Introduction

Proximal spinal muscular atrophies (SMAs) are a group of clinically variable motor neuron disorders characterized by degeneration of spinal cord anterior horn cells. SMAs are generally classified into types I to III according to age at onset and highest motor milestone achieved.1, 2 SMA type III is the most clinically variable form, with symptoms onset before (type IIIa) or after (type IIIb) age 3 years,3 normal achievement of motor milestones, variable severity of scoliosis, tendon retractions and joint contractures, and eventual loss of walking ability.

Type I–III SMAs are autosomal recessive conditions caused by loss of function of the survival motor neuron (SMN1) gene.4 Irrespective of phenotypic severity, both copies of the SMN1 gene are absent in about 95% of cases, whereas 2–3% of patients are compound heterozygotes typically with one allele deleted and subtle mutations in the other.5 Complete loss of the SMN protein is embryonically lethal,6 but SMA patients obtain reduced amounts of the protein from a nearly identical gene copy, SMN2, present (with SMN1) in a duplicated and inverted region of 5q13.4 Because of alternative splicing, most SMN2 transcripts lack exon 7 (SMN-del7) so that insufficient amounts of functional protein are produced. In fact, SMN protein levels are reduced in spinal cord and cell cultures from SMA patients, and correlate inversely with phenotypic severity.7, 8, 9 SMN2 copy number can also vary, and patients with high copy number often have a milder phenotype.10, 11, 12

At present, there is no effective treatment for SMA. Some therapeutic approaches aim to increase the amount of SMN protein produced by SMN2 through promoter activation, reduction of exon 7 alternative splicing, or both.13, 14, 15, 16, 17, 18, 19, 20, 21, 22 Some of these approaches are being investigated in ongoing or planned clinical trials, and great efforts have been done to identify the most appropriate clinical outcome measures for patients affected from various severities.23 In this view, it would be very useful to have reliable biomarkers of disease severity and response to treatment.23

In the present study, we investigated associations between clinical phenotype and molecular characteristics in adult patients with type III SMA, with the aim of evaluating available molecular biomarkers for possible use as surrogate endpoints in clinical trials on SMA. Clinical phenotype was assessed by tests of muscle strength and function. Molecular evaluation comprised determination of SMN2 copy number, SMN transcript levels (full length and del7), and SMN protein levels.

Materials and methods

Patients and clinical evaluation

A total of 45 patients (29 male, 16 female, Table 1), mean age 36.8 years (range 18–56 years) with diagnosis confirmed by molecular analysis were recruited to an ongoing double-blind placebo-controlled multicenter trial to assess the safety of salbutamol (EudraCT No. 2007-001088-32). All patients enrolled in the double-blind trial were included in the present study. At clinical evaluation, 26 were ambulant and 19 were wheelchair bound (mean age at loss of walking, 20 years). Based on age of onset, 15 were type IIIa and 30 were type IIIb. No patients reported onset of symptoms over 18 years of age. Written informed consent was obtained from all patients, and the study was approved by the Ethics Committee of each participating Centre.

Table 1 Selected clinical and molecular characteristics in a cohort of type III adult SMA patients
Full size table

Patients were comprehensively evaluated at baseline. Only selected variables are reported here as potential outcome measures. Muscle strength was assessed by manual muscle testing of 18 muscle groups (elbow flexors and extensors, finger flexors and extensors, thigh flexors and extensors, leg flexors and extensors, foot dorsiflexors) and graded from 0 to 5 according to the Medical Research Council (MRC) scale.24 The force of maximum voluntary isometric contraction (Newtons, N) was assessed in elbow flexor, handgrip, three-point pinch, knee flexor, and knee extensor25 for 30 of the 45 patients, using a hand-held myometer (CIT Technics, Groningen, The Netherlands).

In ambulant patients, motor function was assessed by the North Star Ambulatory Assessment (NSAA) scale.26 Ambulant patients also performed the 6-Min walk test (6MWT) recently shown to be reliable for assessing type III SMA patients.27, 28

Forced vital capacity (FVC, % of predicted) was measured in all patients using a standard spirometer in the sitting position. Lean body mass (grams) was assessed by dual X-ray absorptiometry (DXA)29, 30 and normalized to height (expressed in cm); this evaluation was performed in 20 patients only, in those neuromuscular Centres where the tool was available. Furthermore, DXA was feasible only for patients who did not have severe contractures preventing the access to the examination bed of the instrument.

