Bioinformatic identification of previously unrecognized amyloidogenic proteins

Abstract

Low-complexity domains (LCDs) of proteins have been shown to self-associate, and pathogenic mutations within these domains often drive the proteins into amyloid aggregation associated with disease. These domains may be especially susceptible to amyloidogenic mutations because they are commonly intrinsically disordered and function in self-association. The question therefore arises whether a search for pathogenic mutations in LCDs of the human proteome can lead to identification of other proteins associated with amyloid disease. Here, we take a computational approach to identify documented pathogenic mutations within LCDs that may favor amyloid formation. Using this approach, we identify numerous known amyloidogenic mutations, including several such mutations within proteins previously unidentified as amyloidogenic. Among the latter group, we focus on two mutations within the TRK-fused gene protein (TFG), known to play roles in protein secretion and innate immunity, which are associated with two different peripheral neuropathies. We show that both mutations increase the propensity of TFG to form amyloid fibrils. We therefore conclude that TFG is a novel amyloid protein and propose that the diseases associated with its mutant forms may be amyloidoses. Low-complexity domains (LCDs) of proteins have been shown to self-associate, and pathogenic mutations within these domains often drive the proteins into amyloid aggregation associated with disease. These domains may be especially susceptible to amyloidogenic mutations because they are commonly intrinsically disordered and function in self-association. The question therefore arises whether a search for pathogenic mutations in LCDs of the human proteome can lead to identification of other proteins associated with amyloid disease. Here, we take a computational approach to identify documented pathogenic mutations within LCDs that may favor amyloid formation. Using this approach, we identify numerous known amyloidogenic mutations, including several such mutations within proteins previously unidentified as amyloidogenic. Among the latter group, we focus on two mutations within the TRK-fused gene protein (TFG), known to play roles in protein secretion and innate immunity, which are associated with two different peripheral neuropathies. We show that both mutations increase the propensity of TFG to form amyloid fibrils. We therefore conclude that TFG is a novel amyloid protein and propose that the diseases associated with its mutant forms may be amyloidoses. Low-complexity domains (LCDs) are common, but functionally mysterious regions of proteins in the human proteome, of which several are associated with amyloidoses (1Nolan M. Talbot K. Ansorge O. Pathogenesis of FUS-associated ALS and FTD: Insights from rodent models.Acta Neuropathol. Commun. 2016; 4: 99Crossref PubMed Scopus (63) Google Scholar, 2Pesiridis G.S. Lee V.M.-Y. Trojanowski J.Q. Mutations in TDP-43 link glycine-rich domain functions to amyotrophic lateral sclerosis.Hum. Mol. Genet. 2009; 18: R156-162Crossref PubMed Scopus (226) Google Scholar, 3Kim H.J. Kim N.C. Wang Y.-D. Scarborough E.A. Moore J. Diaz Z. MacLea K.S. Freibaum B. Li S. Molliex A. Kanagaraj A.P. Carter R. Boylan K.B. Wojtas A.M. Rademakers R. et al.Mutations in prion-like domains in hnRNPA2B1 and hnRNPA1 cause multisystem proteinopathy and ALS.Nature. 2013; 495: 467-473Crossref PubMed Scopus (934) Google Scholar). LCDs are characterized by long segments made up of relatively low sequence diversity and are also commonly intrinsically disordered. LCDs are thought to be integral to the self-association of some proteins involved in RNA binding, the formation of membraneless organelles, and the self-association of intermediate filament proteins (4Nott T.J. Petsalaki E. Farber P. Jervis D. Fussner E. Plochowietz A. Craggs T.D. Bazett-Jones D.P. Pawson T. Forman-Kay J.D. Baldwin A.J. Phase transition of a disordered nuage protein generates environmentally responsive membraneless organelles.Mol. Cell. 2015; 57: 936-947Abstract Full Text Full Text PDF PubMed Scopus (859) Google Scholar, 5Elbaum-Garfinkle S. Kim Y. Szczepaniak K. Chen C.C.