Transmembrane protease serine 5: a novel Schwann cell plasma marker for CMT1A

Patient recruitment and consent

Institutional Review Board approval was obtained from University of Iowa and written informed assent/consent were provided by participants under a protocol approved by the ethics board of the NIH Rare Diseases Clinical Research Network (Protocol INC6601). Subjects with CMT1A were identified and evaluated in the Inherited Neuropathy Consortium (INC) clinic in the Department of Neurology at Iowa. Subjects were diagnosed with CMT1A on the basis of clinical evidence of sensory and/or motor peripheral neuropathy (including length-dependent sensory loss, weakness and atrophy of the distal musculature, and decreased deep tendon reflexes), nerve conduction studies, and confirmatory genetic testing for the PMP22 duplication in the subject or affected first-degree relatives. Subjects and normal controls provided two 6-ml EDTA-containing tubes of blood during their visit for plasma extraction performed by standard techniques20 which was performed within 15 min of the blood draw and stored in aliquots stored at −80°C. All subjects were examined clinically by investigators who were certified by the INC for the proper administration of the CMTNSv2, a validated 9 item, 36 composite score based on patients’ symptoms (three items), examination findings (four items), and electrophysiology (two items).21 CMTESv2 scores were also calculated which included the seven items based on patients’ symptoms and examination findings in the CMTNSv2 but excludes the physiological results. Thus, the CMTES has a maximum score of 28 rather than 36 points.21 These scores were then subjected to Rasch modification to generate CMTNS-R and CMTES-R.10 Subjects with CMT1B, CMT1X, and CMT2A evaluated at the University of Iowa were also included and provided serum samples in an identical manner.

Nerve conductions

Ulnar motor conduction velocities (MNCV) were performed by standard techniques22 with recording over the belly of the Abductor Digiti Minimi (ADM) with stimulation at the wrist and below the elbow. Maximum compound muscle action potential (CMAP) amplitudes were recorded using baseline to peak measurements in mV.

Plasma assays

Aliquots from the plasma samples were screened by Olink immuno-PCR profiling, by an in-house developed Imperacer immuno-PCR method for TMPRSS5, and by the Quanterix NfL methods. Olink (Uppsala, Sweden) profiling was performed using the manufacturer’s recommended protocols.23, 24 Initially, a subset of the samples was run on the Neurology Panel (92 proteins). Subsequently, the full set of samples was run on five Olink panels (neurology includes TMPRSS5 assay, neurology exploratory includes NfL assay, immuno oncology, immune response and inflammation panels [total 398 unique proteins]). Subsequent to the initial single Neurology Olink panel profiling, aliquots from the same samples were assayed by an in-house developed Imperacer immuno-PCR platform TMPRSS5 assay (Chimera Biotech; Dortmund, Germany) using the manufacturer’s recommended protocols and using R& D Systems anti-human TMPRSS5 mouse monoclonal catalogue number MAB24951, and Goat polyclonal Catalogue number AF2495 antibodies. Quanterix NfL assays were performed according to the manufacturer’s protocols using the NfL-Lite R kit on a single molecule array (Simoa TM) HD-1 instrument (Quanterix, Lexington, MA).

Statistical analysis methods

Data were analyzed using GraphPad Prism v7.03 (GraphPad, San Diego CA). Pearson correlation coefficients were calculated for each marker relative to neuropathy score (CMTES-R, CMTES-N, Ulnar CMAP, Ulnar MNCV). Correction for age effect was performed by partial correlation calculated using software package R v3.5.0 (Open source).

AUC curve program

The TMPRSS5 and NfL ROC curves and AUC calculations were generated using R packages ROCR and pROC (R v 3.5.0 Open source).

Cell culture experiments

S1625 rat Schwann cells were transfected with Sox10 siRNA (Ambion, 4390771) or control siRNA using Lipofectamine 3000 as described.26 RNA was isolated from three replicate cultures 48 h after transfection using Tri Reagent (Ambion) and relative levels RNA were determined quantitative RT-PCR using the Comparative Ct method.27 Primers for Tmprss5 and Actb were as follows: rTmprss5F: CAGTCTACTGTGCTGAGGAATG; rTmprss5 R: GAGACTTGCCCTGTATGTTAGG; rActb F: CACCCGCGAGTACAACCTTC; rActb R: CCC ATA CCC ACC ATC ACA CC.

