Comparative Interactome Analysis of α-arrestin Families in Human and Drosophila

Reviewer #1 (Public Review):

The study provides a complete comparative interactome analysis of α-arrestin in both humans and drosophila. The authors have presented interactomes of six humans and twelve Drosophila α-arrestins using affinity purification/mass spectrometry (AP/MS). The constructed interactomes helped to find α-arrestins binding partners through common protein motifs. The authors have used bioinformatic tools and experimental data in human cells to identify the roles of TXNIP and ARRDC5: TXNIP-HADC2 interaction and ARRDC5-V-type ATPase interaction. The study reveals the PPI network for α-arrestins and examines the functions of α-arrestins in both humans and Drosophila. The authors have carried out the necessary changes that were suggested, and the manuscript can now be accepted.

Comments: I would like to congratulate the authors and the corresponding authors of this manuscript for bringing together such an elaborate study on α-arrestin and conducting a comparative study in drosophila and humans.

Introduction: The introduction provides a rationale behind why the comparison between humans and Drosophila is performed.

Results: The results cover all the necessary points concluded from the experiments and computational analysis. The images are elaborate and well-made. The authors have a rigorous amount of work added together for the success of this manuscript. The authors have provided a database of network of α-arrestins in both humans and Drosophila which can be used by other researchers working in the same subject to study the interacting genes.

Discussion: the authors have utilized and discussed the conclusion they draw from their study. But could highlight more on ARRDCs and why it was selected out of the other arrestins.

References: the authors have considered the suggestion and added the necessary references.

The authors have provided future work directions associated with their work.

Reviewer #2 (Public Review):

In this manuscript, the authors present a novel interactome focused on human and fly alpha-arrestin family proteins and demonstrate its application in understanding the functions of these proteins. Initially, the authors employed AP/MS analysis, a popular method for mapping protein-protein interactions (PPIs) by isolating protein complexes. Through rigorous statistical and manual quality control procedures, they established two robust interactomes, consisting of 6 baits and 307 prey proteins for humans, and 12 baits and 467 prey proteins for flies. To gain insights into the gene function, the authors investigated the interactors of alpha-arrestin proteins through various functional analyses, such as gene set enrichment. Furthermore, by comparing the interactors between humans and flies, the authors described both conserved and species-specific functions of the alpha-arrestin proteins. To validate their findings, the authors performed several experimental validations for TXNIP and ARRDC5 using ATAC-seq, siRNA knockdown, and tissue staining assays. The experimental results strongly support the predicted functions of the alpha-arrestin proteins and underscore their importance.

Reviewer #3 (Public Review):

Lee, Kyungtae and colleagues have discovered and mapped out alpha-arrestin interactomes in both human and Drosophila through the affinity purification/mass spectrometry and the SAINTexpress method. Their work revealed highly confident interactomes, consisting of 390 protein-protein interactions (PPIs) between six human alpha-arrestins and 307 preproteins, as well as 740 PPIs between twelve Drosophila alpha-arrestins and 467 prey proteins.

To define and characterize these identified alpha-arrestin interactomes, the team employed a variety of widely recognized bioinformatics tools. These analyses included protein domain enrichment analysis, PANTHER for protein class enrichment, DAVID for subcellular localization analysis, COMPLEAT for the identification of functional complexes, and DIOPT to identify evolutionary conserved interactomes. Through these assessments, they not only confirmed the roles and associated functions of known alpha-arrestin interactors, such as ubiquitin ligase and protease, but also unearthed unexpected biological functions in the newly discovered interactomes. These included involvement in RNA splicing and helicase, GTPase-activating proteins, and ATP synthase.

The authors carried out further study into the role of human TXNIP in transcription and epigenetic regulation, as well as the role of ARRDC5 in osteoclast differentiation. It is particularly commendable that the authors conducted comprehensive testing of TXNIP's role in HDAC2 in gene expression and provided a compelling model while revised the manuscript. Additionally, the quantification of the immunocytochemistry data presented in Figure 6 convincingly supports the authors' hypothesis.

Overall, this study holds important value, as the newly identified alpha-arrestin interactomes are likely aiding functional studies of this protein group and advance alpha-arrestin research.

Author Response

The following is the authors’ response to the original reviews.

Reviewer #1

The study provides a complete comparative interactome analysis of α-arrestin in both humans and drosophila. The authors have presented interactomes of six humans and twelve Drosophila α-arrestins using affinity purification/mass spectrometry (AP/MS). The constructed interactomes helped to find α-arrestins binding partners through common protein motifs. The authors have used bioinformatic tools and experimental data in human cells to identify the roles of TXNIP and ARRDC5: TXNIP-HADC2 interaction and ARRDC5-V-type ATPase interaction. The study reveals the PPI network for α-arrestins and examines the functions of α-arrestins in both humans and Drosophila.

