Genetic, structural and clinical analysis of spastic paraplegia 4

Background: Spastic paraplegia type 4 (SPG4), resulting from heterozygous mutations in the SPAST gene, is the most common form among the heterogeneous group of hereditary spastic paraplegias (HSPs). Objective: To study genetic and clinical characteristics of SPG4 across Canada. Methods: The SPAST gene was analyzed in a total of 696 HSP patients from 431 families by either HSP-gene panel sequencing or whole exome sequencing (WES). We used Multiplex ligation-dependent probe amplification to analyze copy number variations (CNVs), and performed in silico structural analysis of selected mutations. Clinical characteristics of patients were assessed, and long-term follow-up was done to study genotype-phenotype correlations. Results: We identified 157 SPG4 patients from 65 families who carried 41 different SPAST mutations, six of which are novel and six are CNVs. We report novel aspects of mutations occurring in Arg499, a case with homozygous mutation, a family with probable compound heterozygous mutations, three patients with de novo mutations, three cases with pathogenic synonymous mutation, co-occurrence of SPG4 and multiple sclerosis, and novel or rarely reported signs and symptoms seen in SPG4 patients. Conclusion: Our study demonstrates that SPG4 is a heterogeneous type of HSP, with diverse genetic features and clinical manifestations. In rare cases, biallelic inheritance, de novo mutation, pathogenic synonymous mutations and CNVs should be considered.


Introduction
Spastic paraplegia type 4 (SPG4, OMIM #182601) is the most frequent form of either sporadic or familial hereditary spastic paraplegias (HSPs), caused by heterozygous mutations in the SPAST gene. [1][2][3] With more than 80 potentially causative loci or genes reported to date, 4 HSPs are known to affect 1-10/100,000 of the population, 5 and autosomal dominant (AD) HSPs comprise 43%-80% of them. [5][6][7][8][9][10][11] Among all AD-HSPs, 70-80% are categorized as "pure" 12 with a phenotype limited to pyramidal signs in the lower limbs, with or without deep sensory loss and sphincter disturbances. [12][13][14][15] Of all pure AD-HSPs, about 40% are caused by SPAST mutations. 12 SPAST encodes spastin, which is a protein from the AAA (ATPase associated with various cellular activities) family of ATPases. 16,17 Spastin controls different aspects of microtubule dynamics (e.g. microtubule number, motility, length, disassembly and remodeling), and hydrolyses ATP to cleave microtubules, 18, 19 a necessary step in axonal transport. 20 Its mechanism involves binding to the C-terminus of tubulin and severing tubulin subunits from the microtubule in an ATP hydrolysis-dependent manner. 19 The structure of human spastin residues 323-610 was solved by cryoelectron microscopy. 21 The structure reveals an AAA+ ATPase homohexamer ( Figure 1A), with ATPase active sites located at the interface between every adjacent subunit ( Figure 1B). 22, 23 A tubulin peptide runs through the channel, suggesting a handover-hand mechanism of substrate translocation, with five subunits interacting with the peptide forming a spiral staircase and one displaced from the peptide/substrate ( Figure 1A). 21 To date, more than 200 SPAST mutations have been found. 3,16,[24][25][26] Deletion/insertions, nonsense, and splice site mutations are distributed throughout the gene, while missense mutations are mostly clustered in the AAA domain. 3 Exon deletions may account for 20% of cases in whom point mutations are not detected. 27,28 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint The penetrance of SPG4 is up to 80-90% and is age-dependent, 3 with age at onset (AAO) that may range from infancy to the eighth decade of life. 1,29 Most cases present as juvenile or adult-onset pure spastic paraplegia with urinary sphincter disturbances, pes cavus, and dysarthria.
However, SPAST mutations are known to cause clinically heterogeneous manifestations, and a high variety of signs and symptoms is reported among carriers of different SPAST mutations.
This clinical heterogeneity occurs even among patients harboring the same mutations, 30 suggesting that other factors affect the clinical presentation of SPG4.
Although previous studies have deciphered some aspects of this variability, the genotypephenotype correlations and some rare features of SPG4 are not fully understood. In this study, we analyzed a large cohort of SPG4 patients from Canada to better clarify the genetic and clinical spectrum of the disease.