Molecular assessments

Blood samples were collected into EDTA tubes for DNA extraction, sodium citrate tubes for protein extraction, and PAX blood RNA tubes (BD Biosciences, San Jose, CA, USA) for RNA. The samples were analysed at the Institute of Medical Genetics of Catholic University in Roma.

Genomic DNA was extracted by standard salting out, and conventional RFLP-PCR used to verify SMA diagnosis.31 For patients testing negative for SMN1 mutation by RFLP-PCR, SMN1 copy number was determined (same method as SMN2 copy number); for patients with a single SMN1 copy, sequence analysis of exons 1–7 and exon–intron boundaries was performed (sequence of primers and PCR conditions are available on request).

SMN1 and SMN2 copy number was determined by relative real-time PCR as reported elsewhere.14 SMN2 copy number was determined in all patients.

The presence of the p.G287R (c.G859C)32 variant in SMN2 was determined in all patients by RFLP-PCR. Briefly, 50 ng of genomic DNA were amplified with R1114 as forward primer and G287RDdeR: 5′-ATTTAAGGAATGTGAGCACCTTA-3′ as reverse primer. The latter contains a mismatch (bold) that introduces a restriction site for DdeI in the variant allele. Amplification conditions were: 30 cycles of 94 °C for 1 min, 55 °C for 1 min, 72° C for 1 min. The PCR products were digested with 3U of restriction enzyme DdeI overnight at 37 °C. Next day, the digestion products were separated by electrophoresis on 4% agarose gels. If the G287R variant was present, two bands (208 and 185 bp) were obtained.

RNA was extracted by PAX blood RNA extraction kit (Qiagen, Dusseldorf, Germany), according to the manufacturer’s instructions. SMN2 full length (SMN2-fl), lacking exon 7 (SMN-del7) and total (SMN-fl plus SMN-del7, SMN-tot) transcript levels were assessed by absolute real-time PCR (Tiziano et al33 and Angelozzi et al, in preparation). In patients with the G287R variant, full-length transcripts were determined by an alternative set of Taqman MGB probe and primers.33 GAPDH transcript levels were determined as quality control for RT-PCR and real-time PCR.

For SMN protein analysis, time between blood collection and preparation of samples ranged from a few hours to 2 days. Samples from 43 patients were analyzable. PBMCs were separated through Lympholyte M medium (Macherey-Nagel, Duren, Germany). The pellet was washed in PBS and frozen in fetal calf serum containing 10% DMSO. After thawing, PBMCs were counted by NucleoCounter (Chemometec, Allerod, Denmark) and resuspended in lysis buffer at 2 × 106 cells/ml (instead of 108 cells/ml, as in ELISA protocol, Enzo Life Science, Farmingdale, NY, USA); 2 × 105 cells were loaded onto each ELISA plate. The ELISA kits were kindly provided by the SMA Foundation. SMN protein concentrations were expressed as pg of protein/106 cells.

Statistical analysis

Means, medians, and SD for continuous variables and proportions for categorical variables were calculated. Associations of SMN2-fl, SMN-del7, SMN-tot transcript levels, and SMN protein levels, with clinical characteristics were assessed by linear regression models. A multivariate model was used to take account of the influence of other covariates. Because of small sample size and non-normal distribution of SMN transcript levels,33 the non-parametric Kruskal–Wallis ‘ANOVA’ by ranks (KW) and Mann–Whitney U-test (MW) were used to compare transcript levels between groups (ambulant vs non-ambulant; type IIIa vs type IIIb). Correlations between clinical characteristics were evaluated by Pearson’s r (R) test. The t-test for paired samples was used to compare the performance of groups at different time-points of the 6MWT. Statgraphics (Centurion XV.II, Statpoint Technologies, Warrenton, VA, USA) and SPSS 18.0 (SPSS, Inc., Chicago, IL, USA) were used to carry out the analyses. Differences associated with P<0.05 or, after Bonferroni correction for multiple testing, with P<0.016 were considered significant.

