Multimodal characterisation of the SMN locus in SMA: copy number quantification and hybrid gene identification

Abstract

Spinal muscular atrophy (SMA) is a severe neuromuscular disorder caused by biallelic disruption of the Survival Motor Neuron 1 (SMN1) gene. Accurate quantification of Survival Motor Neuron 2 (SMN2) copy number is essential for patient stratification, prognosis, and treatment decisions, yet remains challenging at high copy numbers and does not fully explain phenotypic variability. We compared complementary molecular approaches in 78 Spanish patients with genetically confirmed SMA. Multiplex ligation-dependent probe amplification (MLPA) and digital PCR (dPCR) showed complete concordance, supporting their reliability for SMN2 quantification. In contrast, the AmplideX PCR/CE SMN1/2 Plus Kit showed discrepancies in seven patients, mainly at clinically relevant thresholds (three vs. four SMN2 copies), indicating higher sensitivity to technical variability. Notably, dPCR resolved a case with >5 SMN2 copies, demonstrating superior resolution. Long-read sequencing (LRS) in ten patients with MLPA-detected rearrangements enabled high-resolution reconstruction of the SMN locus and revealed six previously undescribed SMN hybrid structures. Structural haplotype architecture, rather than copy number alone, may further determine SMN variability and contribute to phenotype heterogeneity. Overall, our findings support an integrated diagnostic strategy in which copy-number quantification is complemented by structural characterisation in complex cases, improving molecular resolution of the SMN locus.

Introduction

5q Spinal Muscular Atrophy (SMA) is a neuromuscular disorder that mostly affects the lower motor neurons located in the spinal cord, leading to gradual weakness and muscle atrophy¹. This disease has an estimated incidence of 1 per 10,000 live births and a carrier frequency ranging from 1 in 40–60². SMA is traditionally classified into four clinical types (I–IV) based on age at onset and motor milestones achieved, with type I being the most severe and type IV the mildest³. Since the advent of new disease-modifying therapies (DMTs), a functional classification that groups patients into non-sitters, sitters, and walkers is preferred⁴.

The most common form of SMA is caused by pathogenic variants in the 5q13 Survival Motor Neuron 1 (SMN1) gene (HGNC:11117)⁵. This gene is mapped in a complex and repetitive region of the genome that has undergone at least one duplication event during evolution, resulting in the presence of a paralogous gene called Survival Motor Neuron 2 (SMN2, HGNC:11118). These features of the SMN locus make it more vulnerable to unequal homologous recombination and genomic rearrangements, including deletions, duplications, gene conversions and SMN hybrid genes⁶. As a result, homozygous deletions and gene conversion events account for 95% of the SMA patients, while the remaining 5% are due to other pathogenic variants in SMN1⁷.

Although SMN1 and SMN2 genes are nearly identical, they differ by 15 sequence changes known as paralogous sequence variants (PSVs)^8–10. Of these, 11 are located in intron six, one in exon seven, two in intron seven and the last one in exon eight. The only coding difference between SMN1 and SMN2 that determines their functional divergence is PSV13 (c.840 C > T), located in exon seven. This variant affects SMN2 pre-mRNA processing by promoting alternative splicing, leading to transcripts lacking exon seven (SMN2∆7) and resulting in production of a largely non-functional protein^9,11. Nevertheless, SMN2 retains the ability to produce small amounts of full-length, functional SMN protein, and therefore acts as a major modifier of disease severity in SMA.

SMN2 copy number is highly variable in the general population and, in SMA patients, higher SMN2 copy numbers are generally associated with milder clinical phenotypes, although this relationship is not absolute¹². Additional genetic modifiers influencing disease severity have also been described. For example, milder phenotypes have been associated with variants such as SMN2 c.859 G > C in exon seven, as well as intronic variants including c.835-44 A > G, c.835-549 A > G, and c.835-1897 C > T^10,13.

Historically, SMN hybrid genes were defined based on exon composition, describing SMN copies in which exon seven corresponded to SMN2 and exon eight to SMN1. The introduction of sequencing-based technologies enabled a more precise molecular definition of hybrid genes, identifying SMN copies carrying the SMN2-defining nucleotide T at position c.840 (PSV13) together with one or more SMN1-specific PSVs arising from gene conversion or unequal homologous recombination. This refinement has revealed a broader spectrum of SMN hybrid configurations at the SMN locus, extending beyond those detectable using exon-based approaches.

SMN hybrid genes have been proposed as potential modifiers of disease severity in SMA. However, their contribution to the phenotype remains unclear. While some studies have reported a higher frequency of SMN hybrid genes in patients with milder disease forms^14,15, others have not identified a consistent association^16,17.

Accurate quantification of SMN2 copy number is essential not only for determining the patient’s prognosis but also for guiding treatment selection. Classically, SMA has been diagnosed primarily by Multiplex Ligation-dependent Probe Amplification (MLPA) or quantitative PCR (qPCR)¹⁸. However, other technologies have been developed to improve SMN2 copy number quantification, detection of SMN hybrid genes, or variants in SMN genes, like digital PCR (dPCR), next-generation sequencing (NGS) and more recently long-read sequencing (LRS), among others^8,19,20.

In this study, we evaluate and compare multiple genomic technologies to identify the most precise and reliable approaches for SMN2 copy number assessment and for the characterisation of complex structural rearrangements and SMN hybrid genes, key genomic features of the SMN locus that may influence phenotypic variability, patient prognosis, and clinical outcome.

Results

Multiplex ligation-dependent probe amplification (MLPA)

Of the 78 SMA patients analysed by MLPA, 13 (16.7%) showed an SMN2 exon seven probe ratio of 0.80–1.20, consistent with two copies. Ratios of 1.30–1.65 (three copies) were observed in 46 patients (59%), while 18 (23.1%) had ratios of 1.75–2.15 (four copies). One patient (1.28%) exhibited a ratio of 2.20–2.65, indicating five SMN2 copies (Fig. 1A).

Fig. 1. Representative results from the MLPA analysis.

