The G3BP1-UPF1-Associated Long Non-Coding RNA CALA Regulates RNA Turnover in the Cytoplasm

Besides transcription, RNA decay accounts for a large proportion of regulated gene expression and is paramount for cellular functions. Classical RNA surveillance pathways, like nonsense-mediated decay (NMD), are also implicated in the turnover of non-mutant transcripts. Whereas numerous protein factors have been assigned to distinct RNA decay pathways, the contribution of long non-coding RNAs (lncRNAs) to RNA turnover remains unknown. Here we identify the lncRNA CALA as a potent regulator of RNA turnover in endothelial cells. We demonstrate that CALA forms cytoplasmic ribonucleoprotein complexes with G3BP1 and regulates endothelial cell functions. A detailed characterization of these G3BP1-positive complexes by mass spectrometry identifies UPF1 and numerous other NMD factors having cytoplasmic G3BP1-association that is CALA-dependent. Importantly, CALA silencing impairs degradation of NMD target transcripts, establishing CALA as a non-coding regulator of RNA steady-state levels in the endothelium.


Introduction
Maintaining equilibrium between RNA synthesis and decay is crucial for cellular function and homeostasis. Following transcription, the fate of a given RNA is directly linked to the protein factors it associates with in a structure and sequence-dependent manner [1]. Functionally, RNA decay serves a dual purpose. While it systemically removes aberrant and potentially toxic transcripts, RNA decay also counteracts transcription by balancing the levels of unmutated mRNA transcripts through degradation. In this way RNA decay enables fast changes in the RNA repertoire in response to external stimuli [2]. Concerning RNA surveillance and quality control, the degradation of defective transcripts is initiated by their identification through specific protein factors (e.g., the exon junction complex) followed by feeding them to the RNA decay machinery [2]. In this context, the nonsense-mediated decay (NMD) pathway, which eliminates mutant transcripts harboring a premature translation termination codon is a major route for RNA degradation [3]. Mechanistically, NMD is directly linked to translation termination [4] and essential to prevent accumulation of C-terminally truncated polypeptides. Key to this process is the ATP-dependent RNA helicase UPF1 (Regulator of nonsense transcripts 1) which catalyzes the remodeling of For silencing of gene expression, cells were transfected with LNAs (50 nM) using Lipofectamine RNAiMax (Life Technologies, Waltham, MA, USA) according to the manufacturer's instructions.

RNA Isolation, RT-qPCR, and RT-dPCR
Total RNA from HUVEC and HeLa cell culture was isolated and DNase digested using RNeasy Mini Kits (Qiagen, Venlo, The Netherlands), according to the manufacturer's instructions, and quantified with a NanoDrop 2000 spectrophotometer (Thermo Fisher, Waltham, MA, USA). cDNA synthesis was done from 500 ng RNA, using random hexamers and M-MLV reverse transcriptase (Thermo Fisher, Waltham, MA, USA). RPLP0-normalized (2 −∆Ct ) quantitative (q) PCR reactions were performed on StepOnePlus real-time PCR cyclers (Thermo Fisher, Waltham, MA, USA) and digital (d) PCR reactions were run on a QIAcuity One system (Qiagen, Venlo, The Netherlands).

In Vitro Sprouting Assays
HUVEC spheroid sprouting assays were conducted as described elsewhere [18]. Briefly, cells were detached after 24 h of transfection and seeded in an EBM-methocel mixture (80: 20) to 96-well U-bottom plates to allow spheroid formation by incubation for 24 h at 37 • C. Successfully formed spheroids were resuspended in a methocel-FCS mixture (80:20) before the same amount of rat-tail collagen type I (Corning Inc., Corning, NY, USA) was added. The spheroids were embedded by plating to 24-well plates and incubated for 24 h at 37 • C under basal conditions or presence of VEGFA (50 ng/mL). The next day, pictures of 10 spheroids per condition were taken using an Axio Observer Z1.0 microscope (Zeiss, Oberkochen, Germany) at 10× magnification. The cumulative sprouting length was determined using the Zeiss AxioVision digital imaging software (version 4.6).