Comments

I will like to congratulate the authors and the corresponding authors of this manuscript for bringing together such an elaborate study on α-arrestin and conducting a comparative study in drosophila and humans.

Introduction:

The introduction provides a rationale behind why the comparison between humans and Drosophila is carried out.

• Even though this is a research manuscript, including existing literature on similar comparison of α-arrestin from other articles will invite a wide readership.

Results:

The results cover all the necessary points concluded from the experiments and computational analysis.

1. The authors could point out the similarity of the α-arrestin in both humans and Drosophila. While comparing α-arrestin in both humans and Drosophila If percentage homology between α-arrestin of both Drosophila and humans needs to be calculated.

Thank you for your insightful feedback. As suggested by reviewer, we determined percentage homology of α-arrestin protein sequences from human and Drosophila using Clustal Omega. This homology is now illustrated as a heatmap in revised Figure S5. Please note that only the values with percentage homology of 40% or higher are selectively labeled.

• Citing the direct connecting genes from the network in the text will invite citations and a wider readership.

Figures:

The images are elaborate and well-made.

1. The authors could use a direct connected gene-gene network that pointing interactions. This can be used by other readers working on the same topic and ensure reproducibility and citations.

We appreciate your valuable comment. Based on the reviewer’s suggestion, we have developed a new website in which one can navigate the gene-gene networks of α-arrestins. These direct connected gene-gene networks are housed in the network data exchange (NDEx) project. Additionally, we have included gene ontology and protein class details for α-arrestins’ interactors in these set of networks, offering a more comprehensive view of α-arrestins’ interactomes.

On page 24 lines 15-18, we have revised the manuscript to introduce the newly developed website, as follows.

“Lastly, to assist the research community, we have made comprehensive α-arrestin interactome maps on our website (big.hanyang.ac.kr/alphaArrestin_PPIN). Researchers can search and download their interactomes of interest as well as access information on potential cellular functions and protein class associated with these interactomes.”

3-1) The co-expression interactions represented as figures should reveal interaction among the α-arrestin and other genes. Which are the sub-network genes does the α- arrestin interact to/ with from the sub-network? The arrows are only pointing at the sub-networks. The figures do not reveal their interaction. Kindly reveal the interaction in the figure with the proper nodes in the figure.

3-2) Figure 2: the network attached in both human and drosophila is well represented. The green lines from α-arrestin indicate the strength of the interaction. Several smaller expression networks are seen. But "α-arrestin" in both organisms seems highly disconnected from all the genes. Connected genes have edges, not arrows. If α-arrestin can be shown connected to these gene-gene networks will help in identifying which genes connect with which gene through α-arrestin. This can be used by other readers working on the same topic and ensure reproducibility and citations.

Thank you for your valuable comment. In response to the reviewer’s recommendation, we’ve added supplementary figure, Figure S4, which illustrates direct interaction between α-arrestin and protein components of clustered complexes (or sub-networks) in addition to the associations shown between α-arrestins and the clustered complexes in Figure 2. We believe that this newly incorporated information regarding direct protein interactions will invite citations and wider readership as the reviewer pointed out.

On page 12 line 27 to page 13 line 5, we have revised the manuscript to cite the direction interactions between ARRDC3 and proteins involved in ubiquitination-dependent proteolysis, as follows.

“While the association of ARRDC3 with these ubiquitination-dependent proteolysis complexes is statistically insignificant, ARRDC3 does interact with individual components of these complexes such as NEDD4, NEDD4L, WWP1, and ITCH (Figure S4A). This suggest their functional relevance in this context, as previously reported in both literatures and databases (Nabhan et al., 2010; Shea et al., 2012; Szklarczyk et al., 2015; Warde-Farley et al., 2010) (Puca & Brou, 2014; Xiao et al., 2018).”

Direct interaction between α-arrestins and protein components of clustered complexes are illustrated in the newly added figure, Figure S4.

4-1) Figure 4. The Protein blot image was blurred. Kindly provide a higher-resolution image.

4-2) Figure 5. B. - The authors can provide images with higher resolution blot images. The bands were not visible.

We appreciate for valuable comment. Unfortunately, the protein blot image was scanned from the original film and the images we provided in the figure represent the highest resolution that we have obtained to date. Raw, uncropped images are shown in Author response image 1 and 2.

Author response image 1.

Raw image of Figure 4B

Author response image 2.

Raw image of Figure 5B

1. Figure: 5. A. - I see non-specific amplifications in the gel images. Are these blotting images? or the gel images that were changed to "Grayscale"? Non-specific amplification may imply that the experiment was not repeated and standardized. Was it gel images or blot images?