In the Coffalyser output, the lower X-axis represents the genomic region analysed, while the upper X-axis shows the corresponding probe name, which indicates its hybridisation site, along with the size of the resulting fragment. The Y-axis displays the final ratio obtained after normalisation. In the ratio chart, the 95% confidence interval over the reference samples is shown as a fixed coloured bar, blue for test probes and green for reference probes, whereas the 95% confidence interval for each sample is displayed as error bars surrounding the final probe ratio (dot). Yellow dots denote inconclusive results, red dots mark absence of signal, purple dots show that the arbitrary border has not been crossed, and blue dots indicate that the upper arbitrary border has been exceeded. Arbitrary borders (red and blue lines) are set at ±0.3 from the mean value of each probe in the reference samples. When a probe ratio in a sample crosses these borders, it may indicate a deletion or duplication, although not necessarily, as the variability of the probe in the reference samples may also cross them. A final ratio (dot) of 0 indicates a homozygous deletion (zero copies); 0.5 indicates a heterozygous loss (one copy); 1 corresponds to two copies; 1.5 to a heterozygous duplication (three copies); and 2 to a homozygous duplication (four copies). Except for the four probes specific to exons seven or eight of SMN1 or SMN2 (marked with the purple box), the remaining probes in the P021 kit recognise sequences that are present in both genes (marked with the orange box). In a diploid individual, these probes will therefore detect a total of four copies. In contrast, the probes specific to exons seven or eight of SMN1 or SMN2 will detect only two copies under the same conditions. A The MLPA analysis reveals five SMN2 copies in patient 78. B Patient 45 had three complete copies of the SMN2 gene and two SMN∆7-8 copy. C The diagram shows a normal copy of the SMN1 and SMN2 genes as well as the positions c.840 (exon seven) and c.1155 (exon eight) which distinguish the two genes. The partial deletion SMN∆7-8 are shown below, followed by a hybrid gene comprising exons seven of SMN2 and eight of SMN1.

Genomic rearrangements affecting SMN genes were identified in ten unrelated patients (12.8%) (cases 17, 18, 32, 33, 45, 56, 62, 63, 65, 66) (Fig. S1). A partial deletion involving exons seven and eight (SMN∆7-8) was found in one patient (Fig. 1B)^10,21 ; specifically, patient 45 carried three SMN2 copies and two SMN∆7-8 copies. In terms of functional copies, patient 45 was considered as three SMN2 copies. In the remaining nine cases, at least one SMN hybrid copy was identified, defined by an SMN2 exon seven signal, together with an SMN1 exon eight signal (Fig. 1C)^22–25. One hybrid copy was detected in patients 18, 33, 63, and 66; two in patients 17, 56, 62, and 65; and three in patient 32 (Fig. S1).

Amplidex PCR/CE SMN1/2 Plus kit

Through the AmplideX PCR/CE SMN1/2 Plus assay, 13 patients showed an SMN2 ratio of 0.663–1.079 (two copies), 43 patients 1.154–1.520 (three copies), and 22 patients >1.595 (four or more copies). Discrepancies with MLPA were noted in seven patients (19, 38, 53, 68, 69, 70, 73). After repeating the assay under identical conditions, two patients (69, 70) still showed discordant values, while five matched MLPA (Table 1).

Table 1.

Comparison of the SMN2 copy number results obtained using MLPA the AmplideX PCR/CE SMN1/2 Plus kit and digital (dPCR) in patients with discrepant results between methods

Patient	SMN2 ex 7 final ratio by MLPA	SMN2 CN MLPA	SMN2 ex 7 copies/ genome by dPCR	SMN2 CN dPCR	SMN2 ex 7 ratio AmplideX 1st result	SMN2 CN AmplideX 1st result	SMN2 ex 7 ratio AmplideX 2nd result	SMN2 CN AmplideX 2nd result
19	1.82	4	3.98	4	1.497	3	1.678	≥4
38	1.94	4	3.98	4	1.373	3	1.772	≥4
53	1.44	3	3.02	3	1.602	≥ 4	1.502	3
68	1.42	3	2.86	3	1.623	≥ 4	1.434	3
69	1.47	3	3.01	3	1.734	≥ 4	1.634	≥4
70	1.44	3	2.96	3	1.62	≥ 4	1.64	≥4
73	1.35	3	2.91	3	1.61	≥ 4	1.498	3
78	2.2	5	5.46	5–6	2.074	≥ 4	2.147	≥4

These discrepancies were observed in the first run, and in two patients, they persisted upon repeat testing. In the case of dPCR, the reported value corresponds to the average between the result obtained using the reference gene and the result obtained using the control sample for comparison. Patient 78 has been included in the table as the results emphasise the variation in sensitivity among the methods for detecting copy numbers greater than four.

CN copy number, Ex exon.

Beyond these seven patients, individual 78 displayed results that, although not deemed discordant, illustrated the differential sensitivity inherent to the methodologies used. By AmplideX PCR/CE SMN1/2 Plus assay, patient 78 showed an SMN2 ratio of 2.074 in the first replicate and 2.147 in the second replicate, indicating the presence of four or more copies. These values were considerably higher than those observed in the other patients with four or more copies (ranging from 1.64 to 1.772), which may suggest the presence of five copies. However, the software is unable to discriminate beyond four copies.

When evaluating AmplideX PCR/CE SMN1/2 Plus kit against MLPA, performance metrics varied according to the SMN2 copy number category (Table 2). For patients with two SMN2 copies (n = 13 by MLPA), concordance was complete, with all diagnostic parameters reaching 100% (sensitivity 13/13, specificity 65/65; 95% confidence interval (CI) ranges 75–100% and 95–100%, respectively). For patients carrying three SMN2 copies (n = 46 by MLPA), AmplideX demonstrated a sensitivity of 89.1% (41/46; 95% CI: 77–95%) and a specificity of 93.8% (30/32; 95% CI: 80–99%), with a positive predictive value (PPV) of 95.3% (41/43; 95% CI: 84–99%) and a negative predictive value (NPV) of 85.7% (30/35; 95% CI: 70–94%). Additionally, for patients with four SMN2 copies (n = 18 by MLPA), sensitivity reached 88.9% (16/18; 95% CI: 67–97%) and specificity 91.7% (55/60; 95% CI: 81–97%), while the PPV was more modest at 76.2% (16/21; 95% CI: 55–89%) and the NPV was 96.5% (55/57; 95% CI: 87–99%). Finally, for the single patient with five SMN2 copies identified by MLPA, sensitivity could not be meaningfully assessed since the kit cannot distinguish more than four copies.