Migration Assays
To determine the migratory capacity of HUVECs, transfected cells were cultured in fibronectin-coated 2-well cell culture inserts (Ibidi, Gräfelfing, Germany) for 24 h. By removing the cell culture inserts a cell-free gap between two confluent cell layers was created and gap closure was subsequently recorded by taking pictures at the indicated time points using an Axio Observer Z1.0 microscope (Zeiss, Oberkochen, Germany). Gap closure was analyzed with respect to the starting points.

In Vitro Permeability Assays
To analyze the permeability of endothelial cells, HUVECs were cultured on fibronectincoated cell culture inserts (ThinCert, 1 µm pore diameter, 24-well, Greiner Bio-One, Kremsmünster, Austria) to create a confluent cell layer and incubated for 1 h at 37 • C with FITC-dextran containing media (70 kDa, 1 mg/mL, Sigma-Aldrich, St. Louis, MO, USA) and the extravasation of FITC-dextran through the endothelial monolayer was analyzed by measuring the fluorescence (λex = 493 nm, λem = 518 nm) in the lower chamber using a GloMax-Multi+ Detection System (Promega, Fitchburg, MA, USA).

Mass Spectrometry
The mass spectrometry data sets, experimental details, and statistics have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository [20] and are publicly available with the data set identifiers PXD033516 and PXD033517. or kept separately. Regardless, extracts were adjusted to 2 mL using 2× IP buffer (100 mM NaCl, 150 mM KCl, 5 mM MgCl 2 , 1 mM DTT, 50 mM Tris-HCl (pH 8.0), 0.05% NP-40 (v/v), 2× protease inhibitor (Thermo Fisher, Waltham, MA, USA)) and 80 U RiboLock (Thermo Fisher, Waltham, MA, USA) were added. Next, 50 µL beads were resuspended in 1 mL lysate, incubated ON at 4 • C and thoroughly washed with 1× IP buffer. After washing with 1× PBS, beads were used for analysis by RT-qPCR, western blot and mass spectrometry.

Cellular Fractionation
Cellular fractionation was done using NE-PER Extraction Kits (Thermo Fisher, Waltham, MA, USA), according to the manufacturer's instructions. Briefly, HUVECs were washed and lysed in ice-cold CER I buffer for 10 min on ice. Next, CER II buffer was added, and samples were further incubated for additional 1 min on ice. Following centrifugation (16,000× g, 5 min, 4 • C), the cytoplasmic supernatant was taken, and nuclei were lysed in NER buffer for 40 min on ice. The nuclear lysate was cleared (16,000× g, 10 min, 4 • C) and RNA was isolated from cytoplasmic and nuclear fractions.

Statistics
Data are presented as means ± SEM and n refers to the number of independent biological replicates. Data normality was assessed using the Shapiro-Wilk normality test. Statistical significance was determined by the two-tailed unpaired t-test or Mann-Whitney U test. Multiple comparisons were performed using two-way ANOVA with Tukey's or Sidak's correction. Probability values of less than 0.05 were considered significant and indicated as follows: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Stimuli-Responsive lncRNA CALA Regulates Endothelial Sprouting and Migration
Located at the interface between the blood and the surrounding tissue, endothelial cells are paramount in sensing, and quickly adapting to environmental changes, thus guaranteeing cellular homeostasis and function [21]. The contribution of lncRNAs to these processes is not well-understood. Searching for stimuli-responsive lncRNAs in published datasets, 7 transcripts caught our interest, which were found to be rapidly activated in response to different external stimuli in numerous cell types [22]. Analyzing the expression levels of these lncRNAs in human umbilical vein endothelial cells (HUVECs), the lncRNA CALA was found to be the highest expressed ( Figure 1A,B). Absolute quantification of CALA levels in HUVECs under basal conditions confirmed a robust expression of~490 copies/ng total RNA ( Figure 1C), identifying CALA as a highly expressed, yet uncharacterized, lncRNA. To analyze the role and function of CALA in the endothelium, we first assayed expression changes of CALA upon different stimuli. Interestingly, ER stress, hypoxia, and starvation stimuli impacting endothelial cell functions [23][24][25] significantly drove CALA expression in HUVECs ( Figure 1D-F and Figure S1A).
guaranteeing cellular homeostasis and function [21]. The contribution of lncRNAs to these processes is not well-understood. Searching for stimuli-responsive lncRNAs in published datasets, 7 transcripts caught our interest, which were found to be rapidly activated in response to different external stimuli in numerous cell types [22]. Analyzing the expression levels of these lncRNAs in human umbilical vein endothelial cells (HUVECs), the lncRNA CALA was found to be the highest expressed ( Figure 1A,B). Absolute quantification of CALA levels in HUVECs under basal conditions confirmed a robust expression of ~490 copies/ng total RNA ( Figure 1C), identifying CALA as a highly expressed, yet uncharacterized, lncRNA. To analyze the role and function of CALA in the endothelium, we first assayed expression changes of CALA upon different stimuli. Interestingly, ER stress, hypoxia, and starvation stimuli impacting endothelial cell functions [23][24][25] significantly drove CALA expression in HUVECs (Figures 1D-F and S1A).  Next, we silenced CALA expression in HUVECs using locked nucleic acids (LNAs) ( Figure 1B) and assessed changes in endothelial cell functions. While transfection of HUVECs with CALA-targeting LNAs led to a~90% decrease in CALA levels ( Figure 1G), we did not observe effects on cell proliferation, apoptosis, or endothelial permeability  Figure S1B-D). In contrast, CALA silencing significantly impaired in vitro endothelial sprouting under basal, as well as under VEGFA-stimulated, conditions ( Figure 1H) and, additionally, cell migration was significantly reduced ( Figure 1I and Figure S1E,F). Taken together, these results indicated that CALA was required for the angiogenic capacity of HUVECs in vitro, presumably through contributing to cell migration.

