Journal of Neurology (2022) 269:3550–3562 https://doi.org/10.1007/s00415-021-10957-0 ORIGINAL COMMUNICATION Muscle MRI characteristic pattern for late‑onset TK2 deficiency diagnosis Cristina Domínguez‑González1,2,3 · Roberto Fernández‑Torrón4 · Ursula Moore5 · Carlos Pablo de Fuenmayor‑Fernández de la Hoz1 · Beatriz Vélez‑Gómez6,7 · Juan Antonio Cabezas6 · Jorge Alonso‑Pérez8 · Laura González‑Mera9 · Montse Olivé3,8,10 · Jorge García‑García11 · Germán Moris12 · Juan Carlos León Hernández13 · Nuria Muelas3,14 · Emilia Servian‑Morilla6,7 · Miguel A. Martin2,3,15 · Jordi Díaz‑Manera3,5,10,17 · Carmen Paradas6,7,16 Received: 8 November 2021 / Revised: 29 December 2021 / Accepted: 30 December 2021 / Published online: 14 March 2022 © The Author(s) 2022 Abstract Background and objective TK2 deficiency (TK2d) is a rare mitochondrial disorder that manifests predominantly as a progres- sive myopathy with a broad spectrum of severity and age of onset. The rate of progression is variable, and the prognosis is poor due to early and severe respiratory involvement. Early and accurate diagnosis is particularly important since a specific treatment is under development. This study aims to evaluate the diagnostic value of lower limb muscle MRI in adult patients with TK2d. Methods We studied a cohort of 45 genetically confirmed patients with mitochondrial myopathy (16 with mutations in TK2, 9 with mutations in other nuclear genes involved in mitochondrial DNA [mtDNA] synthesis or maintenance, 10 with single mtDNA deletions, and 10 with point mtDNA mutations) to analyze the imaging pattern of fat replacement in lower limb muscles. We compared the identified pattern in patients with TK2d with the MRI pattern of other non-mitochondrial genetic myopathies that share similar clinical characteristics. Results We found a consistent lower limb muscle MRI pattern in patients with TK2d characterized by involvement of the gluteus maximus, gastrocnemius medialis, and sartorius muscles. The identified pattern in TK2 patients differs from the known radiological involvement of other resembling muscle dystrophies that share clinical features. Conclusions By analyzing the largest cohort of muscle MRI from patients with mitochondrial myopathies studied to date, we identified a characteristic and specific radiological pattern of muscle involvement in patients with TK2d that could be useful to speed up its diagnosis. Keywords TK2 · Mitochondrial myopathy · MRI Introduction Mutations in this gene produce a deficiency of the enzyme thymidine kinase 2 (TK2), involved in the intramitochon- Recessive mutations in the TK2 gene lead to an ultra-rare drial biosynthesis of pyrimidine nucleotides. Consequently, mitochondrial disorder that manifests primarily as a pro- the balance in the nucleotide pool is disturbed, which gressive myopathy with early respiratory involvement [1–4]. interferes with the synthesis of the mitochondrial DNA (mtDNA), causing mtDNA depletion or multiple mtDNA deletions [5, 6]. Cristina Domínguez-González and Roberto Fernández-Torrón Clinical symptoms of this mitochondrial disorder can contributed equally to this work. start at any age, with a variable rate of progression. The factors that determine the severity of the disease are poorly * Jordi Díaz‑Manera jordi.diaz-manera@newcastle.ac.uk understood, with the age at onset and the presence of mtDNA depletion the generally accepted prognostic factors * Carmen Paradas cparadas@us.es [7, 8]. An early infantile-onset with very rapid progression and survival of fewer than 2 years is distinguished as the Extended author information available on the last page of the article 13 Vol:.(1234567890) Journal of Neurology (2022) 269:3550–3562 3551 most severe phenotype, with all these cases being associated Methods with severe muscle mtDNA depletion. Patients with later- onset have a more heterogeneous clinical course with a mean Patients survival of 13–23 years after the onset of initial symptoms [7, 8]. We recruited patients with the diagnosis of Primary Mito- The infantile and childhood-onset are the most frequent chondrial Myopathy (PMM), defined as an alteration of presentation, corresponding together to around 80% of mitochondrial oxidative phosphorylation of genetic origin patients with TK2 deficiency (TK2d) reported in the litera- that predominantly, but not exclusively, affects the skeletal ture [7, 8]. Patients who start their symptoms from the sec- muscle [11]. Patients came from 9 different Spanish cent- ond decade of life are the least recognized and are probably ers: University Hospital 12 de Octubre in Madrid, Uni- underdiagnosed. Taking all the clinical forms together, the versity Hospital Virgen del Rocío in Seville, University main manifestation is a progressive myopathy, which char- Hospital Juan Ramón Jiménez in Huelva, Hospital de la acteristically affects the facial, cervical, axial, bulbar, and Santa Creu i Sant Pau and Hospital de Bellvitge in Bar- respiratory muscles more prominently and, in some cases, it celona, University Hospital of Albacete, Central Hospital is associated with ptosis and/or ophthalmoparesis, the latter of Asturias in Oviedo, Hospital La Candelaria in Tenerife being more frequent in the later-onset type [9]. Other char- and Hospital Universitari i Politécnic La Fe in Valencia. acteristics of TK2d myopathy that can help to differentiate Clinical data were collected retrospectively from evalu- it from other mitochondrial myopathies are the presence of ations performed during the routine diagnostic process. high levels of CK (frequently > 10×) and the existence of In total, 45 patients were recruited for the study. dystrophic changes in the muscle biopsy, associated with Twenty-two patients with pathogenic variants in nuclear the common nonspecific signs of mitochondrial dysfunction genes involved in