Table 2.

Concordance between AmplideX PCR/CE SMN1/2 Plus kit and MLPA (gold standard) for SMN2 copy number

MLPA (gold)	Amplidex = 2	Amplidex = 3	Amplidex ≥ 4	Total
2 copies (n = 13)	13	0	0	13
3 copies (n = 46)	0	41	5	46
4 copies (n = 18)	0	2	16	18
5 copies (n = 1)	0	0	1	1
Total	13	43	22	78

MLPA classified patients into four categories (2, 3, 4 and 5 copies), while AmplideX reported results as 2, 3, or ≥4 copies.

In addition to SMN1 and SMN2 copy number, the AmplideX PCR/CE SMN1/2 Plus kit detected the c.859 G > C variant in SMN2 in two cases: patient 16 in homozygosity and patient 59 in heterozygosity state (Fig. S2). As expected, no patients with the c.*3 + 80 T > G and c.*211_*212del variants were found. Finally, no hybrid structure or partial deletions were detected in the cohort with this kit.

Digital PCR (dPCR)

The determination of SMN2 exon seven copy number by dPCR showed complete concordance with MLPA. In patient 78, MLPA suggested five SMN2 copies (ratio 2.20), whereas the AmplideX PCR/CE SMN1/2 Plus assay reported only ≥4 copies (ratios 2.074 and 2.147). Digital PCR yielded 5.59 copies per genome initially; across four further replicates the mean was 5.46 (range 5.14–5.71), consistent with five to six copies (Table 1).

On average, ~8,200 valid partitions per gene and sample were obtained, considered optimal under Qiagen (Hilden, Germany) protocols. No positive partitions were observed for SMN1 exon seven, while SMN2 exon seven showed 800–1600 positives depending on variability and copy number. As probes were limited to exon seven of SMN1/SMN2, detection of hybrid genes or deletions outside these regions was not possible.

Long-read sequencing (LRS)

The ten patients with genomic rearrangements were clinically assessed and molecularly characterised using LRS (Figs. 2–5 and Table 3). Variants in SMN2 intron six, exon seven and intron seven were visualysed in Integrative Genomic Viewer (IGV) to evaluate their potential impact on exon seven inclusion in SMN2 transcripts and, consequently, on the clinical phenotype. Despite this comprehensive assessment, no clinically relevant variants were ultimately identified. Paraphase analysis further resolved four distinct haplogroups (S2-1, S2-2, S2-3, SMN2_del_exon78) according to Chen et al.²⁰ (Table 3).

Fig. 3. Characterisation of SMN2 copy number and SMN hybrid structures detected by LRS.

a Patient 17, b patient 18, c patient 32, d patient 32, e patient 45, f patient 56, g patient 62, h patient 63, i patient 65, and j patient 66. The distribution of paralogous sequence variants (PSVs) across haplotypes is shown: each row corresponds to a haplotype, while each column represents a PSV position. PSVs specific to SMN1 are indicated in dark blue, whereas those of SMN2 are shown in light blue.

Fig. 2. Representation of the expected values according to the fitted model and the observed values according to read counts in long-read sequencing (LRS).

a Patient 17, b patient 18, c patient 32, d patient 32, e patient 45, f patient 56, g patient 62, h patient 63, i patient 65, and j patient 66. Bar plots show the comparison between observed (blue) and expected (green) read counts for different SMN2 haplotypes. Magenta dots indicate the estimated copy number of SMN2 in SMA patients.

Fig. 5. Visualisation of LRS results in patient 45 using integrative genomic viewer (IGV).

A 6.3 kb deletion was identified, encompassing from chr5:70.072.862 to chr5:70.079.171 (GRCh38/hg38).

Table 3.

Clinical and genetic data illustrating the variability in disease severity among SMA patients with rearrangements

P	SMA type	Age at onset (m)	Maximum motor milestone achieved	Age at loss of walking with support (y)	Age (y)	Actual functional status	Haplogroup	PSV event	Total SMN CN by LRS	Total SMN CN by MLPA	Compatible results	Phenotype spectrum
17	IIb	16	Walking with support	4	44	Sitter	S2-1	H1	3	3	Yes	Expected
							S2-1	H2
							S2-3	SMN2
18	IIIa	18	Walking without support	NA	5	Walker	ND	H4	3	3	Yes	Expected
							S2-3	SMN2
							S2-3	SMN2
32	IIIa	24	Walking without support	26	33	Sitter	S2-2	H1	4	4	Yes	Weaker
							S2-2	H3
							S2-2	H3
							S2-2	SMN2
33	IIIa	15	Climbing stairs	37	38	Sitter	S2-1	H5	3	4	No	Weaker
							S2-1	H6
							S2-2	SMN2
45	IIb	18	Walking with support	Unknown	47	Sitter	smn2_del_exon78	SMNΔ7-8	4	5	No	Expected
							smn2_del_exon78	SMNΔ7-8
							S2-1	SMN2
							S2-1	SMN2
56	IIIa	36	Walking without support	28	29	Sitter	ND	H4	2	4	No	Weaker
							S2-3	H9
62	IIIa	24	Walking without support	9	73	Sitter	S2-1	H7	3	3	Yes	Expected
							ND	H8
							S2-1	SMN2
63	IIa	10	Sitting without support	NA	6	Sitter	S2-2	H1	3	3	Yes	Expected
							S2-1	SMN2
							S2-2	SMN2
65	IIIa	20	Climbing stairs	13	37	Sitter	S2-3	H4	4	4	Yes	Weaker
							S2-3	H4
							ND	SMN2
							ND	SMN2
66	IIIb	36	Climbing stairs	NA	36	Walker	ND	H4	3	3	Yes	Stronger
							S2-2	SMN2
							S2-3	SMN2

The table summarises the clinical information for each patient, including the identifier (P), clinical type, age at symptom onset (months), the highest motor milestone achieved prior to treatment, age at loss of supported ambulation (years), current age (years) and present functional status. It also provides the results of long-read sequencing (LRS), specifying haplogroups according to Chen et al.²⁰ and the result of the analysis for the PSV haplotype, classified as canonical SMN2 PSV haplotypes or PSV-defined hybrid structures (Hn), as well as the total SMN copy number (CN). The total SMN CN refers to complete SMN2 copies, hybrid copies and those with partial deletions, such as SMNΔ7–8. These results are compared with the total SMN CN obtained by MLPA to assess concordance. Clinical and genetic data are jointly evaluated to determine whether the phenotype aligns with that expected from the natural history (“expected”) or instead represents a more severe (“weaker”) or milder (“stronger”) presentation.

