PERK/EIF2AK3 integrates endoplasmic reticulum stress-induced apoptosis, oxidative stress and autophagy responses in immortalised retinal pigment epithelial cells

Retinal pigment epithelium (RPE) performs essential functions for ensuring retinal homeostasis and is a key site for pathogenic changes leading to age-related macular degeneration (AMD). Compromised proteostasis in RPE results in ER stress and ER stress-dependent antioxidant, apoptosis and autophagic responses. ER stress induces the unfolded protein response (UPR) in which EIF2AK3, encoding the protein kinase RNA-like ER kinase (PERK), acts as a key regulator. Downregulated EIF2AK3 gene expression has recently been identified in AMD using human donor RPE, however the molecular mechanisms that integrate the various ER-mediated cellular pathways underpinning progressive RPE dysfunction in AMD have not been fully characterised. This study investigated the downstream effects of PERK downregulation in response to Brefeldin A (BFA)-induced ER stress in ARPE-19 cells. PERK downregulation resulted in increased ER stress and impaired apoptosis induction, antioxidant responses and autophagic flux. ARPE-19 cells were unable to efficiently induce autophagy following PERK downregulation and PERK presented a role in regulating the rate of autophagy induction. The findings support PERK downregulation as an integrative event facilitating dysregulation of RPE processes critical to cell survival known to contribute to AMD development and highlight PERK as a potential future therapeutic target for AMD.


Results
PERK downregulation increases ER stress through impaired UPR. We first assessed whether ER stress and PERK pathway signalling can be induced in ARPE-19 cells by treating the cells with 1 µg/ml BFA for a time-course of 0-48 h (Fig. 1a-e). Induction of ER stress was confirmed by time-dependent upregulation of GRP78 from the 24-h time point, increasing further at 48-h (Fig. 1a,e). We next investigated whether BFAinduced ER stress initiated canonical PERK pathway signalling in ARPE-19 cells by assessing phosphorylation of PERK (Fig. 1b,e). Phosphorylation of PERK was assessed using the slower migration of total PERK detected by a band at 170 kDa as previously established 8 . Total PERK protein expression did not increase during BFA-induced ER stress, however we observed slower migration of the 170 kDa PERK band, suggestive of PERK phosphorylation at 24-and 48-h time points compared to non-treated (NT) samples at the 0-h time point (Fig. 1b,e). We then examined downstream effectors of PERK including phosphorylation of eif2α at serine 51 relative to total eif2α (p-eif2α (Ser51)/eif2α) (Fig. 1c,e) and CHOP (Fig. 1d,e) protein expression. Both displayed significant time-dependent increase at 24 and 48 h, compared to NT samples at the 0-h time point (Fig. 1c,d). Together this confirmed that ER stress can be induced via canonical PERK signalling pathway in ARPE-19 cells. Next, to assess the role of PERK in regulating the ARPE-19 ER stress-induced response, we carried out and confirmed siRNAmediated PERK knockdown (Fig. 1g,j). siRNA-mediated PERK knockdown was carried out in the presence of BFA-induced ER stress for a time-course of 0-48 h. This was accompanied by a significant downregulation in downstream p-eif2α (Ser51)/eif2α (Fig. 1h,j) and CHOP expression (Fig. 1i,j) at 24-and 48-h time points. BFAinduced ER stress led to increased GRP78 protein expression following PERK knockdown at the 24-h time point, which was sustained at 48-h time point (Fig. 1f,j). Increased GRP78 expression following downregulated PERK pathway signalling suggests an increase in the requirement for ER chaperone-mediated degradation 8 , potentially as a result of increased ER-protein load due to a loss of global translation attenuation following decreased eif2α phosphorylation 8 . Together, these data indicate that ER stress is increased and the overall adaptive UPR is impaired following PERK downregulation in response to BFA-induced ER stress in ARPE-19 cells.