CALA Interacts with Multiple RNPs and Primarily Associates with Cytoplasmic G3BP1
To understand the underlying molecular mechanism of CALA function in endothelial cells, we first assessed the coding potential of CALA which is located on chromosome 17 and constituted out of 6 exons (Figure 2A). RNA-seq data, ribosome profiling and computed coding probabilities [26] (Figure 2A and Figure S2A,B) show that CALA is spliced but does not encode proteins. Using sucrose density gradient ultracentrifugation of mock and proteinase K-treated total cell lysates, we found that CALA was strongly complexed with proteins, based on the substantial shift of the CALA signal away from the high molecular weight fractions towards the lighter ones (fraction #7 → #2-4) upon proteinase K treatment ( Figure 2B). To identify the endogenous protein interactome of CALA, we deployed antisense affinity purification of CALA-RNPs ( Figure S2C). To this end, we designed two distinct desthiobiotinylated 2 O-Me-RNA antisense probes (probe #1, probe #2), as described elsewhere [18,19] (Figure 1B), and purified CALA-RNPs from whole cell lysates. Purified RNA and protein fractions were subjected to RT-qPCR and mass spectrometry, respectively ( Figure S2C). We detected a significant enrichment of CALA over an unspecific non-target control (NTC) for probe #1 and #2 ( Figure 2C). In parallel, mass spectrometry of the co-purified protein fractions enabled the identification of CALA binding partners (PXD033516, Figure 2D). In detail, probes #1 and #2 specifically enriched 128 and 85 proteins, respectively, with 71 proteins being significantly enriched by both probes ( Figure 2E). In contrast to the CALA-specific probes, usage of the NTC control probe did not enrich any proteins ( Figure S2D,E) and comparison to our previously published interactome datasets of the lncRNAs NTRAS [19] and GATA6-AS [18] revealed a CALA specific protein interactome ( Figure S2F).
We first validated the interaction between CALA and hnRNP H1 as a representative for its family members by RNA immunoprecipitation (RIP) and observed a strong interaction of hnRNP H1 with CALA ( Figure S2G). While hnRNPs are well-known to regulate pre-mRNA splicing [29,30], an initial study we conducted excluded a splicing-regulatory function for CALA [19].
For the second identified protein network of cytoplasmic RNP granule proteins, factors involved in mRNA metabolism, were prevalent. Specifically, CALA affinity selection coenriched the multi-functional homologs G3BP1 and G3BP2 ( Figure 2D,F). Besides forming homo-and hetero-multimers [31,32], both proteins engage in RNA/DNA binding and numerous protein interactions [33,34]. In line with this, we also enriched numerous G3BPbinding proteins ( Figure 2D,F): E.g., CAPRIN1 (Cytoplasmic activation/proliferationassociated protein-1), PABPC1 (Polyadenylate-binding protein 1), NUFIP2 (Nuclear fragile X mental retardation-interacting protein 2), ATXN2L (Ataxin-2-like protein), and UPF1 [34] ( Figure 2D,F). Subsequently, we validated the interaction between CALA and G3BP1, the key component of CALA's identified cytoplasmic RNP granule protein network, by RIP from total cell lysate which revealed a substantial fraction of CALA (~20%) to be bound by G3BP1 ( Figure 2G). Since the interactome of CALA clearly separates into nuclear and cytoplasmic protein fractions, we next analyzed the subcellular distribution of CALA and accordingly found CALA present in both compartments, compared to specific marker transcripts ( Figure 2H). Subsequently, we specifically addressed the site of interaction between CALA and G3BP1 by anti-G3BP1 RIPs in cytoplasmic and nuclear fractions and demonstrated an almost exclusive cytoplasmic interaction between both binding partners ( Figure 2I). Taken together, these results revealed that the cytoplasmic fraction of CALA was extensively complexed in multivalent G3BP1-RNPs, the constituents of which are known for their mutual interactions and are functionally related to RNA metabolic processes.