mtDNA synthesis or maintenance were (ragged-red COX-negative fibers) [3]. However, these fea- included: 16 patients with recessive mutations in the TK2 tures can complicate the diagnosis of TK2d myopathy erro- gene, 7 patients with recessive mutations in the POLG neously leading to the suspicion of a muscular dystrophy. gene, 1 patient with a heterozygous mutation in the TWNK We have previously analyzed the clinical features of a gene, and 1 patient with recessive mutations in the RRM2B series of 18 patients with late-onset TK2d, and identified the gene. In addition, 20 patients with primary alterations in existence of a possible pattern of involvement in the muscle mtDNA were also included: 10 patients with a single MRI scans performed in seven of these patients [3]. This large-scale deletion in mtDNA and 10 patients with a point pattern was characterized by the predominant involvement mutation in any of the mtDNA genes (4 patients had the of the gluteus major, semitendinosus, sartorius, and gastroc- m.8344A > G mutation in the MTTK gene, 3 patients the nemius medialis muscles in the lower limbs, which was the m.3243A > G mutation in the MTTL1 gene and 3 patients only examined region. had a mutation in the MTCO1, MTTS1, and MTTN genes, In this work, we extended the muscle MRI series of the respectively). All of them manifested predominantly with lower limbs of adult patients with TK2d to 16 and compare muscular symptoms, although not exclusively. The muscle them with the results of 29 patients with a diagnosis of pri- involvement was manifested as chronic external ophthal- mary mitochondrial myopathy due to different molecular moparesis (CPEO) (47%), ptosis (56%), facial weakness defects. We also compared the TK2d MRI pattern with that (33%), limb muscle weakness (31%), cervical weakness of other genetic myopathies that share some clinical char- (29%), dysphagia (20%), and respiratory insufficiency acteristics with TK2d, such as facioscapulohumeral dystro- (20%). The muscular symptoms and signs of each patient, phy (FSHD) due to the weakness of the facial muscles and as well as the molecular diagnosis, are detailed in Table 1. the oculopharyngeal muscle dystrophy (OPMD) due to the involvement of the ocular and bulbar muscles. Our objective is to evaluate the diagnostic value of the Muscle‑imaging acquisition and analysis muscle-imaging pattern to facilitate the identification of adult patients with TK2 mutations. Early and accurate diag- Muscle MRI studies were performed as part of the routine nosis is extremely important in this specific mitochondrial diagnostic process in a 1.5 T MR scanner at the evaluating disorder since a very promising therapy is currently under institution and were subsequently analyzed and scored by development, which could modify the natural history of the the same neurologist (R F–T), with experience in neuro- disease [10]. muscular disorders and muscle imaging. The evaluator was blind to the clinical manifestations and final diagnosis. We analyzed lower limb axial T1-weighted sequences and short-tau inversion recovery (STIR) sequences. Muscles 13 3552 Journal of Neurology (2022) 269:3550–3562 Table 1  Clinical characteristics summary Table 1  (continued) 1 TIM E MUSCULAR SYMPTOMS GENETIC DIAGNOSIS mtDNA variants were reference to the GeneBank number UPPER LOWER FR OM FACI NEC LIMBS LIMBS EXCE RESPI NC_012920.1 I DIS C RATOR GEN / PT AL K RCISE DYSP CK GENOM MUTATION (% of D EA P Y OS WEA WEA PRO DIS PRO DIS INTOL HAGI heteroplasmy in muscle if SE E INSUF E/RefSe IS KNE KNE XIM TA XIM TA ERAN A applicable and available) ON O FICIEN q1 SS SS AL L AL L CE SET CY (yea rs) from pelvis, thighs and lower legs were bilaterally scored 1 in axial T1-sequences with the semi-quantitative Mercuri 16 N mtDNA Single large-scale deletion 2 6 x2 mtDNA Single large-scale deletion 3 4 6 8 N N mtDNA mtDNA Single large-scale deletion Single large-scale deletion visual scale (MVS) modified by Fisher [12]: 0: normal 5 6 6 6 N X2 mtDNA mtDNA Single large-scale deletion Single large-scale deletion appearance; 1: mild involvement, less than 30% of individ- 7 36 X3 mtDNA Single large-scale deletion ual muscle volume; 2: moderate involvement, 30–60% of Single large-scale deletion 8 43 N mtDNA (55%) Single large-scale deletion individual muscle volumes; 3: severe involvement, > 60% 9 10 N mtDNA 1 0 11 N mtDNA (60%) Single large-scale deletion of individual muscle; 4: end stage, all the muscle is 1 1 14 X5 MTCO1 m.5992G>A p.(Gly30Asp) (44%) severely affected, replaced by increased density of con- 1 2 1 24 X2 MTTL1 m.3243A>G nective tissue and fat, with only a rim of fascia and neuro- vascular structures distinguishable. Total Mercuri score in 40 X2 MTTL1 m.3243A>G 3 1 30 N MTTL1 m.3243A>G a single patient was calculated by adding all values of the 4 1 6 X3 MTTK m.8344A>G 5 1 6 27 X6 MTTK m.8344A>G Mercuri score of all single muscles. 1 X1 11 MTTK m.8344A>G 7 0 1 X1 3 MTTK m.8344A>G 8 1 43 0 NA MTTS1 m.7497G>A (93%) Genotype‑MRI correlation 9 2 25 NA MTTN m.5692T>C (85%) 0 2 7 N POLG NM_00 [c.2243G>C p.(Trp748Ser )] ; We divided the cohort into TK2d patients and patients with 1 2693.2 [c.1760C>T p.(Pro587Leu)] POLG other mitochondrial conditions. To establish a pattern of 2 [c.914G>A p.Ser305Asn] 2 9 NA NM_00 2693.2 heterozygous muscle MRI fat replacement in time and correlations to POLG [c.752C>T p.(Thr251Ile) ; 2 3 7 NA NM_002 693.2 c.1760C>T p.(Pro587Leu)] ; [p.(Trp486Ter)] Medical Research Council's scale (MRC scale), hierarchi- 2 17 X2 - POLG NM_002 [c.2573C>T p.(Thr858Ile)] cal analysis, and graphical representation as a heatmap was 4 X2 heterozygous 0 693.2 POLG performed using R software, V.3.1.3. 2 [c.2573C>T p.(Thr858Ile] 8 X4 NM_002 5 heterozygous 693.2 POLG 2 6 13 N NM_002 693.2 [c.2864A>G p.(Tyr955Cys)] heterozygous Muscle MRI in TK2dvs other myopathies POLG 2 [c.2864A>G p.