We appreciate your insightful comment. The images in Figure 5A represent western blot bands from co-immunoprecipitation assay for analysis of the interaction between TXNIP and HDAC2 proteins. Since immunoblotting using immunoprecipitates can usually detect some non-specific bands from heavy (~ 50 kDa) and light (~25 kDa) chains of the target antibody or from multiple co-immunoprecipitated proteins, we assume that the vague non-specific bands in Figure 5A might be a heavy chain of TXNIP or HDAC2 antibody or an unclear non-specific band. Because target bands showed strong intensity and very clear pattern compared to the non-specific bands in the co-immunoprecipitation assay, we believe that this data is sufficient to support the interaction of TXNIP with HDAC2. Finally, In the revised Figure 5A, we’ve modified the labeling for different experimental conditions, namely siCon and siTXNIP treatments, and added expected size of proteins (kDa), as shown below.

1. Figure 5. A. RT-PCR analysis: What was your expected size of the amplifications? the ladder indicated is in KDa. Is that right?

We appreciate your insightful questions. As mentioned above, Figure 5A shows the blotting images of co-immunoprecipitation analysis, and the ladder indicates the molecular weight (kDa) of protein markers. For clearer interpretation, the expected size of target proteins has been added in Figure 5A in the revised manuscript.

1. How were the band intensities determined?

Thank you for your question. For quantification of immunoblot results, the densities of target protein bands were analyzed with Image J, as we described in the Materials and Methods.

Discussion:

The authors have utilized and discussed the conclusion they draw from their study. But could highlight more on ARRDCs and why it was selected out of the other arrestins. The authors have provided future work directions associated with their work.

1. Why were only ARRDCs presented amongst all the arrestin in the main part of the manuscript?

We’re grateful for your valuable feedback. The reason we focused on α-arrestins was that α-arrestins have been discovered relatively recently, especially when compared to more established visual/ β-arrestin proteins in the same arrestin family but the biological functions of many α-arrestins remain largely unexplored, with notable exceptions in the budding yeast model and a few α-arrestins in mammals and invertebrate species. Most importantly, comparative study highlighting the shared or unique features of α-arrestins is yet to be undertaken. To gain a more comprehensive understanding of these unexplored α-arrestins across multiple species, we’ve centered our research on the ARRDCs within the arrestin protein family.

On page 21 lines 8-17, we’ve edited the manuscript to emphasize the importance of a comparative study on α-arrestins, as detailed below.

“According to a phylogenetic analysis of arrestin family proteins, α-arrestins were shown to be ubiquitously conserved from yeast to human (Alvarez, 2008). However, compared to the more established visual/ β-arrestin proteins, α-arrestins have been discovered more recently and much of their molecular mechanisms and functions remain mostly unexplored except for budding yeast model (Zbieralski & Wawrzycka, 2022). Based on the high-confidence interactomes of α-arrestins from human and Drosophila, we identified conserved and specific functions of these α-arrestins. Furthermore, we uncovered molecular functions of newly discovered function of human specific α-arrestins, TXNIP and ARRDC5. We anticipate that the discovery made here will enhance current understanding of α-arrestins.”

1. The discussion could be elaborated more by utilizing the data.

We appreciate your insightful feedback. Based on the reviewer’s suggestion, we’ve enhanced the discussion in the manuscript to provide a clearer interpretation of our results. First, we’ve added description of conserved protein complexes significantly associated with α-arrestins, stated on page 22 lines 5-12 and lines 23-26.

Page 22 lines 5-12: “The integrative map of protein complexes also highlighted both conserved and unique relationships between α-arrestins and diverse functional protein complexes. For instance, protein complexes involved in ubiquitination-dependent proteolysis, proteasome, RNA splicing, and intracellular transport (motor proteins) were prevalently linked with α-arrestins in both human and Drosophila. To more precisely identify conserved PPIs associated with α-arrestins, we undertook ortholog predictions within the α-arrestins’ interactomes. This revealed 58 orthologous interaction groups that were observed to be conserved between human and Drosophila (Figure 3).”

Page 22 lines 23-26: “Additionally, interaction between α-arrestins and entities like motor proteins, small GTPase, ATP binding proteins, and endosomal trafficking components were identified to be conserved. Further validation of these interactions could unveil molecular mechanisms consistently associated with these cellular functions.”

Secondly, we’ve added description of role of ARRDC5 in osteoclast maturation, as stated on page 23 lines 22-24.

“Conversely, depletion of ARRDC5 reduces osteoclast maturation, underscoring the pivotal role of ARRDC5 in osteoclast development and function (Figure S9A and B).”

Lastly, we examined the association between α-arrestins’ interactomes and human diseases, incorporating our findings into the discussion. The newly introduced figure based on the result is Figure S10.

