Selective autophagic clearance of protein aggregates is mediated by the autophagy receptor, TAX1BP1

Misfolded protein aggregates can disrupt cellular homeostasis and cause toxicity, a hallmark of numerous neurodegenerative diseases. Protein quality control by the ubiquitin proteasome system (UPS) and autophagy is vital for clearance of aggregates and maintenance of cellular homeostasis1. Autophagy receptor proteins bridge the interaction between ubiquitinated proteins and the autophagy machinery allowing selective elimination of cargo2. Aggrephagy is critical to protein quality control, but how aggregates are recognized and targeted for degradation is not well understood. Here we examine the requirements for 5 autophagy receptor proteins: OPTN, NBR1, p62, NDP52, and TAX1BP1 in proteotoxic stress-induced aggregate clearance. Endogenous TAX1BP1 is both recruited to and required for the clearance of stress-induced aggregates while overexpression of TAX1BP1 increases aggregate clearance through autophagy. Furthermore, TAX1BP1 depletion sensitizes cells to proteotoxic stress and Huntington’s disease-linked polyQ proteins, whereas TAX1BP1 overexpression clears cells of polyQ protein aggregates by autophagy. We propose a broad role for TAX1BP1 in the clearance of cytotoxic proteins, thus identifying a new mode of clearance of protein inclusions.

proteins, thus identifying a new mode of clearance of protein inclusions. 23

Main 24
Maintenance of cellular and organismal health is intricately connected to protein quality 25 control. A balance exists between protein translation, folding, and degradation that maintains the 26 stoichiometry and function of cellular protein complexes and organelles. If this balance is 27 perturbed, the accumulation of misfolded proteins can be toxic to the cell and is associated with 28 disruption of cellular function and numerous neurodegenerative diseases such as amyotrophic 29 lateral sclerosis (ALS), Huntington's disease (HD), and Alzheimer's disease (AD) 3 . A number of 30 protein quality control pathways exist within the cell to forestall dysfunction. Molecular 31 chaperone systems function to monitor and refold proteins if possible, while misfolded or 32 damaged proteins are targeted for elimination via autophagy or the UPS 4,5 . The UPS is generally 33 responsible for routine turnover of short-lived proteins and targeted degradation of soluble or 34 solubilized misfolded proteins while macroautophagy, a catabolic process, culminates in the 35 lysosomal degradation of long-lived proteins, damaged organelles, and portions of the 36 cytoplasm 6-9 . Since the proteasome can only accommodate unfolded polypeptide chains, it is 37 generally thought that autophagy is responsible for the removal of insoluble protein 38 aggregates 1,10 . If either autophagy or the UPS is hindered, acute or chronic proteotoxic stress, 39 such as that caused by the expression of mutated proteins in neurodegenerative disease, can 40 result in the selective accumulation of these aggregation-prone proteins [11][12][13] . In vivo, inhibition of 41 autophagy results in intracellular protein aggregation contributing to neuronal cell death and 42 neurodegeneration in mice 14,15 . 43 In recent years, numerous studies have highlighted the ability of autophagy to selectively 44 eliminate specific substrates, including mature ribosomes, endoplasmic reticulum, intracellular 45 pathogens, mitochondria, and protein aggregates [16][17][18][19] . In these cases, the autophagic machinery 46 employed in nonselective bulk degradation of cytosolic material is targeted to specific cargo. 47 Stimulation of autophagy is a promising therapeutic strategy in the treatment of protein 48 aggregation diseases and has been shown to enhance turnover of aggregated proteins, such as 49 TDP-43, in neuronal ALS models and huntingtin protein in Huntington disease models [20][21][22] . Key 50 to this approach is an understanding of how autophagy is selectively targeted to aggregates. In 51 addition to the specificity mediated by E3 ubiquitin (UB) ligases that target their cognate 52 substrates, there is selectivity in delivery to the proteasome via UB receptors 23 and in the 53 recruitment of autophagic machinery via autophagy receptors which associate with 54 polyubiquitylated proteins, thus linking substrates to the appropriate degradation machinery 16 . 