Population
A total of 696 HSP patients from 431 families were recruited in eight medical centers across Canada (Montreal, Quebec, Ottawa, Toronto, Hamilton, Calgary, Edmonton, and Vancouver) as part of CanHSP, a Canadian consortium for the study of HSP. Details about the diagnosis and recruitment process has been previously reported. 31 Clinical assessments were done including family history, demographic data, developmental history, AAO, and HSP core symptoms (lower extremity weakness and spasticity, extensor plantar responses, hyperreflexia, and bladder dysfunction). Other neurological, as well as non-neurological clinical presentations of HSP, were also assessed. For a group of the patients, the Spastic Paraplegia Rating Scale (SPRS), which is . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 20, 2021.
an indicator of severity in HSP, was measured. 32, 33 For another subset of patients, brain and/or spine MRI were performed. The disease was deemed pure unless the patient had at least one sign not attributable to the lateral corticospinal tract or pyramidal tract, including ataxia, intellectual disability, cognitive decline, language development delay, extrapyramidal signs, visual dysfunction, epilepsy, deafness, dysarthria, optic atrophy, peripheral neuropathy, and dystonia; in which case the disease was classified as complex.
All the data is stored in a central database at McGill University. All patients signed informed consent forms and the institutional review board approved the study protocol.

Genetic and data analysis
DNA was extracted from peripheral blood using a standard procedure. 34 Initially, HSP-gene panel sequencing was performed on 379 patients. Then, 194 genetically undiagnosed patients, and additional 206 patients who were not analyzed with panel sequencing (400 patients in total from 291 families), went through whole exome sequencing (WES).
For WES, the Agilent SureSelect Human All Exon v4 kit for capture and targeted enrichment of the exome was used. To analyze the WES data, we used a list of 785 HSP-related genes or genes associated with similar neurological disorders which cause spasticity (Supplementary Table 1). Illumina HiSeq 2000/2500/4000 system was used for sequencing captured samples. Using Burrows-Wheeler Aligner (BWA), the sequence reads were then aligned against the human genome (GRCh37 assembly). 35 We used the Genome Analysis ToolKit (GATK) 36 and Annotate Variation (ANNOVAR) 37 for variant calling and annotation, respectively. We excluded variant calls with a genotype quality less than 97 and less than 30x depth of coverage. Integrated Genomics Viewer was used to visually inspect the detected . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint variants, and suspicious variants were validated by Sanger Sequencing. Sanger Sequencing was also used for assessing sporadic patients' parents, to determine if the proband had a de novo mutation.
SPAST variants (NM_014946) were initially selected based on identifying missense and LoF alleles, including frame-shift, splice-site, nonsense, and copy number variations (CNVs) with a minor allele frequency less than 0.01 in gnomAD. 38 The variants' pathogenicity has been determined using VarSome, 39 according to the American College of Medical Genetics and Genomics (ACMG) guideline. Variants classified as "Benign" and "Likely Benign", as well as intronic splice site variants higher than ±3 were excluded from the analysis. To detect CNVs, ExomeDepth 40 was used on WES data, followed by 48 selected samples that went through Multiplex ligation-dependent probe amplification (MLPA) testing (MRC Holland, Amsterdam, The Netherlands) to confirm or exclude suspected SPAST CNVs. InterPro 41 was applied to identify domains and corresponding sites in the protein.

Statistical analysis
To determine the association between two categorical variables, one categorical variable with one continuous variable, and two continuous variables, Pearson chi-squared test, Mann-Whitney U test, and Spearman's rank correlation coefficient were used, respectively. P-value was set at <0.05 and Bonferroni correction for multiple comparisons was applied when necessary. SPSS was used to perform all statistical analyses.