CALA Impacts the Composition and Integrity of Cytoplasmic G3BP1-RNPs Driving mRNA Decay
To analyze the function of CALA within the identified G3BP1-positive RNPs, we first performed anti-G3BP1 RIPs in cytoplasmic lysates, followed by mass spectrometry (PXD033517, Figure 3A, left). Strikingly, this analysis revealed an overlap of 68 proteins (40%) that were also recovered by CALA antisense affinity selections ( Figure 3B, Table S1), highlighting a shared protein interactome. Among the proteins, we identified the G3BP1 binding partners G3BP2, CAPRIN1, PABPC1, NUFIP2, ATXN2L, and UPF1 [34], strongly supporting the notion that CALA is also part of this complex. We next repeated the anti-G3BP1 RIPs in CALA-silenced HUVECs and determined changes in the G3BP1 interactome by mass spectrometry (PXD033517, Figure 3A, right). After excluding an effect of CALA silencing on G3BP1 expression ( Figure S3A,B), we observed 36 of the G3BP1-interacting proteins to be differentially associated with G3BP1 upon CALA silencing. While only 3 proteins showed an augmented G3BP1 interaction, the remaining 33 proteins were significantly decreased in their G3BP1 interaction ( Figure 3C). Gene ontology analysis assigned the majority of these proteins to NMD [35] (Figure 3C,D and Figure S3C). Among the NMD proteins, we found reduced interactions between G3BP1 and UPF1 ( Figure 3C,D). UPF1 is a key regulatory factor of NMD and other mRNA decay pathways, thereby broadly affecting mRNA fate (extensively reviewed in [8]). Furthermore, interactions of the RNA helicase MOV10 with G3BP1 were reduced ( Figure 3C,D). MOV10 has been reported to be involved in mRNA decay [36], and was shown to be especially important for proper UPF1-mediated mRNA decay, as the translocation of MOV10 on target mRNAs ensures removal of secondary structures and proteins, enabling decay [37]. Additionally, CALA silencing disrupted the interaction between G3BP1 and ribosomal proteins like RPS6, RPS8, RPS11, and RPL23A ( Figure 3C,D). Interestingly, G3BP1 has been previously shown to interact with 40S subunits [38] and with respect to mRNA decay, UPF1 was reported to interact with 40S subunits upon ribosome stalling and the initiation of NMD [39]. Together, these findings support the notion that CALA is involved in NMD.  Given these results and the implication of NMD in RNA decay of non-mutant transcripts, we sought to examine the impact of CALA on RNA decay. Therefore, we measured the expression levels of well-known non-mutant NMD targets [40,41] upon CALA silencing under homeostatic conditions. Strikingly, CALA silencing significantly Given these results and the implication of NMD in RNA decay of non-mutant transcripts, we sought to examine the impact of CALA on RNA decay. Therefore, we measured the expression levels of well-known non-mutant NMD targets [40,41] upon CALA silencing under homeostatic conditions. Strikingly, CALA silencing significantly induced the mature transcript levels of GADD45A, GAS5, and RP9P ( Figure 3E-G and Figure S3D-F), while we could not observe changes in their precursor levels ( Figure 3E-G and Figure S3D-F). Of note, correctly spliced, non-mutated GADD45A and GAS5 transcripts are generally reported to be key targets of NMD, as their RNA levels must be kept low for preventing apoptosis and growth arrest [42,43]. While we could not detect a profound change in cell cycle progression, the observed impairment of angiogenesis and migration was in agreement with the role of GADD45A [44] and GAS5 [45] in endothelial cells.
In summary, our data indicated that CALA is required for the stabilization of cytoplasmic G3BP1-positive RNPs implicated in mRNA decay (Figure 4). We showed that the loss of CALA led to the disintegration of G3BP1-RNPs and, ultimately, to the stabilization of target transcripts, as evidenced by an increase in RNA expression levels ( Figure 4). Overall, these results describe CALA as a regulator of RNA decay in the cytoplasm. Of note, correctly spliced, non-mutated GADD45A and GAS5 transcripts are generally reported to be key targets of NMD, as their RNA levels must be kept low for preventing apoptosis and growth arrest [42,43]. While we could not detect a profound change in cell cycle progression, the observed impairment of angiogenesis and migration was in agreement with the role of GADD45A [44] and GAS5 [45] in endothelial cells.
In summary, our data indicated that CALA is required for the stabilization of cytoplasmic G3BP1-positive RNPs implicated in mRNA decay (Figure 4). We showed that the loss of CALA led to the disintegration of G3BP1-RNPs and, ultimately, to the stabilization of target transcripts, as evidenced by an increase in RNA expression levels ( Figure 4). Overall, these results describe CALA as a regulator of RNA decay in the cytoplasm.