55 9 TAX1BP1 (Fig. 4Bvii, 4D). This is consistent with a prior report that only the ZF2 of TAX1BP1 173 is capable of binding UB 25 . An N-and C-terminally truncated mutant (ΔSKICH/ΔZF) did not 174 localize to aggregates but instead formed aberrant clusters in the cytosol and was inferior to the 175 ZF mutants in its ability to rescue clearance (Fig. 4Bx, 4D, Supplementary Fig. 4A). 176 Very similar results were observed in triple knockout (TKO: TAX1BP1 KO, OPTN KO,177 NDP52 KO) cells ( Supplementary Fig. 4B, 4C), ruling out potential redundancy that may mask 178 requirements for TAX1BP1 domains through functionally related autophagy receptors. Low, 179 near-endogenous expression of TAX1BP1 rescued aggregate clearance to that of WT cells while 180 higher expression increased clearance compared to WT cells (Fig. 4E,4F,Supplementary Fig. 181 4C). In contrast, neither the ΔSKICH, ΔZF nor ΔSKICH/ΔZF mutants were able to fully rescue 182 aggregate clearance demonstrating the need for these protein-interaction and UB-binding 183 domains (Fig. 4F, Supplementary Fig. 4C). Thus, full length TAX1BP1 was able to rescue 184 aggregate clearance in the absence of NDP52 and OPTN, indicating that TAX1BP1 promotes 185 aggrephagy (Fig. 4F). 186 We also examined whether TAX1BP1 overexpression in WT cells beyond the level of 187 endogenous TAX1BP1 expression could promote aggregate clearance, thus addressing the 188 therapeutic potential of TAX1BP1. TAX1BP1 protein expressed in WT cells behaved similarly 189 to the endogenous protein, remaining diffuse in untreated cells and colocalizing with UB-stained 190 foci upon puromycin treatment (Fig. 4G, Supplementary Fig. 4D). Aggregate formation upon 2h 191 puromycin treatment was decreased in both low (~20% decrease) and high (~30% decrease) -192 overexpressing TAX1BP1 cells (Fig. 4H). Clearance of aggregates in WT cells with low 193 overexpression of TAX1BP1 was similar to that in WT cells expressing only endogenous 194 TAX1BP1 (Fig. 4I, Supplementary Fig. 4D), whereas, higher levels of TAX1BP1 expression 195 increased clearance upon puromycin washout -only ~10% of cells retained UB-positive foci in 196 contrast to more than 30% of WT cells (Fig. 4G, 4I). 197

Overexpression of WT TDP-43 and polyQ huntingtin fragments is cytotoxic in in 198
vitro and in vivo model system 20,37-44 . Stimulation of autophagy increases clearance of WT 199 and mutant TDP-43, huntingtin fragments and attenuates cytotoxicity 20,43,[45][46][47][48] . Therefore, WT, 200 TAX1BP1 KO, and high-expressing TAX1BP1-rescue cells were exposed to a variety of 201 stresses, including low dose proteasome inhibition and expression of aggregation-prone proteins, 202 such as the ALS-associated, EGFP-TDP-43, and varied length model substrates carrying 203 expanded glutamine tracts (polyQ) expressed from exon 1 of the huntingtin-encoding gene 204 ( Supplementary Fig. 5A, 5B). Cells were either infected or transfected on day 1 and assessed 205 daily for six days. Loss of TAX1BP1 decreased viability compared to WT cells upon exposure to 206 each form of proteotoxic stress (Fig. 5A, Supplementary Fig. 5C). Expression of TAX1BP1 in 207 the knockout line was able to partially restore cell viability (Fig. 5A, Supplementary Fig. 5C). 208 To determine whether TAX1BP1 can aid in clearance of Htt-polyQ proteins, we infected and therefore is a useful control when comparing to the aggregate-forming HttQ74-EGFP and 213 HttQ103-EGFP proteins. HttQ23-EGFP expression was robust and as expected the protein 214 remained largely diffuse and cytosolic in both WT and TAX1BP1 KO cells (Fig. 5B). In 215 contrast, TAX1BP1-deficient cells exhibited increased focal aggregates of both Htt74Q-EGFP 216 and Htt103Q-EGFP compared to WT cells (Fig. 5B, 5C, 5D). Rescue of the TAX1BP1 KO line 217 with expression of FLAG-TAX1BP1 restored clearance of HttQ74-EGFP and Htt103Q-EGFP to 218 that of WT cells, indicating that TAX1BP1 is highly effective in directing clearance of 219 huntingtin polyQ proteins (Fig. 5B, 5C, 5D). Endogenous TAX1BP1 also colocalized with 220 HttQ103-EGFP aggregates in WT cells (Fig. 5E) appearing to enclose aggregates in many 221 instances (Fig. 5F, Supplementary Figure 5D). 