(Tyr955Cys)] 16 X2 NM_002 7 heterozygous 693.2 2 8 50 TWNK N NM_021 [c.1070G>C p.(Arg357Pro)] heterozygous We compared the muscle MRI pattern in TK2d with FSHD 830.5 2 13 RRM2B N NM_015 [c.329G>A p.(Arg110His)] and OPMD since they share some clinical characteristics 9 homozygous 3 713.4 TK2 [c.323C>T p.Thr108Met ] and might be included in the differential diagnosis. Access 16 X2 NM_004 0 614.5 TK2 homozygous to muscle MRI raw data from FSHD [13] and OPMD [14] 3 [c.604_606del (p.Lys202del)] 1 14 X4 NM_004 614.5 homozygous was given by their authors. Boxplots were generated for 3 4 NA TK2 [c.604_606del (p.Lys202del)] 2 NM_004 homozygous individual muscles for patients in four groups: (1): TK2d 614.5 3 3 18 X1 8 TK2 NM_004 [c.323C>T p.Thr108Met] homozygous patients, (2) other mitochondrial conditions, (3) FSHD, and (4) OPMD. The right limb was always used. Median MVS in 614.5 TK2 3 [c.604_606del (p.Lys202del) 31 X2 NM_004 the TK2d group was in turn compared to the median score in 4 homozygous] 614.5 TK2 3 [c.323C>T p.Thr108Met] 3 X6 NM_004 5 614.5 TK2 homozygous FSHD, OPMD, and the other mitochondrial conditions using 3 [c.604_606del (p.Lys202del)] 6 30 X2 NM_004 614.5 homozygous a Mann–Whitney test for unpaired samples. To demonstrate TK2 3 7 NA X3 NM_004 614.5 [c.323C>T p.Thr108Met] homozygous the key muscles which differ between groups, those muscles 3 8 38 TK2 N NM_004 [p. c.604_606del (p.Lys202del)] homozygous showing a significant difference were selected for display. 614.5 3 7 TK2 X2 NM_004 [c.623A>G p.Thr208Cys] To compare multiple variables on a similar scale, we used 9 homozygous 4 X2 614.5 TK2 [p. c.604_606del (p.Lys202del)] principal component analysis (PCA). Using this approach, we 12 NM_004 0 0 614.5 TK2 homozygous took all Mercuri scores for each muscle selected in each patient 4 [c.323C>T p.Thr108Met] and positioned that individual on a 2D chart of principal 13 X4 NM_004 1 homozygous 614.5 TK2 [c.415G>A p.Ala139Thr] ; components, in a way that maximizes the variation between 4 10 X2 NM_004 [c.40_41insGGGGC 2 614.5 p.Leu14Argfs*40] individuals. Each variable, therefore, becomes a vector along TK2 4 [c.623A>G p.Tyr208Cys] 30 X5 NM_004 3 homozygous 614.5 4 4 24 X1 6 TK2 NM_004 [c.623A>G p.Tyr208Cys] homozygous which an individual is positioned. Clusters in one area of the 614.5 4 40 X1. TK2 NM_004 [c.323C>T p.Thr108Met] chart or following a particular vector, indicate a group of indi- 5 5 homozygous 614.5 viduals with similar characteristics, while a group with distinct The colored cells represent the cases where the symptom or sign is characteristics form a separate cluster, which may be oriented present along a different vector line where variability is greatest for CK creatine kinase, NA not available, N normal 13 Journal of Neurology (2022) 269:3550–3562 3553 13 3554 Journal of Neurology (2022) 269:3550–3562 ◂Fig. 1  Heatmap of muscles of the lower limbs in patients with mito- In the entire cohort, the gluteus maximus, sartorius, gas- chondrial myopathies, disease duration, and MRC scale. A Heatmap trocnemius medialis, semitendinosus, semimembranosus, showing involvement of pelvic, thigh and leg muscles. Patients and muscles are ordered according to hierarchical clustering with increas- and tensor fascia latae were the most affected muscles, ing grading in fat replacement severity from the bottom to the top even though the fat muscle replacement was mild in many (patients—rows) and from the left to the right (muscles—columns). cases ((Median Mercuri (MM) 1, range: 0–4). The popliteal, The score of a muscle in a patient is indicated by the color of the tibialis posterior, and flexor digitorum longus were never or square. Grey squares mean that data are not available. The column in the top left contains information related to the genetic data of the seldomly affected (MM 0). patient (legend in the bottom left). We have also included a column with information about the muscle strength of the patients measured Genotype and muscle MRI pattern using the MRC scale in red and a column to the far right with infor- mation about the disease duration (time from onset of symptom to the MRI in years). B Heatmaps showing the median Mercuri score To study if there were different radiological patterns regard- of the muscles of the pelvis, thigh and leg areas. The three group ing the molecular diagnosis, we divided the cohort into of diseases and muscles are ordered according to hierarchical clus- three different groups: Group 1: patients with an mtDNA tering. The column in the top right contains information related to single large-scale deletion; Group 2: patients with a point the genetic data of the patient (legend in the bottom left). Group 1: mtDNA single large-scale deletion; Group 2: Point mtDNA muta- mtDNA mutation (MTCO1, MTTL1, MTTK, MTTS1, and tion (MTCO1, MTTL1, MTTK, MTTS1, and MTTN genes); Group 3: MTTN genes); Group 3: patients with mutations in nuclear Mutation in nuclear genes involved in mtDNA maintenance (POLG, genes involved in mtDNA synthesis or maintenance (POLG, TWNK, RRM2B and TK2 genes). No pattern of muscle involvement TWNK, RRM2B, and TK2). We found that muscle MRI was identified in any group involvement was more prominent in patients with muta- tions in nuclear genes (mean total MRC score: 23.6 ± 17.7) this second cluster of individuals. Those muscles showing a than in those with point mtDNA mutation (mean total significant difference between groups were used to generate a MRC score = 13.2 ± 12.7) and those with single large- three-principal component analysis. This approach was used scale mtDNA deletion (mean total MRC score = 2.4 ± 3.5) to compare the key muscles in TK2d to other mitochondrial (Kruskal–Wallis, p < 0.001)) (Fig. 1). Thereby, muscle MRI diseases, TK2d to FSHD, and TK2d to OPMD. To compare was