On page 24 lines 10-14, we’ve added discussion on Figure S10 as follows.

“We further explored association between α-arrestins’ interactomes and disease pathways (Figure S10). Notably, the interactomes of α-arrestins in human showed clear links to specific diseases. For instance, ARRDC5 is closely associated with disease resulting from viral infection and cardiovascular conditions. ARRDC2, ARRDC4, and TXNIP share common association with certain neurodegenerative diseases, while ARRDC1 is implicated in cancer.”

Supplementary figures:

The authors have a rigorous amount of work added together for the success of this manuscript.

1. The reference section needs editing before publication. Maybe the arrangement was disturbed during compiling.

Thank you for your valuable comment. Based on the reviewer’s suggestion, we have rearranged the reference section to enhance its clarity. Below are excerpts from the update reference section in the manuscript.

“Adenuga, D., & Rahman, I. (2010). Protein kinase CK2-mediated phosphorylation of HDAC2 regulates co-repressor formation, deacetylase activity and acetylation of HDAC2 by cigarette smoke and aldehydes. Arch Biochem Biophys, 498(1), 62-73. doi:10.1016/j.abb.2010.04.002

Adenuga, D., Yao, H., March, T. H., Seagrave, J., & Rahman, I. (2009). Histone Deacetylase 2 Is Phosphorylated, Ubiquitinated, and Degraded by Cigarette Smoke. American Journal of Respiratory Cell and Molecular Biology, 40(4), 464-473. doi:10.1165/rcmb.2008-0255OC

Akalin, A., Franke, V., Vlahovicek, K., Mason, C. E., & Schubeler, D. (2015). Genomation: a toolkit to summarize, annotate and visualize genomic intervals. Bioinformatics, 31(7), 1127-1129. doi:10.1093/bioinformatics/btu775

Alvarez, C. E. (2008). On the origins of arrestin and rhodopsin. BMC Evol Biol, 8, 222. doi:10.1186/1471-2148-8-222”

1. many important references were missing.

We appreciate and agree with the reviewer’s comment. In response to the reviewer’s recommendation, we’ve thoroughly reviewed the manuscript and below are sections of the manuscript where around 20 new references have been added.

On page 8 lines 12-14:

“Utilizing the known affinities between short linear motifs in α-arrestins and protein domains in interactomes(El-Gebali et al., 2019; UniProt Consortium, 2018) “

On page 8 lines 19-22:

“One of the most well-known short-linear motifs in α-arrestin is PPxY, which is reported to bind with high affinity to the WW domain found in various proteins, including ubiquitin ligases (Ingham, Gish, & Pawson, 2004; Macias et al., 1996; Sudol, Chen, Bougeret, Einbond, & Bork, 1995)”

On page 9 lines 3-6:

“Next, we conducted enrichment analyses of Pfam proteins domains (El-Gebali et al., 2019; Huang da, Sherman, & Lempicki, 2009b) among interactome of each α-arrestin to investigate known and novel protein domains commonly or specifically associated (Figure S3A; Table S5).”

On page 9 lines 7-10:

“HECT and C2 domains are well known to be embedded in the E3 ubiquitin ligases such as NEDD4, HECW2, and ITCH along with WW domains (Ingham et al., 2004; Melino et al., 2008; Rotin & Kumar, 2009; Scheffner, Nuber, & Huibregtse, 1995; Weber, Polo, & Maspero, 2019)”

On page 10 lines 12-16:

“In fact, the known binding partners, NEDD4, WWP2, WWP1, and ITCH in human and CG42797, Su(dx), Nedd4, Yki, Smurf, and HERC2 in Drosophila, that were detected in our data are related to ubiquitin ligases and protein degradation (C. Chen & Matesic, 2007; Ingham et al., 2004; Y. Kwon et al., 2013; Marin, 2010; Melino et al., 2008; Rotin & Kumar, 2009) (Figure 1E; Figure S2F).”

On page 13 lines 20-21:

“Given that α-arrestins are widely conserved in metazoans (Alvarez, 2008; DeWire, Ahn, Lefkowitz, & Shenoy, 2007), “

On page 14 lines 12-17:

“The most prominent functional modules shared across both species were the ubiquitin-dependent proteolysis, endosomal trafficking, and small GTPase binding modules, which are in agreement with the well-described functions of α-arrestins in membrane receptor degradation through ubiquitination and vesicle trafficking (Dores et al., 2015; S. O. Han et al., 2013; Y. Kwon et al., 2013; Nabhan et al., 2012; Puca & Brou, 2014; Puca et al., 2013; Shea et al., 2012; Xiao et al., 2018; Zbieralski & Wawrzycka, 2022) (Figure 3).”