222 Because TAX1BP1 overexpression can promote clearance of aggregates beyond that in 223 WT cells and because we found TAX1BP1 highly and specifically-expressed in brain ( Figure  224 2D, 2F), we examined whether TAX1BP1 could provide a protective effect in iPSC-derived 225 neurons exposed to huntingtin proteins. iPSC-derived neurons with or without stable expression 226 of TAX1BP1 were infected with virus expressing either the non-aggregating control construct, 227 HttQ23-EGFP, ( Figure 5G) or the aggregate-forming HttQ103-EGFP ( Figure 5H) and imaged 228 daily for 15 days to assess viability via nuclei count (NLS-BFP). The rates of cell death, assessed 229 by comparing slopes obtained from lines fitted to cell number over time, did not differ between 230 WT and TAX1BP1-overexpressing neurons infected with the non-aggregate-forming HttQ23-231 EGFP ( Figure 5G). However, TAX1BP1 overexpression significantly improved survival in 232 neurons exposed to the aggregate-forming HttQ103-GFP protein ( Figure 5H), consistent with our 233 observations in HeLa cells. 234 Here we report a broad role for TAX1BP1 in protein homeostasis. Our results 235 demonstrate that loss of TAX1BP1 leads to decreased ability to target insoluble protein for 236 degradation. Overexpression of TAX1BP1 further promotes aggregate clearance and rescues cell 237 viability upon exposure to varied proteotoxic insults, including translation stress, proteasome 238 inhibition, and exposure to aggregate-prone proteins such as TDP-43 and huntingtin-expanded 239 polyQ model substrates (Supplementary Figure 5E). 240 Aggrephagy has potential to mitigate neurodegenerative proteinopathies. Focus on the 241 proteins that provide specificity in targeting aggregates, such as TAX1BP1, may be valuable, as 242 these are likely deciding factors in directing the autophagy response. Using single or 243 combinatorial knockouts of OPTN, NDP52, TAX1BP1, NBR1, and p62, we observed distinct 244 roles for these varied autophagy receptor proteins in aggrephagy. Though they may function en 245 masse to maintain protein homeostasis, the individual proteins exhibit functional and spatial 246 distinctions at the subcellular and tissue level. For example, autophagy receptor proteins are 247 recruited independently to distinct microdomains surrounding bacteria and mitochondria where 248 they perform nonredundant roles during xenophagy and mitophagy, respectively 26,49 . 249 Furthermore, TAX1BP1 is robustly expressed in human brain lysate as well as primary rat 250 cortical neurons, distinguishing it from NDP52. TAX1BP1 is also associated with the insoluble 251 protein fraction in primary rat cortical neurons exposed to proteotoxic stress. Amongst the other 252 autophagy receptor proteins, only p62 showed similar association with the insoluble fraction. 253 TAX1BP1 and its paralog, NDP52, share similar domain structures; however, recent studies 254 suggest structural differences in the organization of the SKICH domains as well as distinct 255 ATG8 binding affinities 25,50 . TAX1BP1 is also more broadly conserved than NDP52 among 256 mammals 25 , suggesting that TAX1BP1 fills nonredundant essential roles. Future studies 257 examining TAX1BP1 expression levels during aging and in animal models of neurodegenerative 258 disease are warranted. 259 TAX1BP1 recruitment to protein aggregates requires the C-terminal ubiquitin-binding 260 domain and its function in promoting aggrephagy further necessitates the N-terminal SKICH 261 domain. However, none of our TAX1BP1 mutants were completely dead in terms of rescue 262 effect, suggesting that other cellular functions of TAX1BP1 may be involved. One such role is to 263 act as a Myosin VI cargo adaptor protein for mediating autophagosome maturation, which could 264 contribute to clearance dependent upon other upstream factors, thus explaining partial rescue 25 . 