-H. Eckmann C.R. Myong S. Brangwynne C.P. The disordered P granule protein LAF-1 drives phase separation into droplets with tunable viscosity and dynamics.Proc. Natl. Acad. Sci. U. S. A. 2015; 112: 7189-7194Crossref PubMed Scopus (577) Google Scholar, 6Kato M. Han T.W. Xie S. Shi K. Du X. Wu L.C. Mirzaei H. Goldsmith E.J. Longgood J. Pei J. Grishin N.V. Frantz D.E. Schneider J.W. Chen S. Li L. et al.Cell-free formation of RNA granules: Low complexity sequence domains form dynamic fibers within hydrogels.Cell. 2012; 149: 753-767Abstract Full Text Full Text PDF PubMed Scopus (1214) Google Scholar, 7Zhou X. Lin Y. Kato M. Mori E. Liszczak G. Sutherland L. Sysoev V.O. Murray D.T. Tycko R. McKnight S.L. Transiently structured head domains control intermediate filament assembly.Proc. Natl. Acad. Sci. U. S. A. 2021; 118e2022121118Crossref Scopus (10) Google Scholar, 8Sysoev V.O. Kato M. Sutherland L. Hu R. McKnight S.L. Murray D.T. Dynamic structural order of a low-complexity domain facilitates assembly of intermediate filaments.Proc. Natl. Acad. Sci. U. S. A. 2020; 117: 23510-23518Crossref PubMed Scopus (10) Google Scholar, 9Murray K.A. Hu C.J. Seidler P. Hughes M.P. Salwinski L. Sawaya M. Pan H. Eisenberg D.S. Identifying amyloid-related diseases by mapping mutations in low-complexity protein domains to pathologies.Nat. Struct. Mol. Biol. 2022; (In press)Google Scholar), but not all proteins with LCDs exhibit these functions. Whereas, subsequent dissociation of these complexes is a hallmark of the normal function of LCDs, proteins with LCDs may become prone to aggregate irreversibly into pathogenic amyloids because of missense mutations, which encourage protein misfolding (3Kim H.J. Kim N.C. Wang Y.-D. Scarborough E.A. Moore J. Diaz Z. MacLea K.S. Freibaum B. Li S. Molliex A. Kanagaraj A.P. Carter R. Boylan K.B. Wojtas A.M. Rademakers R. et al.Mutations in prion-like domains in hnRNPA2B1 and hnRNPA1 cause multisystem proteinopathy and ALS.Nature. 2013; 495: 467-473Crossref PubMed Scopus (934) Google Scholar, 10Murray D.T. Zhou X. Kato M. Xiang S. Tycko R. McKnight S.L. Structural characterization of the D290V mutation site in hnRNPA2 low-complexity-domain polymers.Proc. Natl. Acad. Sci. U. S. A. 2018; 115: E9782-E9791Crossref PubMed Scopus (25) Google Scholar, 11Nomura T. Watanabe S. Kaneko K. Yamanaka K. Nukina N. Furukawa Y. Intranuclear aggregation of mutant FUS/TLS as a molecular pathomechanism of amyotrophic lateral sclerosis.J. Biol. Chem. 2014; 289: 1192-1202Abstract Full Text Full Text PDF PubMed Scopus (82) Google Scholar). Some examples of amyloidogenic LCD-containing proteins are FUS, TDP43, and HNRNPA1, which are all associated with ALS. Because LCDs are often disordered, cross-β structures are present in LCD condensates (12Guenther E.L. Cao Q. Trinh H. Lu J. Sawaya M.R. Cascio D. Boyer D.R. Rodriguez J.A. Hughes M.P. Eisenberg D.S. Atomic structures of TDP-43 LCD segments and insights into reversible or pathogenic aggregation.Nat. Struct. Mol. Biol. 2018; 25: 463-471Crossref PubMed Scopus (107) Google Scholar), and many proteins containing them form amyloids in disease; we speculate that these domains may be more susceptible to mutations that cause the formation of an amyloid to be energetically favorable. Under this assumption, we chose to focus our search for novel amyloidogenic proteins on those proteins that contain an LCD (Fig. 1). This search for unidentified amyloidogenic proteins based on pathogenic mutations expands on previous work (9Murray K.A. Hu C.J. Seidler P. Hughes M.P. Salwinski L. Sawaya M. Pan H. Eisenberg D.S. Identifying amyloid-related diseases by mapping mutations in low-complexity protein domains to pathologies.Nat. Struct. Mol. Biol. 2022; (In press)Google Scholar, 13Kato M. Zhou X. McKnight S.L. How do protein domains of low sequence complexity work?.RNA. 2022; 28: 3-15Crossref PubMed Scopus (2) Google Scholar) by considering a larger subset of the human proteome. Our approach is agnostic to details about the queried proteins (besides identifying LCDs based on amino acid sequence) such as their functions or the diseases with which they are associated. Also, while our approach does identify many known amyloidogenic proteins, our focus is solely on those that have never been documented to form amyloid fibers either in vivo or in vitro. Here, we advance computational screening methods to identify mutations that may cause a functional LCD to