FKRP-dependent glycosylation of fibronectin regulates muscle pathology in muscular dystrophy

The muscular dystrophies encompass a broad range of pathologies with varied clinical outcomes. In the case of patients carrying defects in fukutin-related protein (FKRP), these diverse pathologies arise from mutations within the same gene. This is surprising as FKRP is a glycosyltransferase, whose only identified function is to transfer ribitol-5-phosphate to α-dystroglycan (α-DG). Although this modification is critical for extracellular matrix attachment, α-DG’s glycosylation status relates poorly to disease severity, suggesting the existence of unidentified FKRP targets. Here we reveal that FKRP directs sialylation of fibronectin, a process essential for collagen recruitment to the muscle basement membrane. Thus, our results reveal that FKRP simultaneously regulates the two major muscle-ECM linkages essential for fibre survival, and establishes a new disease axis for the muscular dystrophies.

5. Figure 1 -the names and colors used to describe the groups are confusing and misleading. Use of the same colors to represent different things in the same figure is confusing to the reader. Also, consider using "wild type" or WT instead of fkrp+/+ to be consistent with in-text writing.
6. All figures -I am not a fan of the showing points outside the 95% confidence range but I will leave this up to the authors' discretion.
7. All figures -remove the two-sided arrow to indicate comparison between groups and use brackets instead -cleaner look.
8. The correct abbreviation is LGMD2I (with a capital I) or the newer abbreviation LMGDR9.
Reviewer #2: Remarks to the Author: Wood et al. reported that FKRP, one of causative genes of dystroglycanopathies, has a new function. They used FKRP-mutated zebrafish and human patient myoblasts, and they found FKRPdependent sialylation, specific N-glycans, of fibronectin is important for the formation of fibronectin-collage axis. Based on the results, they proposed sialylation of N-glycan-fibronectin regulates muscle pathology in muscular dystrophy. However, to reach the conclusions, the authors should provide more convincing data.
1) The authors considered three possibilities of FKRP-dependent sialylation and thought Fig. 4e model is most possible. However, they did not show direct binding myosin10 and FKRP. Direct binding data of normal and mutant FKRP proteins to myosin10 is necessary. It is also necessary to show direct binding of fibronectin-myosin10 because they performed proteomic analysis of fibronectin-pull down fraction and found myosin10 as one of candidate proteins. It also requires to perform similar proteomic experiments using the Golgi-enriched fraction, or at least after removal of cytosol.
2) The authors claimed myosin10 localized in the luminal side of Golgi. However, it could not conclude by STED study. Actually, location of Golgi(58k) is reported in the cytosol side of Golgi. They should give concrete data of location myosin10 by different methods.
3) The authors concluded that FKRP did not catalyze fibronectin sialylation directly (Ex Fig.9). line 892, there is no description how to prepare enzymes? Does this FKRP protein have ribitol-Ptransferase activity?
4) It is of interest that the authors found the increment of ST6Gal1 expression in FKRP mutants although they tried to find down regulated genes. However, it is unclear why they could not find any difference total sialylation (Ex Fig. 6) in control and patients although ST6Gal1 increased? 5) Lines 824, 825, pathological information including severity of CMD patient and LGMD2I patient was not described at all. They may be different. IIH6 staining patterns of both should be shown. However, sialylation of both fibronectin did not show any difference (Fig. 3b), suggesting that sialylation level of fibronectin does not correlate directly with disease severity. By the way, in Fig.  3a, Red box may be Blue box. Line 479, g should be removed. It is noted that this review could not see the increase of asialobianntenary N-glycan peak in patients. Why? 6) Line 240, CDP-ribitol should be ribitol-phosphate. 7) Lines 553~555, this explanation is wrong, because fukutin and FKRP sequentially contributed the a-dystroglycan-glycan formation. 8) Two reference sections could be found (page 12 and page 48). The latter one may be correct. 9) Fig. 4, why fibronectin staining patterns of patients are different from control? 10) Table 1, why the authors did not examine myosin 9? Refs 32 and 33 are related to myosin 9 (non-muscle myosin IIA). Manuscript: NCOMMS-20-10420 FKRP-dependent glycosylation of fibronectin regulates muscle pathology in muscular dystrophy. Below we systematically address each of the reviewer's concerns. Corresponding text changes are highlighted in blue within the manuscript. Overall, this is a well-written manuscript, except for a few gene nomenclature issues, with a key finding on the modification of fibronectin requiring functional FKRP protein which is disrupted in LGMD2I and WWS. However, there are key aspects of their findings that need clarification to fully understand the nature of the regulatory steps of FKRP on fibronectin sialylation.
Major Comments: Q1.1. "The authors state at that α-dystroglycan glycosylation status has been shown that it does not track with disease progression or severity. Does the N-glycan sialylation status of fibronectin track with either disease progression or severity?" Response. Our original data did detect a dramatic drop, but of similar magnitude, in the sialylation levels of fibronectin in both of the patient samples tested (Figure 3a, b). However, although this data did not show a difference in sialylation levels between the two patient samples, we did observe a highly significant difference in fibronectin-collagen binding between these two samples, that correlated with clinical severity (Fig 3c). This was also the case for the Golgi defects we characterised (Fig 4), both for the Myosin10-fibronectin colocalisation and the Myosin10 fragmentation phenotype, which tracked with clinical severity. Thus, although we could not resolve disease-specific severity differences in the levels of fibronectin sialylation, we believe it likely that there are patient specific changes in sialylation levels on fibronectin that we have not been able to detect with current approaches. Given that only 3 of the 7 potential biantennary glycan sites on fibronectin reside in the collagen binding domain, and that there is a limit to the amount of fibronectin we can pull down from patient myoblast samples, our N-glycan MS/MS technique may not have been sensitive enough to detect subtle changes in fibronectin sialylation that may occur on individual functionally important sites. In an attempt to resolve this question, we undertook glycopeptide analysis of fibronectin derived from cultured patient cells with a view to improve our sensitivity of detection of sialylation levels on fibronectin glycosylation sites within the collagen binding domain. Although the Packer lab is one the global leaders in undertaking such analyses we were unable to provide data that distinguishes between the two stages of the disease in terms of fibronectin sialylation. using this technically challenging approach. Unfortunately, this technology works well for high concentrations of protein obtained from traditional protein purification techniques, not rare muscle patient samples with limited passage potential. Nevertheless, the decrease seen in both patients, relative to the control, does implicate fibronectin sialylation as a significant factor in the disease. We have now also added additional text to the results at line151-161 and line 233 page 10 explaining the cellular phenotypic tracking with clinical severity more clearly.