265 Additionally, TAX1BP1 is best studied for its role in negatively regulating nuclear factor-kb 266 (NF-κB) and interferon regulatory factor (IRF) 3 via an interaction with the deubiquitinase, A20, 267 13 thus restricting pro-inflammatory signaling and immune response 51,52 . Though links between 268 TAX1BP1's role as an autophagy adaptor and the immune response are yet to be understood, it 269 is well known that protein aggregation, inflammation, and autophagy are intertwined 53 . Studies 270 have shown that exposure to disease-associated protein aggregates can elicit innate immune 271 response in glial cells and that LPS-induced inflammation results in enhanced aggregate 272 formation in disease models suggesting a synergistic relationship between proinflammatory 273 response, proteostasis and neurodegeneration 54,55 . Autophagy may function to downregulate 274 inflammatory signaling and TAX1BP1 may be an important link between detection and 275 monitoring of cellular protein aggregates and the inflammatory response. 276 Notably, ectopic expression of TAX1BP1 in knockout cells was able to rescue aggregate 277 clearance to WT levels and increased overexpression of TAX1BP1 in both the knockout and WT 278 cells was able to reduce aggregate levels below those observed in WT cells. Furthermore, 279 TAX1BP1 overexpression in IPSC-derived neurons was protective against huntingtin aggregate-280 induced toxicity. TAX1BP1 thus plays a general role in promoting aggrephagy and future studies 281 aimed at increasing TAX1BP1 expression or stability in vivo present promising therapeutic 282 potential in addressing proteinopathies. An increased understanding of targeting specificity of 283 selective autophagy receptor proteins for protein aggregates may make autophagy-stimulating 284 approaches more specific and effective in treatment of protein misfolding diseases. 285 286

Acknowledgements 287
We thank Drs. Alicia Pickrell, Malavika Raman, Achim Werner, David Beck and Richa Lomash 288 for critical reading of the manuscript. We also thank the NINDS Light Imaging Facility, the 289 NICHD Imaging Facility, and the NHLBI Flow Cytometry Core Facility. We thank Dr. Harm

Competing Interests 302
The authors declare no competing financial interests.

Cell Culture and reagents 305
HeLa and HEK293T were cultured in DMEM (Life Technologies) supplemented with 306 10% (v/v) FBS (Gemini Bio Products), 10 mM HEPES (Life Technologies), 1 mM sodium 307 pyruvate (Life Technologies), 1mM non-essential amino acids (Life Technologies) and 2 mM 308 glutamine (Life Technologies). HeLa cells were acquired from the ATCC and authenticated at 309 the Johns Hopkins GRCF Fragment Analysis Facility using STR profiling. Testing for 310 mycoplasma contamination was performed bimonthly using the PlasmoTest kit (InvivoGen). 311 Plasmids were transfected using either X-tremeGENE 9 (Roche), polyethylenimine 312 For acute treatments, HeLa cells were grown on poly-D-lysine-coated (Sigma-Aldrich, 373 P7280) coverslips and treated with either 5 μg/ml of puromycin (Invivogen) for 2 h or 1 μM 374 MG132 (Sigma-Aldrich) for 8 h at 37°C. Cells were either fixed for imaging or assessed for 375 clearance. For clearance, cells were washed three times in DMEM-10% FBS and released into 376 drug-free medium for 3 h at 37°C. For long treatments, HeLa cells were grown on poly-D-lysine-377 coated coverslips and treated with 1 μg/ml of puromycin or 1 μM MG132 for 18 h at 37°C. 378 Coverslips were fixed and processed for microscopy, as outlined above. 379 For huntingtin poly-Q-protein clearance assays, HeLa cells were seeded on poly-D-380 lysine-coated coverslips and transduced the following day with lentivirus expressing either 381 HttQ23-EGFP, HttQ74-EGFP or HttQ103-EGFP for 24 hours with 8 µg/ml polybrene. Cells   permutation test (a.k.a randomization test) was used. In brief, the delta mean score of the groups 498 was compared to random delta mean scores of shuffled groups iterated 10,000 times. P value 499 was determined by calculating the number of times the delta score was higher in the shuffled 500 group than in the ground true group 58 . The number of observations used in each experimental 501 series is included in Table 2. Code is provided in Supplementary Information.  502 503

Data Availability 504