become amyloidogenic (9Murray K.A. Hu C.J. Seidler P. Hughes M.P. Salwinski L. Sawaya M. Pan H. Eisenberg D.S. Identifying amyloid-related diseases by mapping mutations in low-complexity protein domains to pathologies.Nat. Struct. Mol. Biol. 2022; (In press)Google Scholar). We define an amyloid as an irreversible fibrous protein aggregate with a cross-β-sheet scaffold. The common methods of experimental identification of amyloid are the binding of such as or and the to separation of and separation of are in a of diseases from to L.C. protein PubMed Scopus Google Scholar). These amyloidoses are characterized by of amyloid by not lead to and G. Pathogenesis and PubMed Scopus Google Scholar). amyloid drives is to identify the protein to for the disease. We propose that many diseases have amyloid to their this our numerous known amyloidogenic mutations as as many mutations not previously associated with amyloidoses. Among the of mutations, we that two from the protein TRK-fused gene protein increase the amyloid propensity of the The identification of this and other amyloid proteins is for the and the of their associated identify LCDs, we the S. of complexity in amino acid and sequence Chem. Scopus Google Scholar) to the human proteome to amino acid segments as either complexity or low We an LCD as low-complexity of amino with for amino in a Under these human proteins contain We The protein in 2021; PubMed Scopus Google Scholar), in A. PubMed Scopus Google Scholar), and Lee M. S. B. J. D. K. K. Z. A. et to and 2018; PubMed Scopus Google Scholar) for pathogenic missense mutations within the LCDs of these proteins and documented This of mutations while agnostic to the functional for the these mutations are not some that are pathogenic because they increase amyloid propensity but also many that are pathogenic for other to Among the that up LCDs, are the common, by and (Fig. is by the common to be in with of the mutations from (Fig. The common mutation is from to in order by to to and to The common to be in is with This is by mutations, up of the documented These that are especially in normal function of human we to identify the mutations from the of pathogenic mutations that increase the propensity of a functional sequence to form a the common protein amyloid M.R. Hughes M.P. Rodriguez J.A. R. Eisenberg D.S. The amyloid and 2021; Full Text Full Text PDF PubMed Scopus Google Scholar). we a that the propensity of segments within proteins L. R. Eisenberg D. Identifying the proteins of Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). the of segments in the of a for we two on the containing the and containing the mutant of these all containing the of by and are not energetically in segments containing this to have is a common in LCDs (Fig. this a of the to and does not this of the (Fig. in the LCDs the human proteome with propensity to form amyloid their of and mutant segments in LCDs by Because mutation generates the that to the of or with the from to mutant is for The mutations that do not the The and are both of the the for a a of the of the that is the of containing to a with a of and a mutant with a of a mutation that the amyloid of the of in the of low-complexity we amyloidogenic mutations as we a mutation to be amyloidogenic a containing the an the of with amyloid and the containing the mutant an or to with amyloid and Mutations that these to cause a amyloid This mutations, the common to to to and to (Fig. to amino are in this to are they are the common of pathogenic mutation in LCDs in This is amino acid on the of a to be to as as to on of other the and these other amyloid formation energetically that is the in whether the mutation is to be amyloidogenic or not on which into other for the amino acid into is in whether the mutation is to be amyloidogenic or are many documented mutations within LCDs that may drive a functional sequence to become but mutations from and are more to be of this that our approach is of identifying amyloidogenic mutations, we the for known amyloidogenic mutations and of the mutations in hnRNPA1 and and the mutation in hnRNPA2B1 have been shown to formation (3Kim H.J. Kim N.C. Wang Y.-D. Scarborough E.A. Moore J. Diaz Z. MacLea K.S. Freibaum B. Li S. Molliex A. Kanagaraj A.P. Carter R. Boylan K.B. Wojtas A.M. Rademakers R. et al.Mutations in prion-like domains in hnRNPA2B1 and hnRNPA1 cause multisystem proteinopathy and ALS.Nature. 