Reviewer #2

In this manuscript, the authors present a novel interactome focused on human and fly alpha-arrestin family proteins and demonstrate its application in understanding the functions of these proteins. Initially, the authors employed AP/MS analysis, a popular method for mapping protein-protein interactions (PPIs) by isolating protein complexes. Through rigorous statistical and manual quality control procedures, they established two robust interactomes, consisting of 6 baits and 307 prey proteins for humans, and 12 baits and 467 prey proteins for flies. To gain insights into the gene function, the authors investigated the interactors of alpha-arrestin proteins through various functional analyses, such as gene set enrichment. Furthermore, by comparing the interactors between humans and flies, the authors described both conserved and species-specific functions of the alpha-arrestin proteins. To validate their findings, the authors performed several experimental validations for TXNIP and ARRDC5 using ATAC-seq, siRNA knockdown, and tissue staining assays. The experimental results strongly support the predicted functions of the alpha-arrestin proteins and underscore their importance. `

I would like to suggest the following analyses to further enhance the study:

1. It would be valuable if the authors could present a side-by-side comparison of the interactomes of alpha-arrestin proteins, both before and after this study. This visual summary network would demonstrate the extent to which this work expanded the existing interactome, emphasizing the overall contribution of this study to the investigation of the alpha-arrestin protein family.

We greatly appreciate your insightful feedback. In response to the reviewer’s suggestion, we’ve depicted a network of known PPIs associated with α-arrestins (Figure S2C and D). Furthermore, by comparing our high-confidence PPIs to these known sets, we found that the overlaps are statistically significant and the high-confidence PPIs of α-arrestins broaden the existing interactome (Figure S2E).

From page 7 line 26 to page 8 line 8, we’ve detailed this side-by-side comparisons of existing interactome and newly discovered high-confidence PPIs of α-arrestins, as outline below.

“As a result, we successfully identified many known interaction partners of α-arrestins such as NEDD4, WWP2, WWP1, ITCH and TSG101, previously documented in both literatures and PPI databases (Figure S2C-F) (Colland et al., 2004; Dotimas et al., 2016; Draheim et al., 2010; Mellacheruvu et al., 2013; Nabhan et al., 2012; Nishinaka et al., 2004; Puca & Brou, 2014; Szklarczyk et al., 2015; Warde-Farley et al., 2010; Wu et al., 2013). Additionally, we greatly expanded repertoire of PPIs associated with α-arrestins in human and Drosophila, resulting in 390 PPIs between six α-arrestins and 307 prey proteins in human, and 740 PPIs between twelve α-arrestins and 467 prey proteins in Drosophila (Figure S2E). These are subsequently referred to as ‘high-confidence PPIs’ (Table S3).”

1. While the authors conducted several analyses exploring protein function, there is a need to further explore the implications of the interactome in human diseases. For instance, it would be beneficial to investigate the association of the newly identified interactome members with specific human diseases. Including such investigations would strengthen the link between the interactome and human disease contexts.

Thank you for your valuable comment. As suggested by the reviewer, we examined the association between α-arrestins’ interactomes and human diseases, incorporating our findings into the discussion. The newly introduced figure based on the result is Figure S10.

On page 24 lines 10-14, we’ve added discussion on Figure S10 as follows.

“We further explored association between α-arrestins’ interactomes and disease pathways (Figure S10). Notably, the interactomes of α-arrestins in human showed clear links to specific diseases. For instance, ARRDC5 is closely associated with disease resulting from viral infection and cardiovascular conditions. ARRDC2, ARRDC4, and TXNIP share common association with certain neurodegenerative diseases, while ARRDC1 is implicated in cancer.”

Reviewer #3:

Lee, Kyungtae and colleagues have discovered and mapped out alpha-arrestin interactomes in both human and Drosophila through the affinity purification/mass spectrometry and the SAINTexpress method. They found the high confident interactomes, consisting of 390 protein-protein interactions (PPIs) between six human alpha-arrestins and 307 preproteins, as well as 740 PPIs between twelve Drosophila alpha-arrestins and 467 prey proteins. To define and characterize these identified alpha-arrestin interactomes, the team employed a variety of widely recognized bioinformatics tools. These included protein domain enrichment analysis, PANTHER for protein class enrichment, DAVID for subcellular localization analysis, COMPLEAT for the identification of functional complexes, and DIOPT to identify evolutionary conserved interactomes. Through these analyses, they confirmed known alpha-arrestin interactors' role and associated functions such as ubiquitin ligase and protease. Furthermore, they found unexpected biological functions in the newly discovered interactomes, including RNA splicing and helicase, GTPase-activating proteins, ATP synthase. The authors carried out further study into the role of human TXNIP in transcription and epigenetic regulation, as well as the role of ARRDC5 in osteoclast differentiation. This study holds important value as the newly identified alpha-arrestin interactomes are likely aiding functional studies of this group of proteins. Despite the overall support from data for the paper's conclusions, certain elements related to data quantification, interpretation, and presentation demand more detailed explanation and clarification.