Q1.2. In human samples (looking at
LGMD2I and CMD patient cells), they found significant reduction in N-glycan sialylation compare to healthy controls in two peaks, that are composed of NeuAc1Hex5HexNAc4 and NeuAc2Hex5HexNAc4. Can the authors comment whether they believe both sites are important for the interaction with collagen? Or would partial sialylation of just one of these sites be enough to restore collagen interaction? Response. The reviewer raises an interesting question about the sialylation of the mono-and di-biantennary glycans released from the cellular proteins. However, the characterization of the glycans was carried out on N-glycans released from the proteins and as such we unfortunately have no information on the site occupancy of these two structures on the protein.

Q1.3. Increasing glycosylation status of α-dystroglycan has been shown to be beneficial in fkrp-deficient backgrounds. Can the authors discuss the relevant therapeutic targets that their study now opens up for patients with LGMD2I or CMD? Do they believe that somehow increasing sialylation of fibronectin can be a therapeutic target?
Response. This is an interesting point the reviewer makes and is the reason we carried out the rescue experiments by injecting fkrp-/-mutant fish with non-sialylated and fully-sialylated fibronectin. From these analyses, presented in Fig. 3 of the original submission, it is clear that fully-sialylated fibronectin rescues the zebrafish fkrp mutant model, so indeed we think increasing sialylation of fibronectin can be a valid therapeutic approach. However, we know from previous work with laminin that these full-length proteins are unlikely to be used in the clinic as they are difficult to handle and deliver, despite providing excellent proof of principle in animal models. We rather envision investigating a possible therapeutic strategy that stabilises Mysoin10 in the Golgi to improve fibronectin sialylation. We have now added a discussion of the therapeutic relevance of increasing fibronectin sialylation to page 8 line 180 of the revised manuscript in line with the reviewer's request.