Plasma proteins upregulated in Charcot-Marie-Tooth disease 1A (CMT1A)

An initial pilot screen of 15 CMT1A samples from cohort 1 against 10 control samples from control cohort 1 on the Olink Neurology panel had identified four candidate biomarkers that were elevated on average >twofold, P < 0.05 in CMT1A plasma, including TMPRSS5 (2.34-fold P = 0.017), NMNAT1 (4.87-fold P < 0.0001), LAT (2.03-fold P = 0.0083), and MANF (2.75-fold P = 0.0002) (Fig. S1). We then expanded the study by testing samples from both CMT1A cohorts 1 (n = 20) and 2 (n = 27), against control cohorts 1 (n = 20) and 2 (n = 21), on the five Olink panels (Neurology, Neurology Exploratory, Immuno-Oncology, Immune Response, Inflammation [total 398 unique proteins]). In all, 339 of the proteins were detectable in >75% of samples tested, including 91 out of the 92 in the Neurology panel. We identified only TMPRSS5 to be consistently elevated in both cohorts of CMT1A samples compared to both control cohorts (threshold >twofold, P < 0.05) (Fig. 1A). TMPRSS5 was elevated on average by 2.07-fold P < 0.0001 all CMT1A (n = 47) compared to all controls (n = 41), and also to comparable levels in the individual CMT1A cohorts compared to either control cohort. It should be noted that the Olink scale (NPX) is a log2-based scale. TMPRSS5 is a serine protease that is also known as spinesin.28, 29

NfL was elevated on average 1.58-fold P < 0.0001 for all CMT1A (n = 47) relative to all controls (n = 41), and to comparable levels in the individual CMT1A cohorts compared to either control cohort (Fig. 1B). This is comparable to the 1.76-fold P = 0.0001 for the CMT1A subgroup in the previous report,19 measure using the Quanterix SIMOA assay. No other proteins on the five Olink panels tested exceeded this threshold of >1.5-fold P < 0.05 for all the comparisons of individual CMT1A cohorts compared to either of the control cohorts.

The ability of TMPRSS5 levels to discriminate CMT1A patients from controls can be assessed using a receiver operator characteristic curve (Fig. 1C). The area under the curves was 0.913 for TMPRSS5 compared to a value of 0.81 for NfL (Olink data), demonstrating good specificity and sensitivity for discriminating TMPRSS5, and also NfL, from control levels.

To confirm the elevation of TMPRSS5 in CMT1A samples using an orthogonal assay, we developed an in-house Imperacer immuno PCR assay using commercial TMPRSS5 antibodies and tested the same CMT1A cohort 1 and control cohort 1 samples shown in Figure S1. A similar upregulation of TMPRSS5 was observed, average 4.21-fold P < 0.0029 (Fig. S2A), compared to the 2.34-fold P 0.0017 observed by Olink analysis (Fig. S1). A very good correlation was observed between Olink and Imperacer TMPRSS5 levels r = 0.964 P < 0.00001 (Fig. S2B).

To confirm the accuracy of the Olink NfL assay, we determined the correlation between the Olink and Quanterix SIMOA assays in 10 CMT1A cohort 1 and 5 control cohort 1 samples. A very good correlation was observed r = 0.9735 P < 0.0001 (Fig. S2C).

Correlation of TMPRSS5 and NfL with patient clinical data

The correlation of TMPRSS5 and NfL with age was determined (Fig. 2). No significant correlation was found for TMRSS5 (Fig. 2A), whereas as expected,19 a positive correlation was observed for NfL for both controls and CMT1A samples (r = 0.6568 P = <0.0001 and r = 0.4735 P = 0.0008, respectively) (Fig. 2B). While the level was elevated in CMT1A compared to controls, the rate of NfL increases with age from the slope of the correlation plots in is similar for both (Fig. 2B). No significant sex effect was observed for the upregulation of TMPRSS5 or NfL in CMT1A samples (data not shown).

The correlation of severity of neuropathy as assessed by the Rasch-modified CMTES score with TMPRSS5 and NfL CMT1A patient data from Figure 1 was determined. First, we used a binned analysis in which mild (CMTES-R < 5), moderate (CMTES-R 6–10), and more severe cases (11–15 and 16–20) were grouped (Fig. 3). For TMPRSS5 although all bins showed significant upregulation of comparable magnitude fold change (1.95- to 2.41-fold) compared to all controls, no significant correlation with CMTES-R was observed as none of the bins were statistically different P < 0.05 from each other (Fig. 3A). Second, the correlation between CMTES-R and TMPRSS5 for all samples was determined (Fig. 4A) and was not significant (r = 0.1975 P = 0.1833). The same pattern of upregulation of TMPRSS5 was seen in all the equivalent CMTNS-R score bins, and no significant correlation with CMTNS-R (all samples) was observed (r = 0.22 P = 0.15 data not shown). This suggests TMPRSS5 is chronically elevated in the progression of CMT1A. No significant correlation was observed between TMPRSS5 and the electrophysiological parameters CMAP and MNCV (all CMT1A samples r = −0.12 P = 0.45 and r = −0.22 P = 0.158, respectively – data not shown). As expected, following age correction, no correlation was observed between TMPRSS5 and CMTES-R, CMTNS-R, Ulnar CMAP, or Ulnar MNCV (data not shown).