Discussion
Homeostatic gene expression is tightly controlled and regulated at every step of the RNA life cycle. Besides transcription, processing, and RNA transport mechanisms, cytoplasmic RNA decay constitutes an important mechanism regulating ultimate transcript levels. Several pathways are known to regulate RNA decay rates and among these, NMD plays a pivotal role. Initially discovered as a translation-dependent process aiming to eliminate aberrant mRNAs [46], NMD was recently recognized for its role in non-mutant RNA turnover [47]. Thereby, NMD specifically balances transcriptional rates and enables a quick response to cellular stimuli requiring rapid changes of RNA expression levels [47]. Up to now, NMD has been known to be predominantly regulated via NMD factor availability [48,49], competition with other decay pathways [49,50], and by autoregulatory processes [49,[51][52][53].
In our study, we identified the undescribed lncRNA CALA to impact NMD efficiency by affecting the stability and composition of cytoplasmic G3BP1-positive RNPs involved in RNA turnover. Deploying antisense affinity selection of endogenous CALA-protein complexes from whole cell lysates, we identified CALA's protein interactome to be parted in two and in line with the observed nuclear-cytoplasmic distribution of the lncRNA. The co-purification of splicing factors might be predominantly of nuclear origin. We also selected numerous cytoplasmic RNP granule proteins and, among those, the significant enrichment of G3BP1, together with well-known G3BP1 interacting partners including UPF1 [34], was prevalent. Strikingly, anti-G3BP1 immunoprecipitation from cytoplasmic fractions followed by mass spectrometry uncovered a protein overlap of 40% with RNAbased CALA affinity selections, describing CALA to be part of cytoplasmic G3BP1-positive