Ex exon, H hybrid gene, m months, NA not applicable, ND no data, No. number, P patient, y years.

By PSV genotyping, we identified nine recurrent hybrid structures (H1–H9) across the cohort (Fig. 4). Six structures (H2, H4, H6, H7, H8 and H9) are, to our knowledge, reported here for the first time. H4 was the most frequent (patients 18, 56, 65, and 66), followed by H1 (patients 17, 32, and 63); the remaining structures were each observed in a single patient (Table 3).

Fig. 4. Paralogous sequence variants (PSVs) haplotypes for SMN hybrid structures detected by LRS.

H1, H3 and H5 had been described previously^17,47. The remaining six structures have not been reported to date. PSVs of SMN1 are indicated in dark blue, while those of SMN2 are shown in a lighter blue. The positions of these PSVs (intron 6, exon 7, intron 7, or exon 8) are indicated at the top of the figure by a curly brace.

In seven cases, MLPA and LRS agreed on the number of full-length SMN2 and hybrid SMN copies (Table 3). Importantly, in our model, the Poisson estimator produced copy-number assignments that were largely concordant with MLPA, in contrast to the default Paraphase dosage calls.

Discrepancies between MLPA and LRS were observed in three patients: 33, 45, and 56. In patient 45, a 6.3 kb deletion spanning intron six to exon eight of SMN2, corresponding to the SMN∆7-8 structure¹², was detected (Fig. 5). Two SMNΔ7-8 copies were consistently observed by MLPA and LRS, while MLPA additionally identified three full-length SMN2 copies, whereas LRS detected only two.

In addition, in patient 33 MLPA quantified four total SMN copies, one SMN hybrid gene and three SMN2 copies. LRS identified three SMN copies, two of them were hybrid structures (H5, H6) and one was a full length or “canonical” copy of SMN2. Due to probe design limited to exons seven/eight, only H6 was detectable by MLPA.

In patient 56, MLPA detected four SMN copies, two of them hybrid, whereas LRS identified only two SMN copies, both hybrids (H4, H9), and no full length or “canonical” SMN2 copy.

Interestingly, only patient 56 was male. The remaining nine patients (90%) were females, a higher proportion compared with 65% of females in the whole cohort of patients with three and four copies. However, this difference was not statistically significant (RR = 4.93; 95% CI, 0.67–36.51, p = 0.084). All patients with four SMN2 copies in our cohort showed a rather weaker than expected phenotype, with an age at ambulation loss that was below the percentile 12.5 of that described for patients with four copies, adjusted by sex²⁶. In other words, although these patients showed a clinical presentation compatible with four SMN2 copies, they fall within the more severe range of the phenotypic spectrum described for this group. Conversely, patient 66 with three SMN2 copies presented a phenotype within the milder range of the spectrum and remained ambulant at 36 years of age (Table 3).

SMN2 copy number and phenotype

The clinical type of patients and SMN2 copy number determined by MLPA and dPCR were examined jointly. Among 13 patients with two SMN2 copies, 84.6% (11/13) were SMA type I. The remaining two, cases 59 and 16, showed milder phenotypes: patient 16, carrying the c.859 G > C variant in homozygosity, was type IIIb, while patient 59, heterozygous for c.859 G > C, was type II. Of 46 patients with three copies, 26 (56.5%) had SMA type II and 20 (43.5%) type III. Among 18 patients with four copies, 17 (94.4%) had type III and one (5.6%) type IV (Table 4). Patient 78, identified through newborn screening at another hospital, remains asymptomatic.

Table 4.

Summary of combined SMN2 copy number and clinical SMA subtype data in our cohort of 77 patients

SMN2 copy number	Type I			Type II		Type III		Type IV	Total
	Ia	Ib	Ic	IIa	IIb	IIIa	IIIb
2	3	8	–	1	–	–	1	–	13
3	–	–	–	14	12	13	7	–	46
4	–	–	–	–	–	6	11	1	18
Total	11			27		38		1	77

Patient 78 was excluded from the table due to the inability to assign a clinical type, given their asymptomatic status.

Discussion

Our comparative analysis of MLPA, AmplideX PCR/CE SMN1/2 Plus, dPCR and LRS for SMN locus characterisation highlights that no single method can comprehensively address all diagnostic needs in SMA. MLPA is considered to yield the reference or true-positive result, as it remains the established gold standard method for SMN2 copy-number quantification in routine clinical practice (typically one to five copies), combining high reproducibility, moderate cost, and wide availability.

dPCR provides similarly precise quantification and offers superior resolution when copy number exceeds five, although it requires dedicated instrumentation and higher per-sample cost. The AmplideX PCR/CE SMN1/2 Plus kit provides rapid results and additional variant detection, but its performance appears more sensitive to technical variables and tends to underperform in complex structural contexts or high copy numbers.

LRS currently represents the most comprehensive approach for characterising complex hybrid structures and detecting novel configurations. However, its cost, processing time, and bioinformatic complexity still limit routine clinical implementation.

Regarding the relationship between SMN2 copy number and clinical type, our results are consistent with findings from other Spanish cohorts and international studies^12,27. Phenotypic variability among patients with identical SMN2 copy numbers has been previously reported and was also observed in our cohort, particularly among individuals carrying three copies^28–30. Similarly, as described in previous studies, patients carrying two SMN2 copies together with the c.859 G > C variant, either in homozygosity or heterozygosity state, tend to present milder phenotypes than expected based solely on SMN2 copy number³¹.