In contrast, NfL, as expected,19 increased across the CMTES-R score bins (Fig. 3B). The two lowest score bins CMTES-R < 5 and 5 < 10 were not significantly different from the controls (1.11-fold, P = 0.9973; 1.38-fold P = 0.2048, respectively), whereas the top three score bins were significantly higher (1.57-fold P = 0.008, 1.68-fold 0.0001, 2.19-fold <0.0001, respectively). Also, the top score bin CMTES > 20 was significantly upregulated compared to the lowest score bin CMTES < 5 (1.98-fold, P = 0.03). There was a significant correlation between CMTES-R and NfL when plotted for all samples (r = 0.53 P = 0.0001) (Fig. 4B). The same pattern of increase in NfL was seen across CMTNS-R score bins, and a significant correlation was observed with all samples (r 0.54, P = 0.003 data not shown). Following age correction, the correlation of NfL with CMTES-R and CMTNS-R remained significant but with reduced r values (r = 0.38 P = 0.0081 and r = 0.3695 P = 0.0174, respectively). No significant correlation was observed between NfL and the electrophysiological parameter CMAP (all CMT1A samples r = 0.19 P = 0.24, after age correction r = 0.22 P = 0.168). An apparent positive correlation with NCV (all CMT1A samples r = 0.32 P = 0.0418) was not significant following age correction (r = 0.29 P = 0.070).

Specificity for CMT1A

CMT1A is the prototype of the demyelinating form of CMT, so we performed a pilot Olink Neurology panel analysis of samples obtained from two other types of demyelinating CMT: CMT1B caused by mutations in the Myelin Protein Zero gene,30 and CMT1X, caused by mutations in the Gap Junction beta1 gene.31 For comparison, we also included samples from two other axonal forms of CMT: CMT2A, caused by mutations in the Mitofusin 2 gene32 and CMT2E caused by mutations in the NEFL gene.33 The level of TMPRSS5 was determined for a subset of CMT1A and samples of the other CMT subtypes and controls (Fig. 5). The expected increase in TMPRSS5 in CMT1A was observed (average FC 2.4 P = 0.0001), but only a modest increase was seen for CMT1B and CMT1X samples (demyelinating forms) which was not statistically significant (FC 1.26 P = 0.1876 and FC 1.26 P = 0.2003, respectively). No significant increase was seen for CMT2A and CMT2E (axonal forms) (FC 1.11 P = 0.9647 and FC 1.11 P = 0.9156, respectively). Therefore, the elevation of TMPRSS5 may be specific to CMT1A and not elevated in all forms of CMT. As there is a wide range of different specific mutations with a range of disease severity in CMT1B and CMT1X, it is possible that patients with specific mutations within these subtypes may prove to have elevated TMPRSS5 levels. Further testing of additional patient sample with known specific mutations would be needed to investigate this.

Schwann cell-specific expression of TMPRSS

The tissue-specific expression of TMPRSS5 in Schwann cells was evaluated in several ways. First, profiling of human tissues performed by the Broad Institute (gtexportal.org) identified the highest levels in human peripheral nerve (i.e., tibial nerve) (Fig. 6A). Lower amounts are observed in salivary gland and some CNS tissues, but expression of TMPRSS5 was almost negligible in other tissues. Consistent with these data, cell sorting studies in mouse skin showed that expression of Tmprss5 was confined to Schwann cells.34 Many myelin-associated genes are regulated by Sox10 expression, which is an important transcription factor at all stages of Schwann cell development. We had previously reported on Sox10-dependent genes in the S16 Schwann cell line,35 and found that Tmprss5 is regulated by Sox10, which was confirmed by independent quantitative RT-PCR analysis. Sox10 binding by ChIP-seq was analyzed in both peripheral nerve and spinal cord,36 and shows binding of Sox10 to enhancers that are marked by the presence of histone H3K27 acetylation (Fig. 6B). Notably, there is strong Sox10 binding in both oligodendrocytes and Schwann cells to an enhancer located ~11 kb upstream of the gene in the rat genome. Next, we performed a Sox10 RNAi knockdown experiment in cultured S16 rat Schwann cells and demonstrated a substantial, over 90% reduction, in TMPRSS5 mRNA levels (Fig. 6C). Finally, after nerve injury, many SC-specific myelin genes like PMP22 decline dramatically as Schwann cells lose contact with degenerating axons, and then they are reactivated if nerve regeneration occurs. The Tmprss5 gene, similarly declines after nerve injury and then increases again as regeneration/remyelination occurs, indicating that TMPRSS5 is regulated by the Schwann cell differentiation program.37-39