Results

Genotypic characterization of patients at the SMN locus

In 43 of the 45 patients, SMN1 exon 7 was absent. The remaining two patients were compound heterozygotes, missing one copy of SMN1, and with the W102X mutation34 in one case, and the S262I mutation35 in the other (Supplementary Figure 1a).

SMN2 gene copy number was determined in all patients. There were five SMN2 copies in 2 patients, 4 in 29 patients, three in 13 patients and a single copy in the patient with the S262I mutation. Among type IIIa patients, 7 out of 15 (46.7%) had three SMN2 copies, the others had 4 genes. Of the 30 type III b subjects, 21 had 4 SMN2 genes (70%).

The G287R variant (Supplementary Figure 1b) of SMN2 was found in 4/45 (8.9%) patients, all type IIIb with three SMN2 copies. One of these patients was homozygous for the G287R variant, being present in both parents.

Correlations between clinical characteristics

Selected baseline clinical features and molecular characteristics of the patients are shown in Table 1. Total MRC score correlated with handgrip (R=0.78, P<0.00001, n=30, data not shown), elbow flexion (R=0.68, P<0.00001, n=30, data not shown), knee extension (R=0.59, P=0.0006, n=30, data not shown), and knee flexion (R=0.74, P<0.00001, n=30, Figure 1a). Total MRC score correlated weakly with predicted forced vital capacity (R=0.28, P=0.06, n=45, Figure 1b). In ambulant patients, total MRC score correlated strongly with NSAA (R=0.77, P<0.00001, n=26, Figure 1c) and 6MWT (R=0.67, P=0.0002, n=26, Figure 1d). Distance covered during the sixth minute (mean 60.37±20.36 m) was significantly less (P=0.001, n=26) than in the first minute (mean 67.39±18.96 m).

Figure 1
figure 1

Scatter plots showing associations between total MRC score and (a) force of knee flexions in Newtons (n=30, R=0.74, P<0.00001); (b) Forced vital capacity (% of predicted) (n=45, R=0.28, P=0.06); (c) NSAA score (n=26, R=0.77, P<0.00001); and (d) 6-min walk test (meters, n=26, R=0.67, P=0.0002). Straight line: expected distribution; flanking lines, 95% confidence limits; black lines: limits of distribution. (a) and (b) refers to the whole cohort, whereas (c) and (d) to ambulant patients only.

Full size image

There was no correlation between muscle strength (total MRC scale score) or motor function (NSAA score, ambulant patients only) and patient age (MRC: R=−0.065, P=0.67; NSAA: R=−0.26, P=0.18). However, total MRC and NSAA scores (in ambulant patients) did correlate inversely with disease duration (MRC: R=−0.57, P<0.00001, n=45; NSAA: R=−0.48, P=0.01, n=26, Figures 2a and b). Forced vital capacity also correlated inversely with disease duration (R=−0.31, P=0.038, n=45; data not shown).

Figure 2
figure 2

Scatter plots showing associations between time from diagnosis and (a) total MRC score (n=45, R=−0.57, P<0.00001) and (b) NSAA score (n=26, R=−0.48, P=0.01); associations between lean body mass/height and (c) total MRC (n=20, R=0.66, P=0.0015), and (d) NSAA (n=11, R=0.71, P=0.006). Straight line, expected distribution; flanking lines, 95% confidence limits; black lines: limits of distribution. (a) and (c) refers to the whole cohort, whereas (b) and (d) to ambulant patients only.

Full size image

Correlations between motor performance and lean body mass

In all patients tested, lean body mass correlated with total MRC score (R=0.66, P=0.0015, n=20, Figure 2c). In ambulant patients, correlations of lean body mass with other aspects of motor performance were strong (MRC: R=0.82, P=0.0005, n=11; NSAA: R=0.71, P=0.006, n=11; 6MWT: R=0.69; P=0.009, n=11; Figure 2d and data not shown). Lean body mass also correlated inversely with disease duration (R=−0.50, P=0.025, n=20) and the correlation remained after correcting for height (R=−0.52, P=0.019, n=20, data not shown).