Q1.4. Were any neurological and/or vascular phenotypes observed in the fkrp mutant zebrafish? This has been reported in other reports and it would be good to validate the occurrence of these pathological findings. Bailey et al., Skeletal Muscle, 2019 reported altered NMJ formation in fkrp mutant zebrafish, the authors should comment on whether fibronectin disruption affects NMJ formation and/or myofiber detachment potentially independent of FKRP expression functional status.
Response. The reviewer is indeed correct that a number of authors have reported vascular defects in zebrafish using a fkrp-targeting morpholino approach and this finding is also reported in Bailey et al. with additional interesting observations on NMJ formation. However, all these analyses were performed on fkrp morphants not on germline mutants which form the basis of our analyses. In line with the reviewer's comments we have further examined our fkrp germline mutants for vascular and neuromuscular defects. We used angiography with injected Fluorescent-Dextran and independently evaluated the vasculature of Evans blue dye injected larvae to examine vascular permability with the myotome. To study neuronal integrity, we stained for motoneuron arbours and junctions in the myotome. None of these analyses showed vascular and neuromuscular disruption independent of the muscle loss or MTJ disruption within the myotome of fkrp mutants. The only time we could see defects in vascular or neuromuscular pattern were in regions of muscle loss or myosepta disruption in fkrp mutants. This new data is presented in Extended Data Figure 2f-g'' for the vascular analyses and in Extended Data Figure 3j, j' for neuronal defects. There are a number of potential explanations for this discrepancy such as the propensity for off target effects when morpholinos are used or the genetic compensation that is triggered by germline mutants that is not generated by morpholino knockdown 1 . It is unclear at this point which explanation is the most likely. We have added text to manuscript explaining these new data points at pg4. In terms of neurological defects, we also wish to point out that in the original submission we also carried out a detailed analysis of the retina which of course is part of the nervous system, given the muscle eye brain phenotype of some FKRP patients. In zebrafish, the retina replicates the neuronal layers in the human eye and as expected we did find abnormalities in basement membranes and modest defects in retinal layering in our mutants. In regards to a potential role for fibronectin in NMJ formation, a search of the literature suggests no obvious mechanism for a role of fibronectin in NMJ formation. By contrast fibronectin is one of the primary components of fiber attachment in the myotome acting initially as a scaffold at the MTJ and later as a regulator of passive tension through control of myotome elasticity. Hence the loss of passive tension and MTJ boundary crossing phenotype we describe. However, it is highly likely that fibronectin's scaffolding role is multidimensional in this context, and almost certainly some of the roles it plays at the MTJ will be independent of its FKRP-mediated collagen binding activity. Q1.6. The correct zebrafish gene name for dystroglycan is dag1. The authors should be consistent in their naming and change "dag" to "dag1" in the manuscript.
Response. Text altered throughout as suggested.

Q1.7. RNA-sequencing (and ideally mass-spectrometry data) should be deposited in a public database for free access.
Response. We are very happy to deposit all the data sets in public repositories. Specifically, the RNA-seq dataset has been deposited at the NCBI Gene Expression Omnibus (GEO) repository under accession number GSE163213 with a link that will be made public on acceptance of the manuscript. The raw data from PGC-LC-MS/MS N-glycan analysis of fibronectin isolated from control and patient myoblast cell lines ( Figure 3) and whole cell lysates (Extended Figure 6) has been submitted to GlycoPOST under the ID GPST000159. The data is accessible at the following URL upon acceptance of the manuscript: https://glycopost.glycosmos.org/preview/150345145fd6fcbaa8caa. Statements outlining how these data sets can be accessed have been made within the methods on page 52.
Minor Comments: Q1.9. Line 89 -"…loss of function15, we postulated…" Move the comma to after "function". Response. The text has been altered as requested.

Q1.11. Line 219 -Remove the " 's" on fibronectin.
Response. The text has been altered as requested.
Response. The text has been altered as requested.
Q1.12. Figure 1 -the names and colors used to describe the groups are confusing and misleading. Use of the same colors to represent different things in the same figure is confusing to the reader. Also, consider using "wild type" or WT instead of fkrp+/+ to be consistent with in-text writing. Response: We agree with the reviewer the colors on the force graph (figure 1e) are confusing when compared with that of the boundary crossing and fibre detachment analysis (figure 1c and d) and we have altered figure 1e in line with the reviewer's suggestion. In terms of the allele designation and the use of wildtype/WT, we have used the allele designation of fkrp+/+ to signify sibling rather than WT status. We use the designation wildtype/WT only when we refer to the Tübingen wildtype strain and fkrp+/+ when we are referring to a genotyped sibling control where heterozygotes have been excluded from analysis. The broad use of the wildtype designation could potentially mask the reader to the technical design of some of our experiments.

