Gene expression based molecular test proves clinical validity as diagnostic aid for the differential diagnosis of psoriasis and eczema in formalin fixed and paraffin embedded tissue

Patients and material sampling

FFPE samples for this retrospective study were derived from the Biobank Biederstein and consisted of 107 patients (psoriasis, n=27; eczema, n=48, disease subtypes, n=18 and cases with contradictory diagnosis n=14)). Final diagnosis of patients was determined by at least two dermatologists taking a comprehensive view onto the clinical picture, family history, laboratory parameters and histological assessment. Characteristics of all patients are shown in Table 1. Patients treated with immune-efficient medication prior to material sampling were excluded from the study. The study was approved by the local ethical committee (project number 2773/10 and 515/17). Generated data and codes can be provided upon request.

Characteristics of FFPE tissue and isolation of total RNA

Skin biopsies (Ø3-6mm) from patients with a potential diagnosis of psoriasis or eczema were taken under local anesthesia by a dermatologist and immediately fixed in neutral buffered formalin for 8-48 hours at room temperature or 4°C. Fixed biopsies were paraffin embedded according to standard procedures (one biopsy per paraffin cassette). Unstained FFPE skin biopsies (<10 years of age) were used to isolate mRNA. Here, tissue sections with a thickness of 10 μm were generated from each FFPE skin sample. The first sections containing both epidermis and dermis were trimmed (at least 30μM) and discarded to avoid usage of degraded material. mRNA was extracted from in total three FFPE sections using a magnetic bead-based purification method according to the manufacturer’s protocol (InnuPREP MP Basic FFPE RNA Kit, IST Innuscreen, Germany). The mRNA eluates were analyzed immediately after the extraction by real-time PCR using specifc primer and hydrolysis probes.

Primer design and Real-time PCR

For specific amplification of the genes of interests NOS2 and CCL27 as well as of the reference genes, TBP and SDHAF2, primers and hydrolysis probes were designed using the primer design tool PrimerBLAST from NCBI (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). Oligonucleotides were synthesized by biomers (Germany). Probes were labeled with an Atto 647N fluorophore at the 5’-end and quenched with a BHQ2 quencher at the 3’-end.

Primer and probe sequences were as follows (forward, reverse and probe): 5’- TTCCTGGTTTGACTGTCCTTACC-3’, 5’- T7TTCACTGTGGGGCTTGC-3’ and 5’- CGGGGAGGCAGTGCAGCCAGC-3’ for NOS2; 5’- ACCCTACAGCAGCATTCCTAC-3’, 5’- AGCTTGTCTGAGAGTGGCTTTC-3’ and 5’- ACCCAGCACTGCCTGCTGTACTCAGCTC-3’ for CCL27; 5’- ATGGATCAGAACAACAGCCTGC-3’, 5’- CATTGGACTAAAGATAGGGATTCCG-3’ and 5’-TCAGGGCTTGGCCTCCCCTCAGGGTGCC-3’ for TBP; 5’- CGGTGTCTACAGTGTTCTCGAC-3’, 5’- GATGTCACACTGAGCAAAGGAG-3’ and 5’- CGCTGATGCTTGCTCTGTCAAGGCACAGCC-3’ for SDHAF2.

Real-time PCR reactions were carried out in 96-well plates using a Thermo Fisher Scientific one-step RT-PCR master mix. Thermocycling and fluorescence detection were performed in a qTower³ (Analytik Jena, Germany) and amplification results were analyzed using the qPCRsoft 4.1 software (Analytik Jena).

FFPE-based molecular classifier (MC) built-up

The molecular classifier was built using L2 regularized logistic regression. The molecular classifier predicts the disease state (probability for psoriasis) directly from the normalized CT values of NOS2 and CCL27 normalized to the reference genes SDHAF2 and TBP. For classification the psoriasis and eczema subtypes are aggregated to their respective high level labels (eczema or psoriasis). The model was calculated using LinearRegression implementation from scikit-learn with an L2 regularization strength of 100. In addition, the uncertainty of the model is estimated based on the predicted probability (cases with a predicted probability between 45% and 55% are predicted as undetermined). The model was trained on 77 patients with a clear diagnosis of psoriasis and eczema (atopic dermatitis (AD) and nummular eczema (NE)) and achieved an AUC of 0.97 +/- 0.06 on the training data. Furthermore, the model achieved a sensitivity = 91% and a specificity = 91% on the evaluation set which includes the training data plus 18 additional patients with prediagnosed subtypes of psoriasis and eczema.