Given the availability of multiple SMN2 copy-number quantification methods and the discrepancies reported between techniques and laboratories, we compared different technologies used in SMA genetic diagnosis to evaluate their relative accuracy in SMN2 copy-number determination^32,33.

Our findings demonstrate complete concordance between MLPA and dPCR in determining SMN2 copy number, further supporting their reliability for accurate quantification. For dPCR, protocol optimisation and the inclusion of reference controls with higher SMN2 copy numbers could further improve accuracy in patients with more than five copies. Although individuals carrying five or more SMN2 copies, such as patient 78, are generally expected to remain asymptomatic or paucisymptomatic even in the absence of treatment, a confirmed genetic diagnosis of SMA remains clinically relevant. Currently, no therapies are approved in Spain for individuals with five or more SMN2 copies. However, identifying these patients remains important for long-term clinical follow-up, natural history characterisation, and informed reproductive counselling.

In contrast, the AmplideX PCR/CE SMN1/2 Plus kit showed lower concordance, particularly in samples with intermediate copy numbers (three or four). The improvement observed after assay repetition suggests increased sensitivity to technical variables and operator handling. Additionally, the kit failed to detect the case with more than four copies of SMN2, due to inherent limitations of the analysis software.

Genomic rearrangements at the SMN locus were found in 12.8% of patients by MLPA, consistent with the recombinogenic nature of this region⁶. The most common partial deletion, SMNΔ7–8³⁴, whose breakpoints were defined by Ruhno et al.¹⁰, was identified in one individual (patient 45), carrying two SMNΔ7–8 copies. AmplideX PCR/CE SMN1/2 Plus and dPCR did not detect this deletion because probe targets lie within the deleted interval.

The remaining nine patients in our cohort with genomic rearrangements carried at least one SMN hybrid copy, in addition to one to three SMN2 copies. MLPA can identify hybrid copies involving exons seven and eight; however, hybrid structures affecting introns six and/or seven, such as H5 and H9, cannot be detected due to the absence of specific probes in that region.

No SMN hybrid were detected using the AmplideX PCR/CE SMN1/2 Plus kit. This is because the AmplideX PCR/CE SMN Plus analysis software evaluates exon and intron seven, but no exon eight, for each copy. In hybrids H1-H8, exon and intron seven were derived from SMN2, leading the software to interpret these copies as complete SMN2 alleles. In the case of H9, a portion of intron seven appears to derive from SMN1, while exon seven originates from SMN2, meaning that detection by AmplideX PCR/CE SMN1/2 Plus kit could theoretically be possible. However, the exact primer binding sites within intron seven are not disclosed by the manufacturer, limiting full assessment of detection capability.

Overall, hybrid SMN copy results obtained by AmplideX PCR/CE SMN1/2 Plus and MLPA may therefore be compatible. Similarly, dPCR did not detect hybrid structures, as it relies on probes specific to exon seven of SMN1 and SMN2. However, given its multiplexing capacity, dPCR could potentially be expanded to interrogate additional SMN regions^19,35–37. The high sequence homology between SMN1 and SMN2 remains a major challenge for designing paralog-specific primers and probes.

Although the contribution of SMN2 structural configuration beyond copy number remains under debate, it has been proposed as a potential modifier of clinical presentation⁶. In this context, LRS provided a higher-resolution view of the rearrangements initially detected by MLPA, enabling more accurate structural characterisation in ten patients and revealing six hybrid structures not previously described. Additionally, LRS enabled identification of the most common SMN2 haplogroups in European populations (S2-1, S2-2, S2-3, and SMN2_del_exon78)²⁰. We explored potential genotype–phenotype correlations within the LRS cohort, focusing on whether hybrid gene structures and/or specific haplotypes were associated with clinical presentations that were either milder or more severe than expected. SMN hybrids were more frequently observed in women, and approximately 40% of patients showed a more severe phenotype than predicted based on natural history. However, previous studies have associated the presence of hybrid structures with milder phenotypes^14,17, highlighting the current uncertainty regarding their clinical impact.

To investigate this further, we evaluated two complementary levels of genetic variation. First, PacBio LRS enabled assignment of SMN2 haplogroups (S2-1, S2-2, S2-3…), which capture broader sequence divergence across the locus. Second, we performed fine-scale structural characterisation across the critical intron six–exon eight region, defining additional variation by the detection of hybrid structures (H1, H2, H3…). Divergence from the canonical SMN2 configuration often reflected hybridisation with SMN1-like motifs, which may influence SMN2 splicing and/or expression.

However, no consistent phenotypic trend emerged when analysing these subgroups independently. Establishing robust genotype–phenotype correlations will likely require direct assessment of PSV composition at the haplotype level, analysis of larger patient cohorts, and functional studies quantifying the amount of functional SMN protein produced by each hybrid copy.

Discrepancies between LRS and MLPA were observed in three cases, particularly in total SMN copy number quantification and in the presence of complex hybrid structures. These differences likely reflect the intrinsic limitations of each approach. First, MLPA and other PCR-based methods have restricted probe designs that limit detection of certain rearrangements. In fact, we cannot rule out the possibility that the remaining patients may present complex structures not detected by MLPA, specifically if they do not involve exons seven and eight, as reported in previous studies¹⁷.

In contrast, LRS may underestimate SMN copy numbers when sequencing depth is suboptimal and the bioinformatic interpretation of long-read data favours minimal-copy-number solutions. The latter is precisely what occurs in patient 56, where LRS identified a 1:1 ratio between hybrid structures H4 and H9. This distribution is compatible with the presence of a single copy of each hybrid (the most parsimonious inference), although it would also support the existence of two copies of each, which aligns with the MLPA results. It is important to note that our method does not yield absolute copy-number estimates; rather, it provides relative haplotype proportions (in this case, 1:1). When these proportions are statistically evaluated using a penalised Poisson model, the simplest interpretation, namely, one copy of each haplotype, is preferred. In the case of patients 33 and 45, the total number of SMN copies was discordant between MLPA and LRS. Increasing LRS depth, for example, by targeting the SMN locus instead of the entire genome, and refining haplotype-assignment algorithms should improve concordance across technologies and yield more reliable SMN copy number estimates.