normal in all but one patient with single mtDNA dele- if both FSHD and TK2d groups showed a similar degree of tion (group 1) in whom fat replacement signs in soleus, asymmetry, boxplots of individual asymmetry were gener- semimembranosus and long head of the biceps femoris were ated. These were plots showing the difference in Mercuri score detected. Among the ten patients with point mtDNA muta- between the left and right within each patient individually. tions (group 2), only three patients had a normal muscle MRI. Nevertheless, and as shown in Fig. 1, the fat replace- Statistics ment pattern was not consistent being gluteus maximus (MM 2), long head of biceps femoris (MM 2), semimembranosus We used the Shapiro–Wilk test to confirm that none of our (MM 1.5), and adductor magnus (MM 1.5) the most affected variables was normally distributed. As such, non-parametric muscles. In patients with mutations in nuclear genes (group statistic tests were used for the analysis. Kruskal–Wallis test 3), gluteus maximus (MM 3) was the most affected muscle, was used to compare quantitative variables across multiple followed by the sartorius, vastus lateralis, vastus medialis, groups while the Wilcoxon test was used for paired compari- tensor fascia latae, semimembranosus, semitendinosus, and sons of muscle fat infiltration. The χ2 test was used to compare gastrocnemius lateralis (MM 1). the qualitative variables. To investigate correlations between muscle function tests and MRI findings, Spearman’s rank cor- Muscle MRI pattern in TK2d relation was used (coefficient reported as ρ). The correlation was considered significant if the p value was less than 0.05 and We focused our attention on the muscle MRI pattern of ρ was 0.6 or higher. Statistical analyses were performed using patients with TK2 deficiency as this is the only mitochon- IBM SPSS Statistics, V.21 (IBM, Armonk, New York, USA). drial myopathy with a specific treatment that can change the natural history of the disease and, therefore, providing a correct diagnosis to the patients is extremely important. Results Gluteus maximus (MM 3), gastrocnemius medialis (MM 2), gluteus medius, sartorius (MM 2) and gracilis (MM Muscle MRI analysis 1.5) were the most affected muscles in patients with TK2 mutations. Gluteus maximus and gastrocnemius medialis All the 45 recruited patients were scanned. The mean age at were affected in all but one patient. At pelvic level, gluteus muscle MRI was 46.4 years (SD 15.8 years) and the mean- maximus and medius were more affected than gluteus minor, time from onset to MRI was 18.2 months (SD 12.9 months). while tensor fascia latae was moderately affected. In the 13 Journal of Neurology (2022) 269:3550–3562 3555 thighs, the gracilis, sartorius, adductor magnus were the and significant involvement of the gluteus maximum, sar- most affected followed by semitendinosus, semimembra- torius, gracilis, obturator externus, gastrocnemius media- nosus, and long head of biceps femoris. Muscles from the lis, and flexor digitorum longus muscles (Fig. 3A, B). In anterior compartment of the thigh were less affected. In the contrast, the rectus femoris, semimembranosus, and tibia- legs, the gastrocnemius medialis, gastrocnemius lateralis, lis anterior muscles were more involved in FSHD (Fig. 4). flexor digitorum, soleus, and peroneus longus were affected When analyzing muscle patterns with PCA, the gluteus whereas the tibialis anterior was seldomly replaced by fat. maximum, obturator externus, vastus lateralis, sartorius, We uncovered a consistent MRI pattern in TK2d patients, and flexor digitorum longus were the most distinctive mus- in which the gluteus maximus and medius, the sartorius, cles in TK2d compared to other mitochondrial diseases and the gastrocnemius medialis were the earliest affected (Fig. 5A) and OPMD (Fig. 5c). PCA analysis between muscles (Figs. 2, 3A, and supplementary Fig. 1). TK2d and FSHD confirmed that the latter were distinc- Correlations between total MRC and muscle MRI were tive muscles in TK2d while short head of biceps femoris, only statistically significant in adductor major (R – 0.64, p semimembranosus, extensor digitorum, and tibialis ante- 0.014), semitendinosus (R − 0.66, p 0.07), extensor digito- rior were those in FSHD (Fig. 5b). rum (R − 0.59, p 0.018) (supplementary table 1). Muscles from FSHD patients were more asymmetri- cally involved than TK2d when comparing right and left TK2d vs FSHD and OPMD MRI pattern comparison side muscles individually. This asymmetry was detected especially in adductor longus, the long and short head of When comparing to other mitochondrial diseases and to biceps femoris, rectus femoris, gracilis, and semitendino- other diseases that could resemble clinically, like FSHD sus in the thigh and tibialis anterior and gastrocnemius and OPMD, muscle MRI from TK2d patients had greater Fig. 2  Heatmap of muscles of the lower limbs in patients with TK2 indicated by the color of the square. Grey squares mean that data are gene mutations, disease duration, and MRC scale. Heatmap showing not available. The column in the top left contains information related involvement of pelvic, thigh, and leg muscles in patients with muta- to disease duration (time from onset of symptom to the MRI in years) tions in the TK2 gene. Patients and muscles are ordered according and the column in the top right provides information about muscle to hierarchical clustering with increasing grading in fat replacement strength of the patients measured using the MRC scale in red. Leg- severity from the bottom to the top (patients—rows) and from the left ends can be found at the bottom on the right to the right (muscles—columns). The score of a muscle in a patient is 13 3556 Journal of Neurology (2022) 269:3550–3562 Fig. 3  Muscle MRI pattern of the lower limbs in TK2d patients