follow up. Patient characteristics and biomarker level at baseline were reported as medians with corresponding interquartile ranges (IQR) for continuous variables and number with corresponding percentages for categorical variables. Differences between DAPSA50 responders and DAPSA50 non-responders were evaluated with Mann-Whitney U test and Chi squared test for continuous and categorical variables, respectively. Differences between treatment groups, TNFi, IL-17Ai, and MTX initiators, were evaluated with Kruskal-Wallis test. A two-component PCA was implemented for dimensional reduction to analysis biomarker signatures before and after treatment [50] examining biomarker signatures and the dynamic interactions between the 32 biomarkers. Prior to the PCA, missing data were imputed by the mean of the individual included biomarker variables. PCA components including corresponding eigenvalues were constructed to analyze variance within the dataset. C1 and C2 were retrieved for analysis of the biomarker signature with the largest amount of explained variance in principal component 1 (C1) and the second largest amount of explained variance in principal component 2 (C2). Correlation between individual biomarkers and the components was visualized in correlation plots with the computational configuration of the PCA ensuring that C1 and C2 were uncorrelated [51]. Loadings/correlation coefficients ranging from −1.0 to +1.0 were presented to assess the correlation of the individual biomarker to the components. Strong contribution to the component was defined by the PCA component loading being considered significant at >0.500. Additional subgrouping of results was included defined by contribution percentages > 5%. Statistical analysis was carried out using statistical software R version 4.1.2 with additional relevant packages.

Principal Component Analysis Correlation Plot
The correlation plots (Figures 1 and 2) were included to visualize the relationship between contribution and correlation of the individual biomarkers to the components. Further, to reduce the high dimensional data to easier understandable two-dimensional data. Uncorrelated components; C1 and C2 were represented by perpendicular axes. Degree of contribution was illustrated by colors black to red with black representing minor contribution and red representing strong contribution to the components. Distance between the vector of individual biomarkers demonstrated correlation between biomarkers. Closely associated variables within the correlation plot were positively correlated, while variables in the opposite quadrant were inversely correlated.

Conclusions
This study was conducted in an exploratory study setup that provided a broad assessment of the relationship between various cytokines in responders and non-responders to the biological therapies, TNFi and IL-17Ai, which demonstrated different cytokine signatures supporting the hypothesis of different phenotypes leading to diverse immune response mechanisms in PsA [52]. It is believed that interesting knowledge has emerged differentiating responders and non-responders of bDMARDs, which further endorses additional examination of the interactions of the dynamic immune response mechanisms in PsA patients to possibly support future individualized treatment strategies. Data Availability Statement: The data underlying this article cannot be shared publicly due to the privacy of the individuals that participated in the study. Data may be shared as part of research collaborations between participating institutions in line with GDPR and if approved by the Parker Institute and Danish authorities.