2013; 495: 467-473Crossref PubMed Scopus (934) Google Scholar). of the mutations in been to aggregation propensity (9Murray K.A. Hu C.J. Seidler P. Hughes M.P. Salwinski L. Sawaya M. Pan H. Eisenberg D.S. Identifying amyloid-related diseases by mapping mutations in low-complexity protein domains to pathologies.Nat. Struct. Mol. Biol. 2022; (In press)Google Scholar). a protein that can form amyloid fibers in N. K. Y. M. J. forms amyloid that in Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar), mutations in the and that are associated with and been shown to cause aggregation X. Lin Y. Kato M. Mori E. Liszczak G. Sutherland L. Sysoev V.O. Murray D.T. Tycko R. McKnight S.L. Transiently structured head domains control intermediate filament assembly.Proc. Natl. Acad. Sci. U. S. A. 2021; 118e2022121118Crossref Scopus (10) Google Scholar). mutation in a pathogenic A.J. J. mutation which the of the gene Genet. Scholar), and an in this protein been shown to formation T. B. A. E. R. E. to an in the binding protein cause Sci. PubMed Scopus Google Scholar). amyloidogenic proteins and also have mutations that on the but these mutations have not been for amyloid These examples that our approach can identify some mutations that to the formation of amyloid our approach novel amyloid mutations, we of the proteins, of its TFG into and its LCD facilitates these to form larger complexes A. N. M. Wang L. A. TFG and J. 2015; PubMed Scopus Google Scholar). The two mutations in TFG that by our and have been associated with and and with Wu Lin Lee novel TFG mutation and TFG 2014; PubMed Scopus Google Scholar, H. M. T. O. J. Y. H. J. B. Y. A. H. Y. K. et TRK-fused gene is in and with J. Genet. 2012; Full Text Full Text PDF PubMed Scopus Google Scholar), and both mutations shown to in aggregation of the We also that the protein and the protein with the mutation are to phase in in the of a but with the the protein forms (Fig. We and the LCD of TFG with to increase in the containing the mutation and containing the mutation for with a that in the of amyloid fibers H. of amyloid structures with PubMed Scopus Google Scholar, M.R. to amyloid formation in PubMed Scopus Google Scholar). mutant the not (Fig. The of fibers from the mutant and the of fibers from the by (Fig. mutant fibers an of amyloid The amyloid of the TFG to fibers also of containing the fibers and to which the fibers the The fibers in an with the of the fibers to the of the The for both mutant fibers the of an amyloid the and the (Fig. We the in the with the because of and the sequence to form a (Fig. we TFG amyloid as by our The common within LCDs to be in are and all of which are known to be for the normal function of many LCDs, especially with to the of J. Lee X. M. S. R. A. D. S. molecular the for phase separation of prion-like RNA binding 2018; Full Text Full Text PDF PubMed Scopus Google Scholar, R. J. K.A. J. of hnRNPA2 low-complexity domain and phase separation by mutation and Cell. 2018; Full Text Full Text PDF PubMed Scopus Google Scholar, S. low-complexity of protein and phase Biol. Chem. Full Text Full Text PDF PubMed Scopus Google Scholar). Because of their functional in LCDs, these can in for to but in this we are in pathogenic mutations in LCDs that in and the in amyloidogenic mutations in LCDs are different pathogenic mutations in LCDs, with the common to have mutations to be amyloidogenic The of mutations (Fig. can to some be in of of a can the a to a pathogenic of can favor formation of to pathogenic are with of which we as the commonly associated with (Fig. and with the to be amyloidogenic (Fig. to the which is in the of protein droplets J. Lee X. M. S. R. A. D. S. molecular the for phase separation of prion-like RNA binding 2018; Full Text Full Text PDF PubMed Scopus Google Scholar). of a larger a an and does not form or M. of amino in PubMed Scopus Google Scholar). This that are other to to the of an amyloid that is with We previously that can into the of in amyloid structures from low-complexity segments M.P. Sawaya M.R. Boyer D.R. L. Rodriguez J.A. Cascio D. L. T. Eisenberg D.S. Atomic structures of low-complexity protein segments that 2018; PubMed Scopus Google Scholar), which may in low-complexity amyloid We have also that lead to