compared to other myopathies. A Muscle MRI from three TK2d patients with different disease duration and weakness severity showed that the most affected muscles were the gluteus maximus (arrows), the sartorius (arrowheads), and the gastrocnemius medialis muscles (asterisks). B Muscle MRI from patients with other mitochondrial myopathies as the one linked to point mutation in the mitochondrial gene MTTN, or other myopathies with similar clinical characteristics as FSHD or OPMD, showed a very different pattern of muscle involvement, highlighting the full preservation of the sartorius muscles (open arrows), even when other muscles are severely degenerated medialis in legs. No degree of asymmetry was detected in they must be distinguished from patients with other progres- the glutei (Fig. 6). sive genetic myopathies, a particularly challenging issue in adult patients, especially in those cases with no prominent ptosis or ophthalmoparesis (25% in our series). From a clini- Discussion cal perspective, the distribution of muscle weakness is criti- cal to suspect a TK2d. Most patients present a combination Mitochondrial myopathies are a wide group of muscle dis- of facial (with or without ptosis and/or ophthalmoparesis), eases characterized by heterogeneous clinical manifestations cervical and axial weakness, with early respiratory insuffi- and molecular causes [15, 16]. Reaching a genetic diagno- ciency requiring mechanical ventilation when they are still sis is crucial to genetic counseling and precision medicine. ambulant [3]. However, in some juvenile-onset patients, the Recessive mutations in the TK2 gene are responsible for a initial diagnostic suspicion may be FSHD, based on promi- syndrome of mitochondrial depletion/multiple deletions in nent facial weakness [18]. On the other hand, some patients mtDNA that can manifest at any age, most often as a pure have a very late onset with progressive CPEO and dysphagia progressive myopathy [2, 4, 7, 8]. In patients with TK2d, as the only symptoms for many years, and they can, there- oral treatment with thymidine and deoxycytidine nucleo- fore, be confused with OPMD. sides has shown to prolong survival in the most severe infan- In a previous work, we described the muscle involvement tile cases and halt the progression of the disease in the other in the MRI of the lower limbs of seven patients with TK2d, clinical forms [10, 17]. It is, therefore, essential to achieve an suggesting a characteristic pattern with potential diagnos- early genetic diagnosis in all patients with TK2d. To do so, tic utility [3]. In the present study, our results confirm the 13 Journal of Neurology (2022) 269:3550–3562 3557 Fig. 4  Boxplots showing values of the Mercuri score of muscles of per every comparison and when significant, Wilcoxon test is applied the pelvis, thigh and leg of TK2d, other mitochondrial myopathies, to compare TK2 and the other three diseases FSHD and OPMD patients. Value of the Kruskal–Wallis test is shown consistency of this pattern in a larger cohort of patients and We have applied a PCA test to the analysis of T1w imag- demonstrate its specificity when compared with the pattern ing data quantified using Mercuri score of two large cohorts of other myopathies, mimicking the clinical presentation of of patients with FSHD and OPMD patients already pub- TK2d, whether of mitochondrial origin or not. lished [13, 14] and have included the results of this cohort In this study, we demonstrated that among patients with of TK2d patients in the analysis, with the aim to (1) validate PMM, muscle fat replacement in the lower limbs is hetero- if there is a different pattern of involvement and (2) identify geneous in patients with mtDNA mutations, without a spe- which muscles can be useful to differentiate one disease cific pattern that distinguishes them as a group. Furthermore, from the other. We have observed that gluteus maximus, in the majority of patients harboring a single large-scale obturator externus, vastus lateralis, sartorius, and flexor mtDNA deletion in our cohort, muscle MRI was normal. In digitorum longus differentiate TK2 myopathy from FSHD the group of patients with nuclear gene mutations, we also and OPMD. In FSHD, involvement of the posterior compart- did not identify a characteristic pattern that allows us to dif- ment of the thigh is a common finding, although the most ferentiate them as a group. However, a hierarchical analysis frequently and severely involved muscle is the semimembra- identified that gluteus maximus, gluteus medius, sartorius, nosus, while in legs the most affected muscles are tibialis gracilis, and gastrocnemius medialis were the most affected anterior, soleus, and gastrocnemius medialis. In addition, muscles in patients with TK2d, compared to mutations in the muscle MRI involvement in FSHD is commonly asymmet- other nuclear genes, this pattern being the only one identified ric, while it is symmetric in TK2d as shown here in Fig. 6 in patients with PMM. Our results also indicated that gluteus [13]. The muscle MRI pattern of OPMD also differs from maximus, gastrocnemius medialis, and sartorius were the TK2d, showing in the pelvis a progressive involvement of earliest affected muscles in patients with TK2 mutations. The glutei muscles, especially gluteus minimus, in the thighs a sartorius is a muscle that is not usually involved or is even preferential involvement of posterior muscles and the adduc- preserved until late stages in many myopathies, so this find- tor muscles, and in the legs, an early involvement of the ing can help in the differential diagnosis. Early involvement soleus and peroneus muscles [14]. of the sartorius muscle has also been described in patients Muscle MRI is a complementary tool used to detect pat- with mutations in SPEN1 [19], RYR1 [20], TNPO3 [21], and terns of muscle