In silico structural analysis
The atomic coordinates of human spastin bound to a glutamate-rich peptide, ADP, BeF 3 and Mg 2+ and D. melanogaster spastin bound to a glutamate-rich peptide, ADP, ATP and Mg 2+ were . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 20, 2021.  Figure 2). Presentation of the disease did not differ significantly between the two types of mutation (Table 1).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. changes all the residue subsequent to residue 365 and inserts a stop codon at position 394. This mutation truncates residues 394 to 616 which includes residues involved in oligomerization and ATP-binding, which will abrogate its enzymatic activity. Gly382 and Pro383 are located in the nucleotide-binding loop 382-389 . 43 Residues Gly385 and Asn386 interact directly with the ADP, in a conformation that is shaped by the Pro383 and Pro384 ( Figure 1C). The mutation p.(Gly382_Pro383insArg) lengthens this loop and likely destabilizes the interaction between the protein and the nucleotide. Phe403 forms a pi-stacking interaction with Phe439 and hydrophobic interactions with neighbouring residues ( Figure 1D). Thus, mutation p.(Phe403Leu) would destabilize the domain and impair the ATPase activity.
In this structure, Arg498 and Arg499 are both involved in coordinating ADP. Arg498 is making direct interaction with the β phosphate of the nucleotide located in the active site, while Arg499 stabilizes helix 455-470 and coordinates BeF 3 , a phosphate structural analog ( Figure 1E).
By homology with D. melanogaster spastin structure (PDB ID: 6P07), those residues seem to be also involved in ATP β and γ phosphate coordination ( Figure 1F). As a result, we can . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; 1 0 hypothesize that the BeF 3 in the active site is mimicking the γ phosphate on an ATP. Mutation p.(Arg498AlafsTer30) causes residues 498 and 499 to be serine and valine residues, which would prevent ATP stabilization. In addition, this frame shift terminates the protein after residue 528, further disrupting the ATP-binding domain. This mutation will therefore abolish the ATPase activity. Finally, mutation p.(Ser597ThrfsTer3) will induce a frame shift at residue 597 and truncate the protein at residue 600. The missing C-terminal residues form a helix involved in intersubunit interactions. This mutation would therefore destabilize the hexameric assembly, which would disrupt the ATPase activity.

Founder French-Canadian mutations and known CNVs
The most frequent SPAST mutation in our cohort, p.(Gly559Asp), which has previously been suggested to be a founder mutation in French-Canadian population 44 , was carried by 8 families (12.3%) and 15 patients in total (9.5%). Seven out of 24 French-Canadian probands (29%) harbored this mutation, while among all the patients with other or unknown ancestral backgrounds, this mutation was present in only one out of 29 probands (3.4%, p=0.011).
Moreover, the mutation p.(Phe403Leu) was only detected in patients with French-Canadian ancestral background, in 5 patients from 2 families, suggesting it may be a founder mutation.

p.Arg499
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 1 1 The range of AAO in all 4 patients from 4 different families, two of which have previously been reported 45 with the p.(Arg499His) mutation was 1-3 years (Mann-Whitney U test; p = 0.003).
The age range at which these four patients started to show symptoms was 1-5. Patients with this mutation in our cohort were more likely to present with motor delay, speech delay, dysarthria, learning disability, progressive cognitive deficits, and upper extremity weakness (Supplementary Table3).
Although statistically insignificant, cases who carried another mutation affecting the same amino acid residue, p.(Arg499Cys), also showed symptoms at a younger age (the range of AAO was 1-5 in three patients and 11-15 in one patient, p = 0.111). When combined together, the two mutations occurred in Arg499 locus were associated with a younger AAO (Mann-Whitney U test; p = 0.004). In silico analysis of Arg499 is discussed above, along with Arg498.

Possible biallelic inheritance in SPG4
The SPAST p.(Ser44Leu) variant has been previously suggested to play modifier role in SPG4. 1,46,47 This variant was detected in one out of the 36 families that had undergone WES. From the three affected siblings, two had undergone WES, and both were heterozygous for the novel pathogenic mutation p.(Ser597ThrfsTer3), as well as for the p.(Ser44Leu) variant. The three siblings had complex HSP with disease onset in early childhood (clinical presentation of these patients is detailed in Supplementary Table 4). One unaffected parent only carried the p.(Ser44Leu) polymorphism, and the other unaffected parent was heterozygous for the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint pathogenic p.(Ser597ThrfsTer3) mutation and wild-type for p.(Ser44Leu) (Supplementary Figure   2). The last physical examination of the parent with the pathogenic mutation at the age between 56 to 60 revealed completely normal findings, including normal reflexes, gait, and motor exam, suggesting that the p.(Ser44Leu) variant may be a modifier of the disease, or that extreme anticipation with undetected clinical effects in the mother exist in this family.
Another patient with consanguineous parents, carried a homozygous mutation, SPAST p.(Tyr51Ter). The patient had started to show symptoms at an age range of 1-5 years and suffered from core HSP symptoms, motor and speech delay, swallowing difficulty, dysarthria, upper extremity weakness, pes cavus, and skeletal abnormalities. Her brain and total spine MRI were normal and her SPRS score was 40 when examined at an age between 16 and 20 years. The patient's parents were both asymptomatic heterozygous carriers.