1. In Figure 1B, it is shown that human alpha-arrestins were N-GFP tagged (N-terminal) and Drosophila alpha-arrestins were C-GFP (C-terminal). However, the rationale of why the authors used different tags for human and fly proteins was not explained in the main text and methods.

We appreciate your valuable comment. Both N- and C-terminally tagged α-arrestins have been used previously. Given that our study aims to increase the repertoire of α-arrestin interacting proteins, where GFP is added might not be a concern. We note that GFP is a relatively bulky tag, and tagging a protein with GFP can potentially abolish the interaction with some of the binding proteins. Follow-up studies utilizing different approaches for detecting protein-protein interactions, such as BioID and yeast two-hybrid, will allow us to build more comprehensive α-arrestin interactomes.

1. In Figure 2A, there seems to be an error for labeling the GAL4p/GAL80p complex that includes NOTCH2, NOTCH1 and TSC2.

Thank you for comment. We double-checked COMPLEAT (protein COMPLex Enrichment Analysis Tool) database for the name of protein complex consisting of NOTCH1, NOTCH2, AND TSC2. The database indeed labeled this complex as the “GAL4p/GAL80p complex”. However, given the potential for mis-annotation (since we could not ascertain the relevance of these proteins to the “GAL4p/GAL80p complex”), we chose to exclude this protein complex from the network. The update protein complex network is illustrated in the revised Figure 2A.

1. In Figure 5, given that knockdown of TXNIP did not affect the levels and nuclear localization of HDAC2, the authors suggest that TXNIP might modulate HDAC2 activity. However, the ChiP assay suggest a different model - TXNIP-HDAC2 interaction might inhibit the chromatin occupancy of HDAC2, reducing histone deacetylation and increasing global chromatin accessibly. The authors need to propose a model consistent with these sets of all data.

We greatly appreciate your detailed feedback. Our data indicates a global decrease in chromatin accessibility (Figure 4C-G) and a diminished interaction between TXNIP and HDAC2 under depletion of TXNIP (Figure 5A). Additionally, we observed an increased occupancy of HDAC2 and subsequent histone deacetylation at TXNIP-target promoter regions (Figure 5C) without any changes in the HDAC2 expression level (Figure 5A) in TXNIP- knockdown cells. From these observations, we infer that the interaction between TXNIP-HDAC2 might suppress the function of HDAC2, a major gene silencer affecting the formation of condensed or accessible chromatin by deacetylating activity. Although we checked whether TXNIP could induce cytosolic retention of HDAC2 to inhibit nuclear function of HDAC2, TNXIP knockdown did not alter its subcellular localization (Figure 5B).

To elucidate the mechanism by which TXNIP inhibits the function of HDAC2, we further investigated the effect of TXNIP on the levels of HDAC2 phosphorylation, which is known to be crucial for its deacetylase activity and the formation of transcriptional repressive complex. However, as shown in the Figure S8C and D, the knockdown of TXNIP did not affect the HDAC2 phosphorylation status, as well as the interaction between HDAC2 and other components in NuRD complex in the immunoblotting and co-IP assays, respectively. The results suggest that TXNIP may inhibit the function of HDAC2 independently of these factors.

Following the reviewer’s suggestion, we carefully provided a proposed model describing the possible role of TXNIP in transcriptional regulation through interaction with HDAC2 and co-repressor complex in Figure S8E.

Description of these newly added figures can be found in the revised manuscript from page 18 line 7 to 27, as outlined below.

“HDAC2 typically operates within the mammalian nucleus as part of co-repressor complexes as it lacks ability to bind to DNA directly (Hassig, Fleischer, Billin, Schreiber, & Ayer, 1997). The nucleosome remodeling and deacetylation (NuRD) complex is one of the well-recognized co-repressor complexes that contains HDAC2 (Kelly & Cowley, 2013; Seto & Yoshida, 2014) and we sought to determine if depletion of TXNIP affects interaction between HDAC2 and other components in this NuRD complex. While HDAC2 interacted with MBD3 and MTA1 under normal condition, the interaction between HDAC2 and MBD3 or MTA1 was not affected upon TXNIP depletion (Figure S8C). Next, given that HDAC2 phosphorylation is known to influence its enzymatic activity and stability (Adenuga & Rahman, 2010; Adenuga, Yao, March, Seagrave, & Rahman, 2009; Bahl & Seto, 2021; Tsai & Seto, 2002), we tested if TXNIP depletion alters phosphorylation status of HDAC2. The result indicated, however, that phosphorylation status of HDAC2 does not change upon TXNIP depletion (Figure S8D). In summary, our findings suggest a model where TXNIP plays a role in transcriptional regulation independent of these factors (Figure S8E). When TXNIP is present, it directly interacts with HDAC2, a key component of transcriptional co-repressor complex. This interaction suppresses the HDAC2 ‘s recruitment to target genomic regions, leading to the histone acetylation of target loci possibly through active complex including histone acetyltransferase (HAT). As a result, transcriptional activation of target gene occurs. In contrast, when TXNIP expression is diminished, the interaction between TXNIP and HDAC2 weakens. This restores histone deacetylating activity of HDAC2 in the co-repressor complex, leading to subsequent repression of target gene transcription.”