Discussion
Homeostatic gene expression is tightly controlled and regulated at every step of the RNA life cycle. Besides transcription, processing, and RNA transport mechanisms, cytoplasmic RNA decay constitutes an important mechanism regulating ultimate transcript levels. Several pathways are known to regulate RNA decay rates and among these, NMD plays a pivotal role. Initially discovered as a translation-dependent process aiming to eliminate aberrant mRNAs [46], NMD was recently recognized for its role in non-mutant RNA turnover [47]. Thereby, NMD specifically balances transcriptional rates and enables a quick response to cellular stimuli requiring rapid changes of RNA expression levels [47]. Up to now, NMD has been known to be predominantly regulated via NMD factor availability [48,49], competition with other decay pathways [49,50], and by autoregulatory processes [49,[51][52][53].
In our study, we identified the undescribed lncRNA CALA to impact NMD efficiency by affecting the stability and composition of cytoplasmic G3BP1-positive RNPs involved in RNA turnover. Deploying antisense affinity selection of endogenous CALA-protein complexes from whole cell lysates, we identified CALA's protein interactome to be parted in two and in line with the observed nuclear-cytoplasmic distribution of the lncRNA. The copurification of splicing factors might be predominantly of nuclear origin. We also selected numerous cytoplasmic RNP granule proteins and, among those, the significant enrichment of G3BP1, together with well-known G3BP1 interacting partners including UPF1 [34], was prevalent. Strikingly, anti-G3BP1 immunoprecipitation from cytoplasmic fractions followed by mass spectrometry uncovered a protein overlap of 40% with RNA-based CALA affinity selections, describing CALA to be part of cytoplasmic G3BP1-positive complexes. Trying to elucidate the function of CALA within those G3BP1-positive cytoplasmic RNPs, we found that CALA silencing leads to a tremendous loss of protein factors, suggesting CALA is of structural importance for cytoplasmic G3BP1-positive complexes. G3BP1, the main cytoplasmic protein factor enriched by CALA affinity selections, has been well studied regarding its role in stress granule formation [32]. However, recent studies also proposed diverse functions for G3BP1 in the absence of cellular stress. For example, G3BP1 plays an important role in RNA metabolism via its endoribonuclease activity [54] and contributes to structure-mediated RNA decay [14]. Our mass spectrometry analysis revealed G3BP1 to interact with several proteins implicated in NMD in a CALAdependent manner, as we found UPF1 and MOV10 to be reduced in their interaction with G3BP1 upon CALA silencing. Both factors are indispensable for efficient RNA decay [8,36]. In agreement, we show a direct implication of CALA in NMD regulation, as silencing of CALA leads to a significant increase in NMD target expression, but, at the same time, to no increase in precursor transcript expression.
Different lncRNAs have been shown to interact with G3BP1 [55,56], UPF1 [37,[57][58][59] and MOV10 [37]. In contrast to these studies, which focus on, and elucidate, the individual interaction case-specifically, we, for the first time, have reported on a lncRNA to control mRNA turnover by being a crucial structural component of cytoplasmic G3BP1positive RNPs (Figure 4). These results are in line with the ability of lncRNAs to act as complex-stabilizing scaffolds, as shown for the abundant lncRNA NEAT1 in the context of paraspeckle formation [60]. Of note, the initial identification of CALA in different cell types [22] along with the ubiquitously expressed CALA binding proteins, strongly argues for a regulatory network not restricted to the endothelium. For example, CALA is additionally well expressed in diverse cancer types, conditions in which NMD is known to play a complex role [61]. Finally, the transcriptional regulation of CALA is not characterized. An in-silico promotor analysis suggests the involvement of the stress-sensitive transcriptional repressor CTCF [62], which is in line with our observed upregulation of CALA in different cellular stress settings. Taken as a whole, our study demonstrates the lncRNA CALA ensures the integrity and functionality of cytoplasmic G3BP1-positive complexes. By doing so, CALA regulates homeostatic RNA decay and gene expression, via a novel, lncRNA-dependent contribution to cytoplasmic decay pathways.

Limitations of the Study
The here presented work aimed to address the cytoplasmic role of G3BP1 and CALA in homeostatic RNA decay, focusing on its mechanism of action. RNA decay, however, is also coupled with stress granule formation. While initial studies done by us suggest that CALA silencing has no direct effect on stress granule formation, we cannot exclude an additional function of CALA under stress conditions. Future studies may, therefore, focus on the role of the identified CALA-G3BP1 interaction beyond homeostasis and comprehensively characterize and describe the target transcripts regulated by CALA-G3BP1 RNPs.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ncrna8040049/s1, Figure S1: Stimuli-responsive lncRNA CALA regulates endothelial sprouting and migration, Figure S2: CALA interacts with multiple RNPs and primarily associates with cytoplasmic G3BP1, Figure S3: CALA impacts the composition and integrity of cytoplasmic G3BP1-RNPs driving mRNA decay; Table S1: List of identified proteins in anti-G3BP1 IP via mass spectrometry and their overlap with proteins identified in antisense affinity selection of CALA (related to Figure 3B), Table S2: List of primer, LNA, and probe sequences (related to Section 2).  Data Availability Statement: The mass spectrometry data sets have been deposited with the Pro-teomeXchange Consortium via the PRIDE partner repository and are publicly available with the data set identifiers PXD033516 and PXD033517.