PERK downregulation impairs ER stress-induced apoptosis in a Bax/Bcl-2-dependent manner.
In conditions of chronic ER stress, the pro-survival UPR switches to PERK-dependent apoptosis via upregulation of CHOP 19  www.nature.com/scientificreports/ of pro-apoptotic CHOP expression (Fig. 1d,e), which is downregulated following PERK knockdown (Fig. 1i,j), confirming that CHOP is regulated in a PERK-dependent manner in ARPE-19 cells. Therefore, we hypothesized that PERK regulates ER stress-induced apoptosis in ARPE-19 cells. Significantly increased apoptosis was detected by flow cytometry following BFA-induced ER stress by 48-h time point (Fig. 2a,b), with cells starting to display specific apoptotic morphology by 24-h ( Fig. 2b-top row). Using the Bcl-2/Bax ratio as an established measure to determine cellular sensitivity to apoptotic stimuli in RPE cell models 27 , we also observed that the Bcl-2/Bax ratio decreased in a time-dependent manner upon BFA-induced ER stress in ARPE-19, supporting the finding that the induced ER stress is associated with increased pro-apoptotic signalling (Fig. 2c,d). We next investigated the role of PERK in regulating ER stress-induced apoptosis following BFA treatment. ARPE-19 cells were transfected with PERK siRNA or a non-targeting control siRNA followed by a 0-48-h time course of 1 μg/ml BFA. Following BFA treatment and PERK knockdown, ARPE-19 cells presented a more Figure 1. PERK downregulation increases ER stress. PERK signalling pathway mediates induction of ER stress in ARPE-19 cells treated with 1 μg/ml BFA through increased expression of GRP78 (a), PERK (b), p-eif2α (Ser51)/eif2α (c) and CHOP (d); representative immunoblots shown at 24-h and 48-h time-points (e). BFAinduced ER stress in cells in which PERK is downregulated with siRNA leads to significant upregulation of GRP78 (f) at 24-h time point and significant downregulation of PERK pathway components PERK (g), p-eif2α (Ser51)/eif2α (h) and CHOP (i) at 24 and 48-h time points when compared to non-targeting control siRNA; (i) representative immunoblots shown. Phosphorylation of PERK was observed from the slower migration of a total PERK band at 170 kDa following treatment with BFA (b,g,e,j). Phosphorylation of eif2α at serine 51 (p-eif2α (Ser 51)) was calculated relative to total eif2a protein expression (c,h). Results are presented as foldchanges of the 0-h time point (a-d) or the NT group in cells transfected with non-targeting control siRNA (f-i). Protein levels were normalised to α-tubulin. Means ± standard deviation are shown (n = 3). Statistical analysis was performed using a one-way ANOVA with Dunnet's post-hoc test (a-d) or two-way ANOVA with Sidak's multiple-comparison test (f-i) (*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001). BFA Brefeldin A, NT nontreated, NS non-significant.  (Fig. 2f). We further found that there was a significant increase in Bcl-2/Bax protein expression ratio in PERK knockdown cells at the 48-h time point (Fig. 2g,h). These data suggest that PERK regulates the rate of apoptosis over a 48-h time course by sensitising ARPE-19 cells to programmed cell death in a Bcl2/ Bax-dependent manner.
PERK downregulation impairs the ER stress-induced antioxidant response through downregulated Nrf2 signalling. The RPE is subjected to a high oxidative burden 28 . A central part of maintaining RPE redox homeostasis is upregulation of antioxidant Nrf2 signalling in response to oxidative stress 29 . Oxidative stress has been shown to upregulate ER stress signalling in RPE and PERK has been shown to directly phosphorylate Nrf2 in response to ER stress 9,10 . Therefore, we hypothesised that PERK may regulate the antioxidant response to BFA-induced ER stress in ARPE-19 cells.
Firstly we investigated Nrf2 phosphorylation at serine 40 (p-Nrf2 (S40)) and NQO1 expression during a 48-h time-course of BFA treatment (Fig. 3a-c) and observed significant time-dependent decreases in p-Nrf2 (S40) (Fig. 3a,c) and NQO1 (Fig. 3b,c). The role of PERK in regulation of the antioxidant response mounted in the same conditions of ER-stress was investigated in PERK-knockdown ARPE-19 cells, which were unable to induce phosphorylated Nrf2 at 24-h and 48-h time points compared to ARPE-19 cells transfected with non-targeting control siRNA (Fig. 3d,f). Following PERK knockdown, NQO1 protein expression was significantly decreased even in the absence of additional ER stress and showed a further significant decrease following BFA-induced ER stress (Fig. 3e,f). Taken together, these data confirm the role of PERK in activation, through phosphorylation of Nrf2 and further demonstrate that downregulation of PERK dysregulates the response to oxidative stress in ARPE-19 cells mediated by the PERK-Nrf2 signalling axis.
PERK modulates the rate of autophagy induction in ARPE-19 cells following ER stress. Autophagy is regulated by sustained ER-stress and PERK pathway signalling was previously shown to be implicated in this process P19 embryonic carcinoma cells constitutively expressing c-jun (C2C5 cells) 25,30 . Having confirmed a regulatory role for PERK in ER-/oxidative stress responses and related apoptosis modulation in ARPE-19 cells, we sought to investigate the role of PERK in regulation of autophagy following BFAinduced ER stress in the presence and absence of the lysosomal inhibitor chloroquine (CQ).