fat replacement that could help in the diag- MYH7 [22], in some congenital myopathies like ACTA1 [23], nosis of patients with myopathies. Although next-generation and in myofibrillar myopathies with mutations in DES [24] sequencing is becoming widely accessible in several centers, or CRYAB [12]. many of the mitochondrial genes are not included in the 13 3558 Journal of Neurology (2022) 269:3550–3562 Fig. 5  PCA analysis. TK2d compared to other mitochondrial myo- digitorum, and tibialis anterior were the muscles that differentiated pathies (A) showed that gluteus maximus, vastus lateralis, obturator most one disease to another. C TK2d compared to OPMD showed externus, sartorius, and flexor digitorum were the muscles that dif- that gluteus maximus, vastus lateralis, obturator externus, sartorius, ferentiated most one disease to another. B TK2d compared to FSHD and flexor digitorum were the muscles that differentiated most from showed that gluteus maximus, obturator externus, sartorius, the short one disease to another head of biceps femoris, semimembranosus, flexor digitorum, extensor panels commercially available, and in several countries, the infiltration have been described for many diseases [25, 26] diagnosis continues to be based on Sanger sequencing. In and there is an increasing body of evidence demonstrating addition, interpretation of NGS results is not always straight- that MRI-based artificial intelligence tools could accurately forward and the addition of clinical, muscle biopsy, and MRI diagnose muscular disorders [27]. Moreover, muscle fat data can be helpful. In this sense, patterns of muscle MRI replacement has been shown to correlate with results of 13 Journal of Neurology (2022) 269:3550–3562 3559 Fig. 6  Boxplots showing the difference in Mercuri score between right and left sides of muscles of the pelvis, thigh and leg of FSHD and TK2d. Results of the Wilcoxon test applied to identify significant differences are shown muscle function tests, and the value of MRI as a possible There are some limitations to this study. The cohort of biomarker to evaluate response to treatments in clinical trials patients with other mitochondrial myopathies is genetically is increasingly being recognized [28, 29]. very heterogeneous and, therefore, it is expected that their Despite all these advantages, there are not many studies muscle MRI are different. Another limitation is the use of a of muscle MRI in mitochondrial myopathies [30, 31]. The semi-quantitative scoring method. Although the method is largest series published to date included 16 patients with commonly used in other studies, inter-interpreter variability mitochondrial myopathy due to different mutations only in is intrinsic to this type of rating scale. the mtDNA (including single large-scale deletions and point In conclusion, analyzing the largest cohort of muscle- mtDNA mutations [32]) and a characteristic pattern of mus- imaging studies from patients with mitochondrial myopa- cle involvement was not identified. thies to date, we have identified a characteristic and specific In this study, we have used conventional qualitative MRI radiological pattern of muscle involvement of the lower imaging techniques, such as T1w, to identify fat replacement limbs in patients with TK2d. This pattern allows: (1) to dis- in muscles. However, in the last 10 years, there has been tinguish TK2d from other mitochondrial myopathies and, a notorious development of quantitative sequences such as therefore, to guide genetic testing, (2) to support the diagno- Dixon or water T2 that can quantify fat replacement and sis when the identified variants in the TK2 gene are classified water content, respectively, and that have been validated as as of uncertain clinical significance, and (3) to differentiate good candidate imaging biomarkers to follow-up patients TK2d from other myopathies with similar clinical manifes- with neuromuscular diseases. These sequences have not been tations, such as FSHD and OPMD, which is particularly used to study patients with mitochondrial myopathies [32]. important now because genetic studies are usually prior- Phosphorus magnetic resonance spectroscopy (31P-MRS) itized over muscle biopsies during the diagnostic process. could be of great interest in this field as it monitors mito- chondrial function in vivo in skeletal muscle [33, 34]. In Supplementary Information The online version contains supplemen- tary material available at https://d​ oi.o​ rg/1​ 0.1​ 007/s​ 00415-0​ 21-1​ 0957-0. fact, 31P-MRS of resting muscle is sensitive and specific for diagnosing mitochondrial myopathies [35] and has been Author contributors CD-G and CP designed the manuscript. CD-G, used in small studies to evaluate therapeutic interventions CP, CPF-F, BV, JAC, JA-P, LG-M, MO, JG-G, GM, JCL-H, NM, and [36]. However, 31P-MRS is a complex technique not avail- MAM performed the patient selection and acquisition of data. RF-T, able in many centers and, therefore, not included in the UM, JD-M, ES-M, and CD-G made the analysis and interpretation of standards of care for mitochondrial diseases [37]. 13 3560 Journal of Neurology (2022) 269:3550–3562 the results. CD-G and CP prepared the first and subsequent drafts of the in thymidine kinase 2 deficiency. Biochem Biophys Res Commun manuscript. All the authors read and approved the submitted version. 310(3):963–966 6. Ashley N et al (2007) Defects in maintenance of mitochondrial Funding This work was supported in part by the Instituto de Salud DNA are associated with intramitochondrial nucleotide imbal- Carlos III and FEDER (PI18/01374 to MA.M and PMP15/00025 to CP, ances. Hum Mol Genet 16(12):1400–1411. https://​doi.​org/​10.