Associations between clinical and molecular data

By linear regression modeling, SMN2 copy number was unrelated to any clinical variable (in all cases P≥0.27, data not shown), or to SMN2 transcript or SMN protein levels (in all cases P≥0.08, data not shown). Similarly, neither SMN2 transcript nor protein levels were influenced by age, years from symptoms onset, or lean body mass (in all cases P≥0.12, data not shown). Transcript and protein levels did not differ between SMA types IIIa and IIIb (MW and KW, P>0.20, data not shown).

In the entire group, neither total nor delta7 SMN2 transcript levels correlated significantly with any clinical characteristic, although SMN2-fl levels correlated weakly with total MRC score (R=0.29; P=0.052, n=45, data not shown), as well as with lower limb MRC score (R=0.29; P=0.049, n=45, data not shown). In ambulant patients only, SMN2-fl levels correlated with total MRC score (R=0.46, P=0.02, n=26, Figure 3a), and with lower limb MRC score (R=0.49, P=0.01, n=26, Figure 3b), and weakly with 6MWT (R=0.37), although this correlation was not significant (P=0.07, n=26, Figure 3c). SMN protein levels did not correlate with motor performance (P≥0.31, data not shown) or with SMN2-fl levels (R=0.23, P=0.18, n=43, Figure 4a); however protein levels did correlate with the SMN2-fl/SMN2-delta7 ratio (R=0.40, P=0.016, n=43, Figure 4b).

Figure 3
figure 3

Scatter plots showing associations of SMN2-fl transcript levels in ambulant patients with (a) total MRC score (n=26, R=0.46, P=0.02), (b) lower limb MRC score (n=26, R=0.49, P=0.01), and (c) 6-min walk test (n=26, R=0.37, P=0.07). Straight line: expected distribution; flanking lines: 95% confidence limits; black lines: limits of distribution.

Full size image
Figure 4
figure 4

Scatter plots showing associations of levels of SMN protein in peripheral blood with (a) SMN2-fl transcript levels (n=43, R=0.23, P=0.18), and (b) the SMN2-fl/SMN2-delta7 ratio (n=43, R=0.40, P=0.016). Straight line, expected distribution; flanking lines, 95% confidence limits; black lines: limits of distribution.

Full size image

Discussion

Several potential therapeutic approaches to SMA are undergoing development or have been tested in recent years.36 Reliable clinical outcome measures and biomarkers are essential to effectively evaluate these approaches. Different motor function measures have been used and validated in SMA patients,23, 26 but some of them are too long to administer, include tasks unbearable for adult patients, or may be used only for patients with moderate phenotypes due to floor or ceiling effects. Moreover, some complications related to the disease, such as scoliosis, retractions, and weight gain, can further impair motor function. Thus, the identification of reliable biomarkers as surrogate endpoints of disease progression and response to treatment has become a matter of urgency.23 The aim of our study was to evaluate the applicability of SMN transcript and protein levels, as surrogate outcome measures in adult type III SMA patients, by relating clinical and molecular data. The clinical variables chosen have been previously validated or used in other SMA studies.23 The molecular techniques we used (absolute real time for transcript analysis, ELISA for protein quantification) are currently considered the most suitable tools for SMN quantification, as they do not make use of normalization against endogenous controls and are therefore unaffected by possible variations in the expression levels of housekeeping genes.

We found that clinical measures correlated strongly with each other, as expected. Similarly, Glanzman et al37 recently found that modified Hammersmith scale score correlated strongly with myometry-measured muscle strength. The motor performance was significantly affected by disease duration but not by age at evaluation, suggesting that in the design of clinical trials this variable could be useful to enroll more homogenous cohorts of patients, rather than age.

Montes et al28 recently evaluated the 6MWT in type III SMA patients spanning a wide age range and found that they showed progressive motor fatigue. We observed similar fatigability in the present series, so it may be also worth investigating whether increased resistance to motor fatigue can be used as a marker of treatment efficacy.