Patients with novel or rarely reported clinical manifestations
In our cohort, four patients (Supplementary Table 5) presented with deafness, one of whom had a de novo SPAST p.(Arg499His) mutation, and the remaining three had familial SPG4. We also report a patient with ocular movement abnormality, in whom extraocular movements showed slightly hypometric saccades but pursuit was smooth, and a mild horizontal gaze evoked rotatory nystagmus. The disease-causing mutation in this patient was p.(Gln434Ter).
One of the signs less reported to date in SPG4 is upper extremity intention tremor, seen in two of the patients in the current study. The causative mutations in these two patients were p.(Arg503Trp) and p.(Leu371Pro). Their respective range of AAO was 46-50 and 26-30, and both presented with a severe form of the disease (respective SPRS scores: 34 and 39). The first . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint patient also had swallowing difficulty and peripheral neuropathy, and both had upper extremity ataxia.
Another clinical feature seen in two of the patients in our cohort was seizures, which is quite rare in SPG4. One of the patients of who carried the mutation p.(Pro489Leu), had a range of AAO of 21-25 years, and suffered from generalized seizures starting at 41-45 years of age. The second patient harbored the splice-site mutation c.1321+1G>A, with AAO of between 1 and 5 years, had episodes of atypical seizures at 56-60.

De novo cases
The parents of the three sporadic cases in our cohort did not carry the causative SPAST mutations. The first patient, with SPAST p.(Ser407Asn), had an AAO of 1-5 years. This case presented with lower extremity weakness and spasticity, upper and lower limb hyperreflexia, extensor plantar response, ankle clonus, and sensory abnormalities. She also showed signs of upper extremity weakness, motor and speech delay, dysarthria, learning disability, and swallowing difficulty. Their brain MRI was normal. The second patient had de novo SPAST p.(Arg499His). This patient started to show symptoms at age range of 1-5, and her symptoms included lower extremity weakness, spasticity and hyperreflexia, extensor plantar responses, and ankle clonus. Furthermore, they had motor and speech delay, dysarthria, and learning disability.
They had a normal brain MRI, and their SPRS score was 37. The last de novo patient, also with a p.(Arg499His) mutation, also with an AAO range of 1-5 years, had an SPRS score of 28, and apart from core symptoms of HSP, had motor and speech delay, deafness, and dysarthria. The two latter patients have been previously reported. 45 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Cases with synonymous mutation and co-occurrence of multiple sclerosis and SPG4
We report three patients from two unrelated families who carried the synonymous mutation SPAST p.(Lys414Lys), previously reported to be pathogenic. 48 The first patient had an AAO of between 6-10 years, and presented with lower extremity spasticity, hyperreflexia, and extensor plantar responses. The second family had two affected individuals, parent and offspring. The parent started to have difficulty walking at the age of 61-65 which led to using a cane, had marked spasticity and brisk reflexes in the lower limbs along with ankle clonus, upgoing plantar responses, urge incontinency, and mild decrease of vibration sensation in the ankles and toes.
The offspring noticed difficulty walking at 21-25 years of age, which progressed slowly. This case had upper and lower limb weakness, bilateral Babinski sign and ankle clonus, and positive Hoffman sign. At an age between 31 and 35, lumbar puncture carried out due to diplopia was positive for oligoclonal bands and suggested a diagnosis of clinically isolated syndrome (CIS), which is considered as the first clinical episode in multiple sclerosis (MS). 49 They were treated with methylprednisone, and after one month, the diplopia resolved.