Correlation of FFPE-based MC probability and PCR results with clinical attributes

Pearson correlation coefficients were calculated to determine the correlation between the FFPE-based MC determined probabilities for psoriasis, the negative delta CT PCR results of CCL27 and NOS2 as well as histological parameters, morphology of the lesion, comorbidities, laboratory parameters and family history.

The FFPE-based molecular classifier (MC) precisely separates eczema from psoriasis and efficiently identifies subtypes of both diseases

To evaluate the potential of FFPE samples for molecular diagnostics, 76 FFPE samples of lesional psoriasis (n=27) and eczema (atopic dermatitis n=28; nummular eczema n=21) were analyzed by probe-based real-time PCR for the expression of NOS2 and CCL27. To calculate a FFPE-based molecular classifier (MC) for differentiation of psoriasis and eczema, a logistic regression model was fitted with a 5-times cross validation. This classifier was designed to predict the probability of psoriasis with a cut-off value of 55% (The range between 45% and 55% would no probability (undetermined). Test specificity for psoriasis was 100% with a sensitivity of 92% and an area under the ROC curve (AUC) of 0.97 +/- 0.06 (Fig. 1A). Two psoriasis patients were mis-classified, and two further psoriasis patients reached a probability between 45-55% and thereby could not be evaluated (Fig. 1B). In a next step, we used the FFPE-based MC to classify subtypes of psoriasis (psoriasis guttata n=5; psoriasis palmoplantaris n=1; generalized pustular psoriasis n=2; and psoriasis pustulosa palmoplantaris n=3) and eczema (dishidrotic eczema n=4; hyperkeratotic rhagadiforme eczema n=3) (Fig. 1B). Whereas all cases of psoriasis guttata, psoriasis palmoplantaris and psoriasis pustulosa pampoplantaris were diagnosed correctly by the FFPE-based MC, cases of dishidrotic and hyperkeratotic-rhagadiform eczema as well as generalized pustular psoriasis tended to be misdiagnosed.

• Download figure
• Open in new tab

Figure 1: The FFPE-based Molecular classifier (MC) dissects clinically clear cases of psoriasis and eczema and is also efficient in subtypes of both diseases.

A) The molecular classifier (MC) was trained on clear cases of eczema (atopic dermatitis (AD n=28), nummular eczema (NE, n=21) and psoriasis (Pso, n=27) using a logistic regression model with a 5-fold cross validation delivering the probability of psoriasis. CT values of NOS2 and CCL27 were normalized to the reference genes SDHAF2 and TBP. The graph indicates the dissection of eczema and psoriasis samples exemplarily for normalization B) FFPE-based MC determined probabilities for psoriasis in clear cases of atopic dermatitis (AD), nummular eczema (NE) and psoriasis (Pso). The grey are between 45%-55% indicates undetermined probabilities C) The FFPE-based MC was used to predict the probability for psoriasis in subtypes of eczema and psoriasis (dyshidrotic eczema, DHE n=; hyperkeratotic-rhagadiform eczema, HRE n=; psoriasis guttata, PG n=5; psoriasis palmoplantaris, PP n=1); generalized pustular psoriasis, GPP n=2; and psoriasis pustulosa palmplantaris, PPP n=2). A probability between 45% and 55% was defined as not evaluable.

FFPE classifier results correlate with clinical attributes of psoriasis and eczema

To understand if the determined probability for psoriasis also correlates with clinical attributes associated with psoriasis or eczema, Person correlation coefficients were determined (Fig. 2). The probability of psoriasis highly and positively correlated with the histological attributes ’dilated dermal capillaries’ (r=0.61), ‘acanthosis’ (r=0.51), ‘neutrophils’ (r=0.39) and ‘parakeratosis’ (r=0.33) (Fig. 2) – all hallmarks for the pathogenesis of psoriasis. This positive correlation was followed by a positive correlation of negative CT values of NOS2 (r=.43; r=0.33; r=0.23; r=0.25, respectively) and a negative correlation with negative delta CT values for CCL27 (r=-0.54; r=-0.5; r=-0.48; r=-0.22, respectively). In line, the probability of psoriasis and the negative delta CT for NOS2 correlated positive with a psoriasis type morphology (r=0.77 and r=0.56) and arthritis (r=0.34 and r=0.35), whereas a negative correlation was determined for an eczema type morphology (r=-0.77 and r=-0.56) and rhinoconjunctivitis allergica (RCA) (r=-0.47 and r=-0.28) (Fig. 2). The probability of psoriasis also correlated positive with the therapeutic efficiency of IL12/IL-23 (r=0.99) and TNF-a inhibition (r=0.56), however negative with the response to Fumarate (r=-0.34) of IL-17 inhibition (r=-0.74). No correlations of the laboratory parameters ‘Stap.aureus’ and specific ‘IgE’ as well as the family history of for psoriasis and eczema were identified.