​ MO and MAM.). CD-G, CP, NM and MO are members of the Euro- 1093/​hmg/​ddm090 pean Reference Network for Neuromuscular Diseases (EURO-NMD) 7. Garone C et al (2018) Retrospective natural history of thymidine and MO is member of XUECs (Xarxes d'Unitats d'Expertesa Clínica kinase 2 deficiency. J Med Genet. https://​doi.​org/​10.​1136/​jmedg​ en Malalties Minoritàries de Catalunya). enet-​2017-​105012 8. Wang J et al (2018) Clinical and molecular spectrum of thymidine kinase 2-related mtDNA maintenance defect. Mol Genet Metab. Data availability statement Data are available upon reasonable request. https://​doi.​org/​10.​1016/j.​ymgme.​2018.​04.​012 The data central to this study are available on request from suitably 9. Rodríguez-López C et al (2020) Clinical, pathological and genetic qualified researchers. spectrum in 89 cases of mitochondrial progressive external ophthalmoplegia. J Med Genet. https://​doi.​org/​10.​1136/​jmedg​ Declarations enet-​2019-​106649 10. Domínguez-González C et al (2019) Deoxynucleoside therapy for Conflicts of interest The authors declare that the research was con- thymidine kinase 2-deficient myopathy. Ann Neurol 86(2):293– ducted in the absence of any conflict of interest. 303. https://​doi.​org/​10.​1002/​ana.​25506 11. Mancuso M, McFarland R, Klopstock T, Hirano M et al (2017) Ethical approval The Institutional Review Board Committee of Hos- International Workshop: outcome measures and clinical trial pital 12 de Octubre “Ethics Committee for Clinical Research—CEIC” readiness in primary mitochondrial myopathies in children and reviewed and approved the study (PMP15-00025) under Number adults. Consensus recommendations. 16–18 November 2016, 16/070. Rome, Italy. Neuromuscul Disord 27(12):1126–1137. https://​doi.​ org/​10.​1016/j.​nmd.​2017.​08.​006 Provenance and peer review Not commissioned; externally peer- 12. Fischer D et al (2008) Distinct muscle imaging patterns in myofi- reviewed. brillar myopathies. Neurology 71(10):758–765. https://​doi.​org/​ 10.​1212/​01.​wnl.​00003​24927.​28817.​9b 13. Tasca G et al (2012) Different molecular signatures in magnetic Open Access This article is licensed under a Creative Commons Attri- resonance imaging-staged facioscapulohumeral muscular dystro- bution 4.0 International License, which permits use, sharing, adapta- phy muscles. PLoS ONE 7(6):e38779. https://​doi.​org/​10.​1371/​ tion, distribution and reproduction in any medium or format, as long journ​al.​pone.​00387​79 as you give appropriate credit to the original author(s) and the source, 14. Alonso-Jimenez A et al (2019) Muscle MRI in a large cohort provide a link to the Creative Commons licence, and indicate if changes of patients with oculopharyngeal muscular dystrophy. J Neurol were made. The images or other third party material in this article are Neurosurg Psychiatry 90(5):576–585. https://​doi.​org/​10.​1136/​ included in the article's Creative Commons licence, unless indicated jnnp-​2018-​319578 otherwise in a credit line to the material. If material is not included in 15. Cohen BH (2019) Mitochondrial and metabolic myopathies. Con- the article's Creative Commons licence and your intended use is not tinuum 25(6):1732–1766. https://​doi.​org/​10.​1212/​CON.​00000​ permitted by statutory regulation or exceeds the permitted use, you will 00000​000805 need to obtain permission directly from the copyright holder. To view a 16. Gorman GS et al (2016) Mitochondrial diseases. Nat Rev Dis copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. Primers 2:16080. https://​doi.​org/​10.​1038/​nrdp.​2016.​80 17. Dominguez-Gonzalez C et al (2020) Growth Differentiation Fac- tor 15 is a potential biomarker of therapeutic response for TK2 deficient myopathy. Sci Rep 10(1):10111. https://d​ oi.o​ rg/1​ 0.1​ 038/​ s41598-​020-​66940-8 References 18. Paradas C, Gutiérrez Ríos P, Rivas E, Carbonell P, Hirano M, DiMauro S (2013) TK2 mutation presenting as indolent myopathy. 1. Saada A, Shaag A, Mandel H, Nevo Y, Eriksson S, Elpeleg O Neurology 80(5):504–506. https://​doi.​org/​10.​1212/​WNL.​0b013​ (2001) Mutant mitochondrial thymidine kinase in mitochondrial e3182​7f0ff7 DNA depletion myopathy. Nat Genet 29(3):342–344. https://​doi.​ 19. Hankiewicz K et al (2015) Whole-body muscle magnetic reso- org/​10.​1038/​ng751 nance imaging in SEPN1-related myopathy shows a homogeneous 2. Béhin A et al (2012) Adult cases of mitochondrial DNA deple- and recognizable pattern. Muscle Nerve 52(5):728–735. https://​ tion due to TK2 defect: an expanding spectrum. Neurology doi.​org/​10.​1002/​mus.​24634 78(9):644–648. https://​d oi.​o rg/​1 0.​1 212/​W NL.​0 b013​e 3182​ 20. Tordjman M et al (2018) Muscular MRI-based algorithm to dif- 48df2b ferentiate inherited myopathies presenting with spinal rigidity. Eur 3. Domínguez-González C et al (2019) Late-onset thymidine kinase Radiol. https://​doi.​org/​10.​1007/​s00330-​018-​5472-5 2 deficiency: a review of 18 cases. Orphanet J Rare Dis 14(1):100. 21. Peterle E, Fanin M, Semplicini C, Padilla JJ, Nigro V, Angelini C https://​doi.​org/​10.​1186/​s13023-​019-​1071-z (2013) Clinical phenotype, muscle MRI and muscle pathology of 4. Chanprasert S et al (2013) Molecular and clinical characteriza- LGMD1F. J Neurol 260(8):2033–2041. https://​doi.​org/​10.​1007/​ tion of the myopathic form of mitochondrial DNA depletion syn- s00415-​013-​6931-1 drome caused by mutations in the thymidine kinase (TK2) gene. 22. Carbonell-Corvillo P et al (2018) A novel MYH7 founder muta- Mol Genet Metab 110(1–2):153–161. https://​doi.​org/​10.​1016/j.​ tion causing Laing distal myopathy in Southern Spain. Neuromus- ymgme.​2013.​07.​009 cul Disord. https://​doi.​org/​10.​1016/j.​nmd.​2018.​07.​006 5. Saada A, Ben-Shalom E, Zyslin R, Miller C, Mandel H, Elpeleg 23. Castiglioni C et al (2014) Muscle magnetic resonance imaging and O (2003) Mitochondrial deoxyribonucleoside triphosphate pools histopathology in ACTA1-related congenital nemaline myopathy. 