in these low-complexity amyloid which may also to their K.A. D. Hughes M.P. Sawaya M.R. Hu C.J. Eisenberg D. to reversible amyloid 2022; Scopus Google Scholar). missense mutations, can into and (Fig. The common mutations that we to be amyloidogenic are to to and to to is the common mutation to be amyloidogenic. a relatively which is the and this to form from to the formation of a for this especially the sequence is amyloid prone other by the as is the for some glycine-rich LCD proteins and forms complexes with of as of its of the TFG mutations that amyloid propensity is a to mutation to is the common which is to amyloid The is a can form with other These increase the of proteins, but these are can cross-β and to the formation of amyloid fibers Y. J. X. K. in diseases 2013; PubMed Scopus Google Scholar, L. M. H. M. L. R. B. T. amyloid formation of the Biol. 2020; 28: PubMed Scopus Google Scholar). This is especially the to mutation a sequence with a in a that is to of as in many LCDs in which are not common (Fig. to mutation in been shown to increase amyloid propensity (9Murray K.A. Hu C.J. Seidler P. Hughes M.P. Salwinski L. Sawaya M. Pan H. Eisenberg D.S. Identifying amyloid-related diseases by mapping mutations in low-complexity protein domains to pathologies.Nat. Struct. Mol. Biol. 2022; (In press)Google Scholar). to is the common of the mutations, which are to increase amyloid is a to the of an amyloid a that is not to form with are more common on the of pathogenic but do in for is the formation of a and to the which is in some structures of amyloid M.R. Hughes M.P. Rodriguez J.A. R. Eisenberg D.S. The amyloid and 2021; Full Text Full Text PDF PubMed Scopus Google Scholar). to mutations in FUS, which are associated with ALS and in our of amyloidogenic mutations, mutations are the common mutations to be amyloidogenic in our of (Fig. is thought of as a up because of of their and by their which forms a with the of the K.A. but not propensity of is by Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). the to form a the also the formation of amyloid by the for which is a to the of the X. Lin Y. Kato M. Mori E. Liszczak G. Sutherland L. Sysoev V.O. Murray D.T. Tycko R. McKnight S.L. Transiently structured head domains control intermediate filament assembly.Proc. Natl. Acad. Sci. U. S. A. 2021; 118e2022121118Crossref Scopus (10) Google Scholar). missense mutations, can into and (Fig. The common mutation to be amyloidogenic is from to is a with a of in the of an amyloid to increase and is to in pathogenic amyloids M.R. Hughes M.P. Rodriguez J.A. R. Eisenberg D.S. The amyloid and 2021; Full Text Full Text PDF PubMed Scopus Google Scholar). This a mutation to to the formation of an amyloid of the TFG mutations that amyloid propensity is a to mutation is to the that the of common mutations to be amyloidogenic may be sequence a of which are amyloidogenic in other the amino acid diversity in LCDs may which mutations increase the of a within their sequence This to mutations more common in the to mutations, the latter of which have propensity is is missense mutations for a of amino acid and some are more because of in and (Fig. This to mutations are the common of mutation the mutations to be but are to mutations, to mutations are with a also more that to mutations are commonly to be amyloidogenic are for to in that mutation with to mutations (Fig. The for the of mutations in this We to on their propensity to form a the of amyloid a based on the of a from the protein and generates an is to a different for but is the and the for all the segments in the is not the for amyloid but is for and is based sequence can be in their we N.C. for the of in 2013; PubMed Scopus Google Scholar) to amyloidogenic regions in TFG and its (Fig. up to different methods and their We and for a of an amyloid containing the mutation but not the The of is experimental amyloid structures and in that can form fibers (3Kim H.J. Kim N.C. Wang Y.-D. Scarborough E.A. Moore J. Diaz Z. MacLea K.S. Freibaum B. Li S. Molliex A. Kanagaraj A.P. Carter R. Boylan K.B. Wojtas A.M. Rademakers R. et al.Mutations in prion-like domains in hnRNPA2B1 and hnRNPA1 cause multisystem proteinopathy and ALS.Nature. 