In the present study, we found no correlation between motor performance and SMN2 copy number. However, similar to our own previous data11 and those of others,10, 12, 37 a higher number of SMN2 genes was found in type IIIb patients, but was not predictive of the clinical phenotype in individual subjects. In the patient bearing the S262I mutation in exon 6, we found a single SMN2 copy, suggesting that this point mutation determines only a mild reduction in SMN protein function. This hypothesis is supported by the previous report of the same mutation in another patient,35 affected by SMA type III as well; however, in that patient, SMN2 gene copy number was not assessed. Also, the frequency of the G287R variant in our cohort was much higher than previously reported (about 9% vs 1%38). This variant has a positive effect in the inclusion of exon 7 in SMN2 transcripts.32, 38 All our patients bearing the G287R variant were type IIIb subjects, thus raising the prevalence of the mutation in this group of patients up to 16% (4/25 individuals). Because of the positive effect of this variant in exon 7 inclusion in SMN2 transcripts, this finding is not unexpected. It is noteworthy that these four subjects had three SMN2 copies, supporting the positive modulating effect of the G287R variant on SMA severity.

We also found a weak correlation between motor performance and SMN2-fl transcript levels, when considering all patients, which was much stronger in ambulant patients. It is likely that in non-ambulant patients, the presence of long-term complications of the condition further worsened motor performance. An alternative hypothesis is that SMN2-fl levels in blood do not reflect those found in target tissues of the disease, such as spinal cord and/or skeletal muscle. However, in our opinion, this hypothesis is less likely, as the correlation of transcript levels with motor function is stronger in the less severely affected patients. As no other transcript assessed (SMN2-del7, total SMN2 transcripts, or SMN2-fl/ del7 ratio) correlated with any baseline clinical characteristic, even in ambulant patients, SMN2-fl appears to have the strongest relation to phenotype. Very recently, some of us (FDT and LR) have collaborated to the BforSMA study.39, 40 Also in that large cohort of young patients spanning the whole phenotypic spectrum of the disease, SMN2-fl levels were significantly higher in the less severely affected subjects, although they were not predictive of the motor performance of single individuals. We found similar results also in our previous study.33 The main difference with the cohort included in the present study is related to the long duration of the disease of our patients and to the associated complications, which may impair the clinical evaluation. Moreover, to our knowledge, longitudinal data regarding SMN level variations with age are not available.

We also found that SMN protein levels were unrelated to baseline clinical characteristics and SMN mRNA levels, except for a weak correlation with the SMN2-fl/SMN-del7 ratio, whose biological significance remains unclear. Lack of correlation between SMN protein levels and motor performance was also found in the study of Crawford et al39 and remains unexplained. It is possible that SMN protein levels do not reflect those found in target tissues of the disease, or that the ELISA method we used requires optimization. It is noteworthy that stabilization buffers are not available for protein samples. On the other hand, for RNA extraction, these buffers allow to preserve samples for relatively long time and provide a ‘snapshot’ of gene expression at the time of sampling. As time between blood sampling and protein extraction (in the context of a multicenter clinical trial) varied considerably, levels of SMN protein could be affected by different variables, such as cell death, sample preservation, higher extractability of SMN protein over time, or post-translational modifications. Indeed, putative loss or increase in SMN protein levels hypothetically related to the factors above cannot be ruled out.

Our finding of strong correlations between several aspects of motor performance and lean body mass is potentially important and suggests that lean body mass, as measured by DXA, might be worth further investigation as an outcome measure in clinical trials on potential therapeutic agents in SMA. On the other side, DXA is not easily feasible in patients with severe contractures, the longitudinal variation of lean body mass in relation to age and disease course is at present unavailable, and the time required to observe a lean body mass increase in response to a potentially effective intervention is also unknown.

To conclude, the results of our study suggest that, in adults with type III SMA, SMN2 copy number, SMN2-del7 transcripts, and SMN protein levels in blood cells are not suitable as markers of phenotypic severity and hence as indicators of response to treatment. SMN2-fl transcript levels appear potentially more useful, as they correlate satisfactorily with motor performance in ambulant patients. Importantly we found that lean body mass shows promise as marker of disease severity and possibly also response to treatment. These findings require verification in larger series patients, of wider range of disease severity and age range (including children). Finally, our data suggest that if not taken into account, the confounding effect of disease duration may impair the identification of potential SMA biomarkers.