13 Journal of Neurology (2022) 269:3550–3562 3561 Muscle Nerve 50(6):1011–1016. https://​doi.​org/​10.​1002/​mus.​ involvement. Acta Neuropathol 138(6):1013–1031. https://​doi.​ 24353 org/​10.​1007/​s00401-​019-​02059-z 24. Olivé M et al (2011) Clinical and myopathological evaluation 32. Hedermann G et al (2018) Progressive fat replacement of mus- of early- and late-onset subtypes of myofibrillar myopathy. Neu- cle contributes to the disease mechanism of patients with single, romuscul Disord 21(8):533–542. https://​doi.​org/​10.​1016/j.​nmd.​ large-scale deletions of mitochondrial DNA. Neuromuscul Disord 2011.​05.​002 28(5):408–413. https://​doi.​org/​10.​1016/j.​nmd.​2018.​02.​008 25. Leung DG (2017) Magnetic resonance imaging patterns of 33. Matthews PM et al (1991) In vivo magnetic resonance spectros- muscle involvement in genetic muscle diseases: a systematic copy of brain and muscle in a type of mitochondrial encephalo- review. J Neurol 264(7):1320–1333. https://​doi.​org/​10.​1007/​ myopathy (MERRF). Ann Neurol 29(4):435–438. https://​doi.​org/​ s00415-​016-​8350-6 10.​1002/​ana.​41029​0416 26. Aivazoglou LU et al (2021) MR imaging of inherited myopathies: 34. Matthews PM, Allaire C, Shoubridge EA, Karpati G, Carpenter a review and proposal of imaging algorithms. Eur Radiol. https://​ S, Arnold DL (1991) In vivo muscle magnetic resonance spec- doi.​org/​10.​1007/​s00330-​021-​07931-9 troscopy in the clinical investigation of mitochondrial disease. 27. Verdú-Díaz J et al (2020) Accuracy of a machine learning muscle Neurology 41(1):114–120. https://​doi.​org/​10.​1212/​wnl.​41.1.​114 MRI-based tool for the diagnosis of muscular dystrophies. Neu- 35. Kemp GJ, Ahmad RE, Nicolay K, Prompers JJ (2015) Quantifica- rology 94(10):e1094–e1102. https://d​ oi.o​ rg/1​ 0.1​ 212/W ​ NL.0​ 0000​ tion of skeletal muscle mitochondrial function by 31P magnetic 00000​009068 resonance spectroscopy techniques: a quantitative review. Acta 28. Diaz-Manera J et al (2018) Muscle MRI in patients with dysfer- Physiol 213(1):107–144. https://​doi.​org/​10.​1111/​apha.​12307 linopathy: pattern recognition and implications for clinical trials. 36. Kornblum C et al (2005) Creatine has no beneficial effect on J Neurol Neurosurg Psychiatry 89(10):1071–1081. https://d​ oi.o​ rg/​ skeletal muscle energy metabolism in patients with single mito- 10.​1136/​jnnp-​2017-​317488 chondrial DNA deletions: a placebo-controlled, double-blind 31P- 29. Alonso-Pérez J et al (2020) New genotype-phenotype correlations MRS crossover study. Eur J Neurol 12(4):300–309. https://d​ oi.o​ rg/​ in a large European cohort of patients with sarcoglycanopathy. 10.​1111/j.​1468-​1331.​2004.​00970.x Brain 143(9):2696–2708. https://​doi.​org/​10.​1093/​brain/​awaa2​28 37. Parikh S et al (2017) Patient care standards for primary mito- 30. Tasca G, Monforte M, Nesti C, Santorelli FM, Silvestri G, Ricci chondrial disease: a consensus statement from the Mitochondrial E (2017) Concentric muscle involvement in POLG-related distal Medicine Society. Genet Med 19(12):12. https://​doi.​org/​10.​1038/​ myopathy. Neuromuscul Disord 27(5):500–501. https://​doi.​org/​ gim.​2017.​107 10.​1016/j.​nmd.​2017.​03.​001 31. Donkervoort S et al (2019) MSTO1 mutations cause mtDNA depletion, manifesting as muscular dystrophy with cerebellar Authors and Affiliations Cristina Domínguez‑González1,2,3 · Roberto Fernández‑Torrón4 · Ursula Moore5 · Carlos Pablo de Fuenmayor‑Fernández de la Hoz1 · Beatriz Vélez‑Gómez6,7 · Juan Antonio Cabezas6 · Jorge Alonso‑Pérez8 · Laura González‑Mera9 · Montse Olivé3,8,10 · Jorge García‑García11 · Germán Moris12 · Juan Carlos León Hernández13 · Nuria Muelas3,14 · Emilia Servian‑Morilla6,7 · Miguel A. Martin2,3,15 · Jordi Díaz‑Manera3,5,10,17 · Carmen Paradas6,7,16 1 9 Neuromuscular Diseases Unit, Neurology Department, Neuromuscular Diseases Unit, Neurology Department, Hospital Universitario 12 de Octubre, Madrid, Spain IDIBELL-Hospital Universitari de Bellvitge, Hospitalet de 2 Llobregat, Barcelona, Spain Hospital 12 de Octubre Research Institute (imas12), Madrid, 10 Spain Institut de Recerca, Hospital de la Santa Creu i Sant Pau, 3 Barcelona, Spain Biomedical Network Research Centre on Rare Diseases 11 (CIBERER), Instituto de Salud Carlos III, Madrid, Spain Department of Neurology. Hospital, Universitario de 4 Albacete., Albacete, Spain Neurology Department, Biodonostia Health Research 12 Institute, Neuromuscular Area, Hospital Donostia, Basque Neuromuscular Diseases Unit. Neurology Health Service, Doctor Begiristain, Donostia‑San Sebastian, Department. Hospital, Universitario Central de Asturias, Spain Asturias, Spain 5 13 John Walton Muscular Dystrophy Research Center, Neurology Department. Hospital, Universitario Nuestra University of Newcastle, Newcastle, UK Señora de Candelaria, Tenerife, Spain 6 14 Neuromuscular Diseases Unit, Neurology Department, Neuromuscular Diseases Unit, Neurology Department, Hospital Universitario Virgen del Rocío/ Instituto de Hospital Universitari i Politecnic La Fe, Neuromuscular Biomedicina de Sevilla, Sevilla, Spain and Ataxias Research Group, Instituto de Investigación 7 Sanitaria La Fe, Valencia, Spain Centro de Investigación Biomédica en Red Sobre 15 Enfermedades Neurodegenerativas (CIBERNED), Madrid, Mitochondrial Disorders Laboratory, Clinical Biochemistry Spain Department, Hospital Universitario 12 de Octubre, Madrid, 8 Spain Neuromuscular Diseases Unit, Neurology Department, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Bellaterra, Spain 13 3562 Journal of Neurology (2022) 269:3550–3562 16 17 Unidad de Enfermedades Neuromusculares John Walton Muscular Dystrophy Research Centre Newcastle (CSUR/EURO‑NMD), Hospital Universitario Virgen del University Translational and Clinical Research Institute Rocío/ Instituto de Biomedicina de Sevilla, Sevilla, Spain Central Parkway, Newcastle upon Tyne NE1 3BZ, UK 13