2013; 495: 467-473Crossref PubMed Scopus (934) Google Scholar, L. R. Eisenberg D. Identifying the proteins of Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar, B. Sawaya M.R. G. M. Eisenberg D.S. of two segments in domains of in Biol. Chem. 2018; Full Text Full Text PDF PubMed Scopus Google Scholar, Boyer D.R. Murray K.A. M. Sawaya M.R. G. Cascio D. B. Eisenberg D.S. prion-like by Biol. Chem. Full Text Full Text PDF PubMed Scopus Google Scholar). are some to many amyloid fibers contain their which can to of Also, does not the sequence of segments that may form fibers in may not be to because of in the of the protein or some other from which can to of These that our of identification the to amyloids that have been by other methods and also amyloid Our to identify mutations that amyloidogenic of the of mutations to be many that have been previously to amyloid as as many mutations that have structural Also, many mutations in proteins known to be to form but the mutations have on their is also to that the mutation the to cause the protein to form an this does not that the protein form an amyloid in disease. This to the and of our These also into the protein we to our TFG previously been shown to aggregate in both and from other proteins in our have also been to aggregate but have not been shown to be and J. M. D. E. K. A. O. J.A. et mutations in with and functional characterization of PubMed Scopus Google Scholar, E. A. S. M. P. P. J. E. L. of amyotrophic lateral Genet. 2020; PubMed Scopus Google Scholar). We are the of these proteins with to amyloid formation. Some other proteins from our of amyloid mutations proteins that have the molecular function protein This not all the known amyloid proteins from the with TFG and but also other proteins such as which is involved in some forms of ALS Chen N. Y. Shi Y. H. H. M. E. S. et al.Mutations in cause and ALS and PubMed Scopus Google Scholar), and in which mutations can lead to because of of the protein U. J. S. S. D. mutations in the and domains of the protein 28: PubMed Scopus Google Scholar). The LCD of TFG to form amyloid fibers containing mutations, in with our (Fig. and the sequence of the protein is to form amyloid fibers with the mutant a (Fig. This is not to our the show propensity in both but the is and the many segments to be to form a and the and of these segments drive its amyloid formation. may have been in the protein and the LCD especially the LCD to a of and the protein is may have been more prone to fibers the LCD in the the sequence of the LCD may have been to form fibers more or in different is the of the domain in the which functions as a of for TFG A. N. M. Wang L. A. TFG and J. 2015; PubMed Scopus Google Scholar). This not present in the may have of the which formation. of the cause of the of both the protein and the LCD to with our this we documented mutations with to amyloidogenic This by the identification of known amyloidogenic mutations as as the formation of amyloid fibers from with mutations not previously as amyloidogenic. Our many unidentified amyloid proteins that to be acid in the human proteome for low complexity with complexity and complexity sequence to be a low-complexity as low complexity with TFG for the and forms a a previously M. Han T.W. Xie S. Shi K. Du X. Wu L.C. Mirzaei H. Goldsmith E.J. Longgood J. Pei J. Grishin N.V. Frantz D.E. Schneider J.W. Chen S. Li L. et al.Cell-free formation of RNA granules: Low complexity sequence domains form dynamic fibers within hydrogels.Cell. 2012; 149: 753-767Abstract Full Text Full Text PDF PubMed Scopus (1214) Google Scholar). protein in to an of to and with by and the by acid by and into TFG a with a but and into containing and TFG to in containing and to a and by the on of The of into a and and mutant TFG LCD to in containing to a of in of to to in a for with and This with with TFG with the the with containing and of and mutant TFG from in aggregation on and for with in for by and for with a of TFG for and with two a the two to and the with the aggregate on an and with for with the on a The that the of this are from the D. S. This D. S. E. is and in the other that they have of with the of this We for with This by the of for of and and and G. M. K. A. and D. S. E. G. M. R. and K. A. M. L. S. and M. P. H. G. M. R. G. M. R. K. A. L. M. P. and R. A. G. M. R. K. A. M. and D. S. E. G. M. R. D. S. E. D. S. E. The is solely the of the and does not the of the

Topics & Concepts