Q1.13. All figures -I am not a fan of the showing points outside the 95% confidence range but I will leave this up to the authors' discretion.
Response. We understand the reviewer's point here, but we believe it is Nature editorial policy to preferably represent all data points on graphs and have tried to adopt this policy in our graphical representations.

Q1.14. All figures -remove the two-sided arrow to indicate comparison between groups and use brackets instead -cleaner look.
Response. We removed all two-sided arrows except for the ones used on the passive force graphs (Fig 2g) where brackets may not convey the analysis.

Q1.15. The correct abbreviation is LGMD2I (with a capital I) or the newer abbreviation LMGDR9.
Response: The text has been altered as requested. We have stuck with the LGMD2I nomenclature as it is the most prevalent in the literature but are also happy to use the newer LMGDR9 designation if the reviewer believes it is more accurate to do so.  Figure 9a, b.

Q2.2. The authors claimed myosin10 localized in the luminal side of Golgi. However, it could not conclude by STED study. Actually, location of Golgi(58k) is reported in the cytosol side of Golgi. They should give concrete data of location myosin10 by different methods.
Response. We are happy to provide the further experiments requested by the reviewer. To undertake these additional analyses, we employed Airyscan super-resolution imaging in conjunction with a different Golgi marker to better localise Myosin10. The Airyscanner approach has the advantage of being able to 3D render the Golgi which allows us to better visualise localisation of Myosin10 as requested by the reviewer. Although we acknowledge that the Golgi 58K protein does localise on the cytosolic side of the Golgi, we wish to point out that it is a widely used general marker of the Golgi membrane. However, to broaden our analyses, in the manner the reviewer requests, we added an additional Golgi marker in the Airyscanner observations, Golgi Reassembly Stacking Protein 2 (GORASP2) 1 , which is a general Golgi membrane protein.
We have now comprehensively extended our investigation to include 3 additional experimental and 3 additional technical replicates, imaging a total of 9 areas for each of the 9 replicates with the Airyscanner method. We also include now an additional supplementary video (2) rendered in 3D software that gives a better view inside the Golgi in super resolution. A rendered panel figure and a sphericity analysis of the new data (which nicely documents the changes that occur within the Golgi of mutant cells) and an analysis of co-localization of the new data using the Golgi as a mask is also provided. The new statistically validated, data confirms our findings from the original STED data. Collectively, this data confirms that Fibronectin is co-localized with Myosin10 in the Golgi of control cells, a distribution that is significantly altered in FKRP patient-derived cell lines. 1. Zhang X, Wang Y. The Golgi stacking protein GORASP2/GRASP55 serves as an energy sensor to promote autophagosome maturation under glucose starvation. Autophagy. 2018;14(9):1649-1651 Fig.9). line 892, there is no description how to prepare enzymes? Does this FKRP protein have ribitol-P-transferase activity? Response. We apologise for the methods oversight we have now added the requested enzyme preparation protocol into the manuscript methods at page 50. On the matter of testing if our FKRP protein has ribitol-P-transferase activity we do appreciate the reviewer's point here.