LRS provides comprehensive characterisation of SMN hybrid structures through haplotype-resolved assembly, although positional information depends on the sequencing strategy used^38,39. Strategies such as ultra-long reads, parental phasing or targeted enrichment can further enable resolution of the precise genomic configuration and formation mechanism of hybrid structures. In contrast, long-range PCR offers a simpler and cost-effective alternative for detecting SMN hybrids based on PSVs; however, shorter amplicons limit the detection of complex configurations and therefore restrict structural resolution^8,17.

In conclusion, combining multiple genetic diagnostic techniques, particularly in cases involving hybrid genes or high copy numbers, enables optimal genotype resolution, more accurate phenotype correlation, and informed therapeutic decisions-making. Although no gold-standard method currently exists for reliable detection of hybrid genes, SMN locus–targeted LRS represents a promising strategy to address this diagnostic gap and may support future clinical diagnostic workflows, particularly in complex cases.

Further research in larger and ethnically diverse cohorts will be essential to clarify the contribution of SMN hybrid copies and haplotype structure to phenotypic variability, and to evaluate their potential role as markers of disease progression and therapeutic response. Such efforts will help refine genotype–phenotype correlations and complement copy number–based assessment.

Methods

Patients and samples

A total of 77 Spanish individuals from 70 families with prior SMA genetic diagnosis at the Genetics Department of La Fe University and Polytechnic Hospital were retrospectively analysed. Between 2015 and 2021, 48 patients were diagnosed using MLPA and, after commercialisation of the AmplideX PCR/CE SMN1/2 Plus kit (Asuragen, Austin, Texas), 29 more between 2021 and 2024. An additional SMA case detected by neonatal screening at another hospital was referred for genetic confirmation and SMN2 copy-number quantification, resulting in 78 patients overall.

Patients were classified in SMA types as follows: SMA type I refers to patients who never achieved independent sitting, with symptom onset within the first two weeks of life (type Ia), between two weeks and three months of age (type Ib) or with onset between three and six months of age (type Ic); SMA type IIa for patients whose maximal motor achievement was independent sitting; SMA type IIb for patients who apart from sitting independently, achieved to stand or walk with support; SMA type IIIa for patients achieving independent walk, whose symptoms appeared before three years of age, SMA type IIIb for patients achieving independent walk, whose symptoms appeared after three years of age and SMA type IV for those in whom muscle weakness begins in the second or third decade of life^18,40. Patients’ functional status (walker, sitter and non-sitter) and World Health Organisation (WHO) motor milestones before the onset of DMT were also collected.

All patients or legal representatives signed informed consent for the genetic study and for the recording of clinical and genetic data in CUIDAME registry⁴¹. The study was approved by the Ethics Committee of Hospital La Fe and conducted in accordance with the Declaration of Helsinki. To unify and compare SMA diagnostic techniques, all patients were analysed by MLPA (reference standard), the AmplideX PCR/CE SMN1/2 Plus kit and dPCR (index tests). Genomic DNA (gDNA) was extracted using the QIAsymphony automated extraction system (Qiagen, Hilden, Germany), and its concentration was determined to be 10 ng/uL using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

Multiplex ligation-dependent probe amplification (MLPA)

MLPA protocol was conducted with the SALSA MLPA Probemix P021 SMA (MRC Holland, Amsterdam, The Netherlands) in 36 patients with A2 lot and 42 with B1 lot. The kit included four probes targeting sequences within exon seven or eight of SMN1 or SMN2 genes, and additional probes to detect sequences shared by both genes (four in A2 lot, 17 in B1 lot) and in other genomic regions as reference (reference probes). It also detects the Neuronal Apoptosis Inhibitory Protein gene (NAIP, HGNC:7634) and its pseudogene, located in the SMN locus. MLPA reactions were run in the ProFlex PCR system thermal cycler (Applied Biosystems, Foster City, CA, USA) following manufacturer’s instructions, and products analysed in the ABI 3500 genetic analyser (Applied Biosystems, Foster City, CA, USA) using ROX 500 as internal size standard. Data analysis was performed with Coffalyser software (MRC Holland, Amsterdam, The Netherlands), comparing with control samples carrying two copies each of SMN1 and SMN2. After normalising the signal of each probe, first against internal reference probes and subsequently against external control samples, the final ratio is obtained. In SMA patients, the SMN2 exon seven probe final ratio estimates copy number: 0.00–0.10 indicates no copies; 0.40–0.65 one copy; 0.80–1.20 two; 1.30–1.65 three; 1.75–2.15 four; and 2.20–2.65 five copies. In the control samples, the copy number of the other genes analysed by MLPA (NAIP in lots A2 and B2, and RAD17, SERF1B and GTF2H2 in lot A2) was not determined; consequently, they could not be used as a reference to estimate the copy number of these genes in SMA patients.

Amplidex PCR/CE SMN1/2 Plus kit

The AmplideX PCR/CE SMN1/2 Plus Kit (Asuragen, Austin, Texas) is a multiplex PCR assay designed to quantify exon seven copy number of SMN1 and SMN2 genes⁴². The kit also enables detection of gene conversions and the genotypic status of clinically relevant variants, including SMN1 c.*3 + 80 T > G and c.211_212del associated with 2/0 carrier status⁴³, as well as SMN2 c.859 G > C, linked to a milder phenotype¹³. It includes a calibrator and control for normalising area ratios and confirming calibrator performance in each batch run. PCR multiplex reaction was performed in the ProFlex PCR system (Applied Biosystems, Foster City, CA, USA) with manufacturer’s recommended reagent volumes. Following amplification, fluorescent PCR products labelled with HEX were analysed in the ABI 3500 genetic analyser (Applied Biosystems, Foster City, CA, USA), using ROX 1000 size ladder for calibration. Electropherogram analysis was conducted with AmplideX PCR/CE SMN Plus Analysis software v1.1.5 (Asuragen, Austin, Texas), which normalises all sample peak areas with the calibrator. The AmplideX PCR/CE SMN Plus Analysis software calculates normalised ratios for SMN1 and SMN2 and translates them into copy numbers. In SMA patients, an SMN1 ratio of 0–0.185 indicates zero copies. For SMN2, 0–0.24 indicates absence, 0.290–0.612 one copy, 0.663–1.079 two, 1.154–1.520 three, and >1.595 four or more copies. Gene conversions between SMN1 and SMN2 appear as hybrid peaks, detectable because the kit includes primers for exon seven and intron seven of both genes. A copy containing SMN1 exon seven with SMN2 intron seven, or vice versa, generates a hybrid peak: an SMN1 hybrid reflects SMN2-to-SMN1 conversion, while an SMN2 hybrid indicates SMN1-to-SMN2 conversion.