Discussion
In this large-scale analysis of SPG4 from CanHSP, we report novel SPAST mutations, the possibility of a founder mutations in the French-Canadian population, novel characteristics of mutations occurring in Arg499, and potential biallelic inheritance. We also report patients with rare or novel clinical manifestations, and co-occurrence of SPG4 and MS.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint Using in silico analysis, we predict that the novel mutations reported in this study cause a loss-of-function of spastin, either by substantially shortening the protein, affecting its ATPase activity, or disturbing the formation of a ternary and quaternary structure necessary for catalysis.
Spastin, a microtubule-cleaving enzyme involved in the cytoskeletal rearrangement, 50-52 also has a role in intracellular trafficking, cytokinesis regulation and resealing of nuclear membrane. [53][54][55] In neurons, it is involved in the regeneration of axons and axonal transport. 51,[56][57][58] Mutations in SPAST could result in a disruption of normal organelle trafficking and distribution, and thus micro-organelle accumulation in axons and swelling of axons [59][60][61][62][63] which could result in HSP phenotype. We looked further into the three patients identified as sporadic and found out that all carried de novo mutations. Consistent with previous reports of more severe manifestations in patients with de novo mutations, 64 all patients showed a severe form of the disease. However, it is important to note that the mutation in two of the de novo patients (Arg499His), which is also carried by 33% of de novo patients, 64 is associated with a younger AAO, and infantile-onset ascending spastic paraplegia (IAHSP); regardless of the mode of inheritance. 65,66 Patients harboring this mutation in our cohort did not show all criteria to be diagnosed as IAHSP (spastic paraplegia progressing to tetraplegia, ocular gaze paralysis, pseudobulbar palsy 67 ), however, they showed a higher probability of suffering from upper extremity weakness, dysarthria, speech and motor delay, learning disabilities and cognitive delay. It has been suggested that amino acid 499 is located in one of the highly conserved regions of the SPAST gene, 18 and mutations that occur in this region could have a significant effect on the hydrophobicity of the protein and its ability to sever microtubules. 45 The mutation in our homozygous patient has been reported previously in a heterozygous patient; 68 however, the presence of an additional mutation or CNVs could not have been . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint determined in that study, as only SPAST had been sequenced. Our patient with the homozygous mutation SPAST p.(Tyr51Ter) showed a severe form of the disease. Previously, there have been two reports of homozygous variants in SPAST. One, with the nonsense mutation p.(Ser545Ter), which also reports a complex and severe form of the disease, 69 and another, with the mutation p. (Leu534Pro), who showed a pure spastic paraparesis at the age of 39. 70 The patient from our cohort and the previously reported patient with p.(Ser545Ter) presented with a more severe form of the disease, probably because they carried nonsense mutations, while the latter patient had a missense mutation. Furthermore, in the latter study, the parents of the proband were asymptomatic carriers of the mutation.
The mode of inheritance observed in the patient with homozygous mutation and the family carrying the potential modifier mutation SPAST p.(Ser44Leu), may be explained by incomplete penetrance and/or anticipation 71-73 rather than biallelic inheritance, which could be considered in prenatal diagnosis and genetic counselling. The synonymous mutation SPAST p.(Lys414Lys) we report in three patients has been previously reported to underlie the disease, and minigene assay suggested that this variant leads to an aberrant splicing effect. 48 Although we did not find other synonymous SPAST mutations in additional HSP patients, it is important to not exclude synonymous variants when analyzing HSP specifically and other diseases in general.
The AAO distribution in our cohort followed the same trend as previously reported, 1,68 however, our results did not replicate previous findings which suggested that the underlying mutation and the patient sex can modify the disease, and that patients with later onset of the disease can have faster progression. 1,24 In our data, we did not find an association between mutation type, sex, or late onset, with the disease course. Table 3 compares the frequency of symptoms not attributable to pure form of the disease in our cohort with previous studies. As . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint shown, we report deafness in SPG4 for the first time. The co-occurrence of SPG4 and MS has been reported before; 74, 75 however, the results of studies assessing an association between HSPrelated mutations and MS are controversial 76,77 and further studies are required to verify such a relation.
Our study has several limitations. For instance, in spite of being one of the largest HSP cohorts, the number of SPG4 patients was limited, especially for a genotype-phenotype correlation study. Furthermore, the families interested in participating in this study were probably those who had remained undiagnosed after the primary SPAST testing; therefore, the proportion of SPG4 patients in this study may be underestimated. In addition, clinical signs and symptoms for some of the patients were missing.
In conclusion, we report one of the largest SPG4 cohorts, with 41 different mutations including 6 novel mutations. We suggest that modes of inheritance other than autosomal dominant could be involved in SPG4. Our study sheds light on some rarely reported or unreported aspects of the disease and helps improve genetic counseling and clinical trials conducted on SPG4. Towards this goal, larger efforts and international collaborations are required.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021.

Financial Disclosures of All Authors (for the Preceding 12 Months)
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. all unrelated to the current study. All other authors report nothing to disclose.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. .348 CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 20, 2021.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 20, 2021. ; https://doi.org/10.1101/2021.07.20.21259482 doi: medRxiv preprint