Q2.3. The authors concluded that FKRP did not catalyze fibronectin sialylation directly (Ex
Unfortunately, however, the reviewer's request to confirm that our FKRP protein has ribitol-P-transferase activity is a very difficult one to achieve. There are several technical challenges to establish a FKRP activity assay. Firstly, the known FKRP donor substrate CDP-ribitol, unlike CMP-Neu5Ac for sialylation, is not commercially available. Secondly, the enzyme activity required to add ribitol-P also requires the production of human ISPD (isoprenoid synthase domain-containing protein) to allow the in vitro synthesis of CDP-ribitol 1 . Thirdly, the CDP-ribitol is added to a phosphorylated tetrasaccharide structure on a-DG. To synthesise this substrate requires production of the a-DG fragment and five enzymes (POMT1/2, SGK196, POMGNT2, B3GALNT2, and fukutin) 2 . Apart from recombinant protein expression and production of all these enzymes, it would also require HPLC method development to purify products for each synthesis step. The complexity of this is further demonstrated in a recent publication by Kuwabara et al. 3 where they further demonstrate that the core M3 Man-6-P peptide with Phosphate on the O-Mannose is necessary for enzymatic activity using Chromatography and MALDI-mass spectrometry. Thus, to perform this alternate assay in our context would require chemical synthesis of CDP Ribitol, isolation of Core M3 Phopho glycopeptide and further extensive analytical methods as shown by Kuwabara et al., 2020 (Methods) and Fig 5, Suppl Fig 5 and Suppl Fig 6 (Results). In fact, the demonstration of this activity is a full manuscript in its own right and I hope the reviewer may concede that it is outside the scope of this study given that sialylation activity is the point under contention for this manuscript.
However, we have confirmed that the FKRP recombinant protein that we have produced is correct by both molecular mass and mass spectroscopic identification of FKRP via SDS-PAGE gel band extraction, which is how most commercially available recombinant proteins are assessed. We provide this data on the production of the protein in Extended Data Figure 9.

Q2.4.
It is of interest that the authors found the increment of ST6Gal1 expression in FKRP mutants although they tried to find down regulated genes. However, it is unclear why they could not find any difference total sialylation (Ex Fig. 6) in control and patients although ST6Gal1 increased? Response. The reviewer raises an interesting point here that had also been unclear to us. Prompted by the reviewer's comments we revisited this issue. In the original sample preparation protocol we performed for this analysis, we analysed the glycans released from the proteins in the cell lysate after the immunoprecipitation of fibronectin. We were concerned that this may have affected the sensitivity of the assay. We therefore repeated our analysis using a different sample preparation method in which glycans were directly released from the membrane proteins of total cell lysates. This has indeed proved to be a better approach, which produced higher sensitivity and improved spectra from which to calculate the total cell protein sialylation. The total glycosylation now matches the RNA seq data as expected in that we see an increase in relative abundance of sialylation, specifically in 2,6 sialylation in the patient samples that correlates with the ST6Gal1 increase in expression. The data is provided in a revised Extended Data Figure 6.
Q2.5. Lines 824, 825, pathological information including severity of CMD patient and LGMD2I patient was not described at all. They may be different. IIH6 staining patterns of both should be shown. However, sialylation of both fibronectin did not show any difference (Fig. 3b), suggesting that sialylation level of fibronectin does not correlate directly with disease severity. By the way, in Fig. 3a, Red box may be Blue box. Line 479, g should be removed. It is noted that this review could not see the increase of asialobianntenary N-glycan peak in patients. Why? Response. We agree with the reviewer that more clinical data is an important addition to the manuscript. We have now included a table (Table 1) detailing clinical phenotype of the patients from which cells were derived. We also include information on both the IIH6 and VIA4 status on sections and referenced publications concerning these patients.
In terms of the same sialylation levels of fibronectin in both patient samples we tackle this question in response to Q1 of Reviewer1 and refer the Reviewer 2 to this answer. The red box is indeed blue and we have now altered the corresponding text. We apologize for this error.
To answer the query as to whether there is a complementary increase of asialylated species with the decrease in sialylation observed in the patients we undertook further analyses of our spectra. As we reported in the original submission the blue boxed peaks m/z 965.9 and 1111.4 in Fig 3a, that represent mono-and di-sialylated biantennary glycans in the control, are lost in the patient samples and the reviewer is correct, that this should lead to a corresponding increase in asialobianntenary N-glycans. Prompted by the reviewer's insightful question we have now revisited our data and we do see the corresponding increase of asialo N-glycans on fibronectin isolated from the mutant cells, relative to the sialylated species (as shown in the following figure extracted from Fig 3a). It confirms the loss of sialylation is accompanied by increase of asialo-species. We provide the data here as we think as it essentially duplicates Fig 3a, but are happy to include it in the manuscript if the reviewer thinks it is a necessary point of emphasis in resubmission.
We can also see a distinct drop in asialobiantennary non-fucosylated (A2G2-m/z 820.39) and fucosylated (F1A2G2-m/z 893.32) peaks in the total cell lysate (Extended Data Fig.6) in the FKRP mutant patient cell lines compared to healthy controls. Accordingly, a complementary increase in relative abundance of asialylated species relative to the sialylated species is observed.