Digital PCR (dPCR)

This technique partitions the sample into thousands of reactions, each ideally with a single DNA molecule where PCR occurs. After amplification, partitions are classified as positive or negative depending on target gene presence. In SMA patients, exon seven of SMN1 and SMN2 were amplified, along with the ribonuclease P RNA component H1 gene (RPPH1, HGNC:19273) as internal control. Primers and probes for SMN2 exon seven and RPPH1 were previously designed^35,44 while those for SMN1 exon seven were designed in-house. All were synthesised at a primer:probe ratio of 3.6 (900 nM:250 nM) by Bio-Rad Laboratories (Hercules, CA, USA) (Table 5).

Table 5.

Summary of the sequence of the primers and probes, and the reaction mix they were part of

Target	Primer or probe	Sequence (5’→3’)	Reaction mix	Reference
SMN1 exon seven	Forward	GCTATCTATATATAGCTATCTATG	1	This study
	Reverse	TGTGAGCACCTTCCTTCTT		44
	Probe	6FAM-TTTGTCTGAAACCCT-3IABkFQ		This study
SMN2 exon seven	Forward	TCCATATAAAGCTATCTATATATAGCTATCTATA	2	44
	Reverse	TGTGAGCACCTTCCTTCTT
	Probe	6FAM-TGTCTAAAACCCT-3IABkFQ
RPPH1	Forward	CTTTGCCGGAGCTTGGA	1 and 2	35
	Reverse	GAGAGTAGTCTGAATTGGGTTATGA
	Probe	HEX-ACCTCACCTCAGCCATTGAACTCAC-3IABkFQ

The paralogous sequence variants (PSVs) that differentiate SMN1 from SMN2 in both intron six and exon seven have been underlined.

FAM 6-Carboxyfluorescein, HEX Hexachlorofluorescein, 3IABkFQ 3’ Iowa Black® FQ.

Each PCR reagent was used at the manufacturer’s recommended volume and concentration for 8500-partition plates (Qiagen, Hilden, Germany). Two digital PCR reaction mixes were tested per patient on the QIAcuity One Digital PCR System (Qiagen, Hilden, Germany) (Table 5). HindIII High Fidelity (10 U/µL, New England Biolabs, Ipswich, MA, USA) was selected as the restriction enzyme, as none of the genes analysed contained its recognition site. Several protocols were assessed, with optimal results obtained using 2 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 48 °C. Imaging parameters were set to 850 ms and gain 6. Data were analysed with QIAcuity Software Suite v.2.5.0.1 (Qiagen), which provides the number of valid, positive, and negative partitions for each gene. For 8500-partition plates, accepted-partition counts close to 8500 were considered valid. In SMA patients, SMN1-positive partitions were expected to be near zero. Copy numbers were expressed relative to the reference gene RPPH1 and a control sample with two SMN copies.

Long-read sequencing (LRS)

Long-read HiFi sequencing (15–20 kb circular consensus sequencing (CCS) reads) was performed in ten retrospectively selected patients previously identified by MLPA as carrying structural rearrangements at the SMN locus, including SMN hybrids and partial deletions (SMNΔ7–8). Seven sequenced patients were of European ancestry and three were of South American ancestry. Genomic DNA was extracted using standard procedures, quantified by Qubit fluorometry (Thermo Fisher Scientific), and quality-assessed via Femto Pulse (Agilent Technologies). Whole-genome HiFi libraries were prepared with the SMRTbell Prep Kit 3.0 (Pacific Biosciences) and sequenced on the Revio platform to a mean depth of 30× following the manufacturer’s recommendations.

HiFi reads were aligned to GRCh38/hg38 using pbmm2 and processed with Paraphase (Pacific Biosciences) to distinguish SMN1- and SMN2-derived sequences based on the nucleotide at position c.840 (SMN1: C; SMN2: T), enabling haplotype reconstruction and SMN haplogroup assignment. In this context, a haplotype represents the set of co-inherited variants belonging to a single physical SMN copy reconstructed from HiFi reads, whereas a haplogroup denotes an evolutionarily related cluster of such haplotypes, as defined by Chen et al.²⁰.

For clarity, we define a ‘PSV haplotype’ as the 16-site PSV signature profile for a single reconstructed SMN gene copy, obtained by genotyping the PSV positions listed in Table 6. We refer to recurrent PSV configurations that mark paralog-switch boundaries as ´hybrid structures’ (H1–H9), whereas SMN2 haplogroups (S2-1, S2-2, S2-3, etc.) were assigned according to Chen et al.²⁰.

Table 6.

Panel of paralogous sequence variants (PSVs) used for hybrid identification and classification

PSV	SMN1 position	Ref SMN1	SMN2 position	Ref SMN2	Gene location	cDNA position
1	chr5:70,950,049	T	chr5:70,074,624	C	Intron 6	c.835-1897
2	chr5:70,950,189	G	chr5:70,074,764	A	Intron 6	c.835-1757
3	chr5:70,950,192	T	chr5:70,074,767	C	Intron 6	c.835-1754
4	chr5:70,950,329	G	chr5:70,074,904	A	Intron 6	c.835-1617
5	chr5:70,950,340	T	chr5:70,074,915	C	Intron 6	c.835-1606
6	chr5:70,950,493	G	chr5:70,075,068	A	Intron 6	c.835-1453
7	chr5:70,950,966	G	chr5:70,075,541	A	Intron 6	c.835-980
9	chr5:70,951,092	A	chr5:70,075,672	C	Intron 6	c.835-849
10	chr5:70,951,392	G	chr5:70,075,972	A	Intron 6	c.835-549
11	chr5:70,951,463	T	chr5:70,076,043	C	Intron 6	c.835-478
12	chr5:70,951,897	G	chr5:70,076,477	A	Intron 6	c.835-44
13	chr5:70,951,946	C	chr5:70,076,526	T	Exon 7	c.840
14	chr5:70,952,094	A	chr5:70,076,674	G	Intron 7	c.*3 + 100
15	chr5:70,952,209	A	chr5:70,076,789	G	Intron 7	c.*3 + 215
16	chr5:70,952,674	G	chr5:70,077,254	A	Exon 8	c.*239

The data presented in this table are derived from Blasco-Pérez, L. et al.⁸ and Monani, U R. et al.⁹. Chromosomal coordinates have been updated to the GRCh38/hg38 reference genome. Chr chromosome, Ref reference.