1. The authors showed that ectopic expression of ARRDC5 increased osteoclast differentiation and function. Does loss of ARDDC5 lead to defects in osteoclast function and fate determination?

We appreciate your valuable comment. We have confirmed the endogenous expression of ARRDC5 in osteoclasts and conducted a loss-of-function study using shARRDC5. As determined by qPCR, ARRDC5 was endogenously expressed very low in osteoclasts. Even during RANKL-induced osteoclast differentiation, the CT value (29-31) for ARRDC5 expression was high in osteoclasts compared to the CT value (17-24) for the expression of marker genes Cathepsin K, TRAP, and NFATc1. Even though its endogenous expression was very low, we generated ARRDC5 knockdown cells by infecting BMMs with lentivirus expressing shRNA of ARRDC5 and subsequently differentiated the cells into mature osteoclasts. After five days of differentiation, we observed a significant decrease in the total number of TRAP-positive multinucleated cells (No. of TRAP+ MNCs) in shARRDC5 cells compared to that in the control cells. This result indicates that the loss of ARRDC5 leads to defects in osteoclast differentiation. Result of this loss-of-function study using shARRDC5 is depicted in Figure S9A and B.

In the revised manuscript, following sentence explaining Figure S9A and B was added on page 19 lines 15-17 as follows.

“Depletion of ARRDC5 using short hairpin RNA (shRNA) impaired osteoclast differentiation, further affirming its crucial role in this differentiation process (Figure S9A and B).”

1. From Figure 6D, the authors argued that ARRDC5 overexpression resulted in more V-ATPase signals: however, there is no quantification. Quantification of the confocal images will foster the conclusion. Also, western blots for V-ATPase proteins will provide an alternative way to determine the effects of ARRDC5.

We appreciate your insightful feedback. As suggested by the reviewer, we quantified V-type ATPase signals using confocal images, which were shown in Figure 6D. The ImageJ program was employed for integrated density measurements, and the integrated density of GFP-GFP overexpressing osteoclasts was set to 1 for relative comparison. The result in the revised Figure 6D revealed a significant increase in V-type ATPase signals in GFP-ARRDC5 overexpressing osteoclasts compared to that in GFP-GFP overexpressing osteoclasts, as outlined below.

We also agree with the reviewer’s comment that Western blot for V-ATPase proteins will be an alternative way to determine the effects of ARRDC5 in osteoclast differentiation. We have confirmed no different expression of V-type ATPase between GFP-GFP and GFP-ARRDC5 overexpressing osteoclasts using qPCR and western blot analysis. The corresponding western blot result is shown in the revised Figure S9C.

In addition, the corresponding qPCR that measures the expression level of V-type ATPase between GFP-GFP and GFP-ARRDC5 overexpressing osteoclasts is shown in Author response image 3.

Author response image 3.

Moreover, based on the references, the V-type ATPase is localized at the plasma membrane during osteoclast differentiation (Toyomura et al., 2003). Although mRNA and protein expression levels were similar in both cells, localization of V-ATPase in plasma membrane was significantly increased in GFP-ARRDC5 overexpressing osteoclasts compared to that in GFP-GFP osteoclasts, as shown in the revised Figure 6D above.

1. The results from Figure 6D did not support the authors' argument that ARRDC5 might control the membrane localization of the V-ATPase, as bafilomycin is the V-ATPase inhibitor. ARRDC5 knockdown experiments will help to determine whether ARRDC5 can control the membrane localization of the V-ATPase in osteoclast.