The datasets generated for all microscopy cell counting experiments are available as 505 supplementary files and noted in the associated figure legends. The datasets generated and 506 analyzed to assess neuron viability are available from the corresponding author on request. 507 Associated code is available as supplementary files. Figures 1D, 1F, 1J that the SKICH domains from different protein families share a common Ig-like fold but 635 harbour substantial structural variations. J. Biomol. Struct. Dyn. 33, 1385-1398(2015. test. c, d, Stably-expressing TAX1BP1 rescue lines were created using either N-FLAG or C-698 FLAG tag at high (H) or low (L) expression levels (See Supplementary Figure 3A, 3B) and 699 exposed to 5 µg/ml puromycin for 2 h, after which cells were either fixed for imaging (c) or 700 washed and followed for a further 3 h in full media (d) and quantified as in (b). e, WT, 701 TAX1BP1 KO, or TAX1BP1 KO + FLAG-TAX1BP1 (H) cell lines were exposed to 5 µg/ml 702 puromycin in the presence or absence of 100 nM Bafilomycin A, after which cells were either 703 fixed for imaging or washed and followed for a further 3 h in full media or in full media 704 containing Bafilomycin A. Larger fields of view shown in Supplementary Figure 3C expression of TAX1BP1 mutants were exposed to 5 µg/ml puromycin for 2 h, after which cells 714 were either fixed for imaging or washed and followed for a further 3 h in full media; scale bar 715 Figure 5. TAX1BP1 mediates aggrephagy of cytotoxic aggregation-prone proteins. a, WT, 733 TAX1BP1 knockout, or TAX1BP1 knockout with stably-expressed TAX1BP1 rescue cell lines 734 were exposed to proteotoxic stressors as indicated on Day 1, then followed for 6 days during 735 which viability was measured by quantification of ATP production. Relative viability represents 736 normalized luminescence displayed as mean ± s.d. from 3 independent experiments; significance 737 was assessed using two-way ANOVA test (****P<0.0001, ***P<0.001, **P<0.01, *P<0.05) 738 with Tukey's post hoc test. P values and normalized viability measurements shown on graphs are 739 for day 6 comparisons. The individual measurements for each time point and conditions used in 740 each experimental series and P values for all comparisons are included in Table 1 rescue experiments: L = low expression, H = high expression. b, Cell lines in (a) were exposed 790 to 5 µg/ml puromycin for 2 h, after which cells were either fixed for imaging or washed and 791 followed for a further 3 h in full media; scale bar 10µm. c, Full field of view images associated 792 with Figure 3E, F showing WT, TAX1BP1 KO, or TAX1BP1 KO + FLAG-TAX1BP1 (H) cell 793 lines exposed to 5 µg/ml puromycin in the presence or absence of 100 nM Bafilomycin A, after 794 which cells were either fixed for imaging or washed and followed for a further 3 h in full media 795 or in media containing Bafilomycin A. d, WT cells exposed to 5 µg/ml puromycin for 2 h in the 796 presence or absence of Bafilomycin A, after which cells were either fixed for imaging or washed 797 and followed for a further 3 h in full media with or without Bafilomycin A; scale bar 10µm. All 798 images are representative of at least 3 independent experiments. 799 800 Supplementary Figure 4. a, Full field of view images of all TAX1BP1 stable mutant expression 801 cell lines exposed to 5 µg/ml puromycin for 2 h, after which cells were either fixed for imaging 802 or washed and followed for a further 3 h in full media. Associated with Figure 4B, C, D. b, 803 Validation of knockout cell lines. c, TKO (triple knockout: TAX1BP1, OPTN, NDP52) cell line 804 with stable expression of TAX1BP1 mutants exposed to 5 µg/ml puromycin after which cells 805 were either fixed for imaging or washed and followed for a further 3 h in full media; scale bar 806 10µm. d, WT cells stably expressing low levels of FLAG-TAX1BP1 were exposed to 5 µg/ml 807 1 ---2 title: ""Selective autophagic clearance of protein aggregates is 3 mediated by the autophagy receptor, TAX1BP1" " 4 output: html_notebook 5 ---6 Images of the size 1200*1200 were segmented using the following 7 segmentation code. as.data.frame(matrix(0,nrow(xy),5)) colnames(df.combine)<-35 c("x","y","Area_real","Areal_roundess","ratio") df.combine$x<-