Because accurate dosage quantification is essential for interpreting SMN2 hybrid architectures, we developed a dedicated probabilistic model by fitting a Poisson model adapted from CASMA⁴⁵ to infer copy number (nᵢ) directly from the number of haplotagged reads assigned to each reconstructed haplotype that operates independently from the copy-number values reported by Paraphase. This approach is consistent with previous long-read and pangenome frameworks supporting allele-specific copy-number estimation from read counts^45–47.

For each haplotype i, let Rᵢ denote the number of reads assigned by Paraphase, rather than using bulk coverage we assume:

$${\mathrm{R}}_{\mathrm{i}}\sim \mathrm{Poisson}\left({\mathrm{\lambda }}_{\mathrm{i}}\right),{\mathrm{with}\mathrm{\lambda }}_{\mathrm{i}}={\mathrm{\lambda }}_{\mathrm{o}}{\mathrm{C}}_{\mathrm{i}}$$

where Cᵢ ∈ {0,…,5} is the candidate copy number of haplotype i, and λ₀ is an unknown global rate representing the expected number of assigned reads per copy in this sample. Importantly, λ₀ does not encode genomic coverage, but rather the average number of phased reads expected per single-copy haplotype. We explored λ0 over a grid from 5 to ∑iRi+5 in 5-read increments. For each combination of candidate copy numbers {Ci} and λ0, we computed the penalised likelihood score:

$$\mathrm{S}={\sum }_{\mathrm{i}}\mathrm{logP}\left({\mathrm{R}}_{\mathrm{i}}\right|{\mathrm{\lambda }}_{\mathrm{i}}={\mathrm{\lambda }}_{\mathrm{o}}{\mathrm{C}}_{\mathrm{i}})-\mathrm{p}{\sum }_{\mathrm{i}}{\mathrm{C}}_{\mathrm{i}}$$

with a penalty p = 5 to discourage unnecessary inflation of copy number (i.e., over-fragmented solutions). The configuration (Ci,λ0) that maximised S was selected as the optimal solution.

This formulation guarantees that Cᵢ = 0 is disallowed when Rᵢ > 0, prevents haplotypes supported by few reads from being spuriously assigned Cᵢ ≥ 1, and yields a parsimonious total copy number per gene (SMN1, SMN2, SMNΔ7–8). The inferred copy numbers Cᵢ are reported as the observed copies, while the expected reads (λ₀ Cᵢ) are compared with the empirical counts to visualise model fit. As an independent verification, base-level depth of coverage across the SMN locus was computed using mosdepth, confirming findings of SMNΔ7–8 haplotypes.

The analysis was complemented by visualisation of exons and coverage.

To characterise the paralog identity and structural composition of each haplotype, we analysed the haplotype-by-variant matrix exported by Paraphase and genotyped a curated panel of PSVs^8,9 spanning intron six to exon eight (Table 6).

PSV trajectories were visualised per haplotype, enabling identification of canonical SMN1, SMN2 or SMN hybrid structures, and paralog-switching boundaries. Each haplotype’s observed bases were compared against SMN1 and SMN2 reference alleles, classifying each site as SMN1-like, SMN2-like, No Coverage, or Other. PSV haplotypes were then classified as “canonical SMN1”, “canonical SMN2”, or “hybrid structure”, based on their PSV signatures. A graphical representation of PSV trajectories was generated for each haplotype to visualise paralog switching and detect structural patterns such as hybrid configurations and Δ7–8 deletions. This PSV-based structural annotation complements the penalised Poisson copy-number estimation, enabling high-resolution, haplotype-resolved reconstruction of the SMN locus that aligns closely with MLPA and supports precise discrimination of hybrid structures.

Finally, each sample was visually inspected in Integrative Genomic Viewer (IGV) to confirm PSV assignment, haplotype consistency, adequate coverage and the absence of misalignment artefacts.

Hybrid structures inferred by PSV haplotyping were labelled Hn (n = 1–9) for clarity.

Author contributions

Conceptualisation: A.B.-S., G.G.-G., JM.M., E.A.; Formal analysis: A.B.-S., T.J., E.A.; Investigation: A.B.-S.; Clinical evaluation: I.O., K.A.-G, NC.Ñ.-G, I.P.-C, JF.V.-C.; Methodology: A.B.-S, S.G.-A, I.L, L.C.-V.; Software: A.B.-S., S.G.-A, I.L.; Supervision: G.G.-G., JM.M., E.A.; Writing-original draft: A.B.-S.; Writing-review & editing: A.B.-S., S.G.-A, L.C.-V., K.A.-G, NC.Ñ.-G, JF.V.-C, E.A., G.G.-G., JM.M.

Data availability

The data presented in this study are available on reasonable request from the corresponding authors. The R scripts used to generate all main and supplementary figures are available from the corresponding author upon reasonable request. Copyright is retained by the authors. The article and all figures comply fully with Springer Nature’s licensing and copyright policy.

Competing interests

K.A.-G; NC.Ñ.-G, I.P.-C and JF.V.-C have received fees for participating in consultancy and educational activities from Biogen, Roche Pharma, and Novartis Gene Therapies. The other authors declare no competing interests. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Supplementary information

The online version contains supplementary material available at 10.1038/s41525-026-00579-8.