Thank you for your insightful comment. V-type ATPase has been reported to play an important role in the differentiation and function of osteoclasts (Feng et al., 2009; Qin et al., 2012). Given that various subunits of the V-type ATPase interact with ARRDC5 (Figure 6A), we speculated that ARRDC5 might be involved in the function of this complex and play a role in osteoclast differentiation and function. As answered above, GFP-ARRDC5 overexpressing osteoclasts showed a similar expression level of V-type ATPase to GFP-GFP cells but exhibited increased V-type ATPase signals at the cell membrane compared to those in GFP-GFP cells (Figure 6D). Additionally, co-localization of ARRDC5 and V-type ATPase was observed in the osteoclast membrane (Figure 6D), as predicted by the human ARRDC5-centric PPI network. On the other side, bafilomycin A1, a V-type ATPase inhibitor, not only blocked localization of V-type ATPase to plasma membrane in GFP-ARRDC5 overexpressing osteoclasts, but also reduced ARRDC5 signals (Figure 6D). These results indicate that ARRDC5 plays a role in osteoclast differentiation and function by interacting with V-type ATPase and promoting the localization of V-type ATPase to plasma membrane in osteoclasts.

V-type ATPase present in osteoclast membrane is important to cell fusion, maturation, and function during osteoclast differentiation (Feng et al., 2009; Qin et al., 2012). GFP-ARRDC5 overexpressing osteoclasts showed a significant increase of V-type ATPase signals in the cell membrane compared to GFP-GFP cells (Figure 6D), and also significantly increased cell fusion (No. of TRAP+ MNCs in Figure 6B) and resorption activity (resorption pit formation in Figure 6C). However, ARRDC5 knockdown in osteoclasts (shARRDC5 cells) showed a significant decrease in No. of TRAP+ MNCs compared to that in the control cells, indicating that the loss of ARRDC5 leads to defects in cell fusion during osteoclast differentiation (Figure S9A and B). As described above, the endogenous expression of ARRDC5 was very low in osteoclasts and could be specifically expressed in a certain timepoint during the differentiation. Therefore, to better understand the interaction with V-type ATPase of ARRDC5 in osteoclasts, ARRDC5 overexpression is more suitable than its knockdown.

Part of the manuscript on page 19 line 21 to page 20 line 6 was edited to support our statement, as outlined below.

“The V-type ATPase is localized at the osteoclast plasma membrane (Toyomura et al., 2003) and its localization is important for cell fusion, maturation, and function during osteoclast differentiation (Feng et al., 2009; Qin et al., 2012). Furthermore, its localization is disrupted by bafilomycin A1, which is shown to attenuate the transport of the V-type ATPase to the membrane (Matsumoto & Nakanishi-Matsui, 2019). We analyzed changes in the expression level and localization of V-type ATPase, especially V-type ATPase V1 domain subunit (ATP6V1), in GFP-GFP and GFP-ARRDC5 overexpressing osteoclasts. The level of V-type ATPase expression did not change in osteoclasts regardless of ARRDC5 expression levels (Figure S9C). GFP signals were detected at the cell membrane when GFP-ARRDC5 was overexpressed, indicating that ARRDC5 might also localize to the osteoclast plasma membrane (Figure 6D; Figure S9D). In addition, we detected more V-type ATPase signals at the cell membrane in the GFP-ARRDC5 overexpressing osteoclasts, and ARRDC5 and V-type ATPase were co-localized at the osteoclast membrane (Figure 6D; Figure S9D).”

1. The tables (excel files) do not have proper names for each table S numbers. Please correct the name of excel files for readers.

We appreciate your valuable comments. In response to the reviewer’s suggestion, we’ve renamed excel files to more appropriate titles for easier readability. List of renamed tables (excel files) are shown below.

Table S1. List of α-arrestins from human and Drosophila Table S2. Evaluation sets of α-arrestins PPIs Table S3. Summary tables of SAINTexpress results Table S4. Protein domains and short linear motifs in the α-arrestin interactomes Table S5. Enriched Pfam domains in the α-arrestin interactomes Table S6. Subcellular localizations of α-arrestin interactomes Table S7. Summary of protein complexes and cellular components associated with α-arrestin Table S8. Orthologous relationship of α-arrestin interactomes between human and Drosophila Table S9. Summary of ATAC- and RNA-seq read counts before and after processing Table S10. Differential accessibility of ACRs and gene expression Table S11. Summary of ATAC-seq peaks located in promoters and gene expression level Table S12. List of primer sequences used in this study

1. http://big.hanyang.ac.kr/alphaArrestin_Fly link does not work. Please fix the link.

We appreciate your comment. In response to the reviewer’s comment, we have made comprehensive α-arrestin interactome maps on our new website (big.hanyang.ac.kr/alphaArrestin_PPIN) and confirmed that users can be re-directed to networks housed in NDEx.

Author response image 4.

Screen shot of the first page of the newly developed website.

Website address: big.hanyang.ac.kr/‌‌‌‌‌‍‍‍‌‌alphaArrestin_PPIN

Author response image 5.

Screen shot of the gene-gene network involving α-arrestin in human.