• Download figure
• Open in new tab

Figure 2: The FFPE-based MC determined probability for psoriasis correlates with disease hallmarks.

The probability for psoriasis as well as the negative delta CT values for CCL27 and NOS2 were correlated with clinical attributes by Pearson correlation.

The FFPE-based MC supports diagnostic decision making in patients with inconsistent diagnosis

To further evaluate the potential of the FFPE-based MC to support diagnostic decision making, 14 patient samples with discrepant clinical and histological diagnosis were analyzed. Within this cohort, 9 patients were assigned a probability for psoriasis and 5 patients were classified as eczema (Fig. 3A). Patient 43, a close to 80 year old woman, developed shortly after a respiratory infection itchy, pustular reddening overall the whole body. Bacterial or viral infections were not detected. Extensive lesional skin at the thigh and knee were surgically excised. Histologically, the lesions presented with infiltration of neutrophils and isolated eosinophils, signs of spongiosis and in part parakeratosis. The patient was diagnosed with psoriasis pustulosa, however, acute generalized exanthematic pustulosis (AGEP) could not be ruled out. In fact, psoriasis pustulosa and AGEP can hardly be distinguished by classical histological examination. Our FFPE-based MC assigned this patient a probability of psoriasis of 74.49%. After initiation of Ixekizumab, the skin status of this patient improved supporting the MC-based classification (Fig. 3B).

• Download figure
• Open in new tab

Figure 3: The FFPE-based MC is able to support diagnosis of unclear cases.

(A) The FFPE-based MC was applied on clinically and histologically challenging cases of patients with a suspected diagnosis of psoriasis or eczema and subtypes of both diseases (n=14). Given patient numbers represent numerical numbering and not biobank IDs. B) Clinical pictures and histological H&E stainings of selected patients that were assigned a probability of psoriasis by the FFPE-based MC.

Patient 59, a mid-sixty male, presented in the clinic with reddish, hyperkeratotic, scaly and itchy skin lesions at both palmae accompanied by erythematous, heavily infiltrated plaques at the back, the capillitium and retroauricular. Histopathological examination showed psoriasiform acanthosis, hyperparakeratosis, a lymphocytic infiltrate and isolated plasma cells leading to the diagnosis of psoriasiform perivascular dermatitis, while clinically diagnostics was in favor for psoriasis. The FFPE-based MC assigned a probability for psoriasis of 99.53%. Skin lesions at the palms and body cleared after 6 months of methothrexate therapy (Fig. 3B).

Patient 62, a male at the end of his sixties, presented at the clinic with confluent, erythematous, crude-lamellar, scaly plaques at both palmae. Histologically, extended hyperparakeratosis, serum crusts, psoriasiforme epidermis hyperplasia combined with spongiosis and a linear lymphocytic infiltrate in absence of neutrophils and eosinophils was described leading to the diagnosis of hyperkeratotic, in part dyshidrotic hand eczema. The FFPE-based MC assigned this patient a probability for psoriasis of 8.25% and thereby a high likelihood of an underlying eczematous disease. Treatment with a topical corticosteroids and altitretinoin improved the skin lesions supporting the MC-based classification (Fig. 3B).

Patient 75, a woman at beginning of her sixties, has been suffering for years under recurrent itchy, sharply demarcated erythematous plaques, with in part lamellar scales and excoriations as clinically seen in both psoriasis and eczema variants. Histologically, a psoriasiform hyperplasia of the epidermis combined with hyperparakeratosis, hypogranulosis, elongated rete-ridges, papillomatosis and focal spongiosis in upper epidermis layers was found. A perivascular dense lymphocytic infiltrate with absence of neutrophils and eosinophils was identified. The patient was diagnosed with a nummular eczema. This is supported also by the FFPE-based MC that assigned a probability of 18.29% to this patient (Fig. 3B) and skin lesions improved with topical corticosteroids and UVB phototherapy.