PERK knockdown impairs the induction of autophagy in ER-stressed ARPE-19 cells in response to serum starvation.
It was previously demonstrated that in the pathological scenario of AMD, autophagy could not be induced by serum starvation in RPE cells 24 . Having established that PERK is essential for the rate of autophagy induction in response to ER stress and that PERK knockdown ARPE-19 cells display impaired autophagic flux, we assessed whether PERK knockdown ARPE-19 cells are capable of inducing autophagy when challenged with serum starvation and ER stress. We carried out PERK knockdown in ARPE-19 cells followed by ER stress induction using BFA under serum present (basal) or serum starved (autophagy induced) conditions. This was performed in the presence or absence of CQ pre-treatment and assessed at a 24-h time point (Fig. 5).When ARPE-19 cells were treated with BFA and CQ (BFA + CQ group) to concurrently induce ER stress and inhibit the lysosome under serum present conditions, there was a significant increase in LC3B-II (Fig. 5a,e) and a small but insignificant increase in P62 protein expression (Fig. 5c,e) compared to BFA without lysosome inhibition (BFA group) in PERK knockdown groups. Conversely, when autophagy was induced using serum starvation following PERK knockdown, a significant decrease in both LC3B-II (Fig. 5b,e) and P62 protein expression (Fig. 5d,e) was observed in samples when comparing cells concurrently treated with BFA and CQ (BFA + CQ group) to cells treated with BFA alone (BFA group). Therefore in a setting of serum starvationinduced autophagy, in addition to BFA-induced ER stress and CQ-induced lysosome inhibition, the significant www.nature.com/scientificreports/ reduction in LC3B-II and P62 expression observed suggests a relative reduction in autophagosome number, which likely reflects an inability of PERK knockdown ARPE-19 to induce autophagy.

Discussion
This study aimed to investigate the molecular determinants that integrate ER stress-mediated pathways known to contribute to dysregulated cellular responses that lead to progressive RPE dysfunction in AMD. This is the first study to our knowledge that identifies an integrative role for PERK downregulation in increased ER stress, impaired autophagic flux, autophagy induction and Nrf2-mediated antioxidant response in a setting of ER stress, resulting from compromised proteostasis using BFA. Furthermore our results support previous findings of impaired ER stress-associated apoptosis 10 , which in the context of the data obtained in this study support a role for PERK downregulation in the impairment of RPE functions known to contribute to the development of AMD.
In this study we observed elevated GRP78 expression following PERK knockdown during BFA-induced ER stress. GRP78 acts as the initiating component of ER stress signalling and its upregulation functionally defines the UPR 23 . The UPR principally acts to attenuate global translation via PERK and eif2α phosphorylation 23 . Previous studies have determined that PERK knockdown in mouse embryonic fibroblasts (MEF) results in increased 35 S-labeled cellular protein in the ER, indicative of an impaired attenuation of global translation and an increase of protein binding to GRP78 due to a loss of PERK kinase activity and eif2α phosphorylation 8 . In this context, our findings support the notion that upregulated GRP78 is an adaptive response intended on delaying induction of apoptotic pathways in ARPE-19 cells. Evidence for such a response was also obtained by studies demonstrating increased survival of chinese hamster ovary cells following GRP78 overexpression, supporting the pro-survival role of GRP78 32 . In the context of AMD, the RPE may display an impaired UPR in a setting of PERK downregulation, however further experimental evidence in human donor tissue or primary cell cultures is required.
Following PERK downregulation, ARPE-19 cells displayed impaired apoptosis induction associated with significant downregulation of CHOP expression. Our evidence of this protective effect corroborates previous findings which demonstrate that adenoviral-mediated PERK knockdown protects ARPE-19 cells from ER stressassociated apoptosis following hydroquinone and visible light exposure (400-800 nm)-induced ER stress via CHOP downregulation 10,11 . Indeed, CHOP has been shown to sensitise cells to apoptosis by downregulating gene expression of the anti-apoptotic protein Bcl-2 and upregulating pro-apoptotic Bax 22,33 . In human foetal RPE, tunicamycin-induced ER stress has been shown to downregulate the Bcl-2/Bax ratio and a further study in ARPE-19 cells has demonstrated regulation of the Bcl-2/Bax ratio in an x-box binding protein (XBP1)-dependent manner following cigarette smoke extract treatment 10,27 . Importantly, we identify that downregulation of PERK and subsequently CHOP regulates the Bcl-2/Bax ratio, which collectively suggests intrinsic apoptosis is induced via alterations in the Bcl-2/Bax ratio and that this is under the regulation of PERK in ARPE-19.
This study did not examine the well documented role of PERK activation resulting in cyclin D1 inhibition via phosphorylation of eif2α leading to cell cycle arrest 34 . Specifically, Brewer and Diehl posited that cell-cycle exit is part of the process of coordinating apoptosis initiation in ER stress and that PERK expression is key to the regulation of this process 34 . In this respect, while PERK knockdown was shown to increase sensitivity to ER-stress apoptosis in embryonic stem cells 8 , work from Chen et al. concluded that PERK knockdown in ARPE-19 cells results in apoptosis inhibition during ER stress 10 . This suggests that the role of PERK in regulating decisions of cellular fate is cell-type dependent. The work of Brewer and Diehl bought forth the theory that the induction of cell-cycle arrest by PERK after activation of the UPR may provide time necessary to re-establish cellular homeostasis. However, it is unclear as to whether this is true for the RPE and further experimentation is required to confirm the dynamics between PERK downregulation, cell-cycle arrest and apoptosis.
As downregulated EIF2AK3 was documented in the RPE of AMD patients 16 , the findings outlined in this study in the model ARPE-19 system could suggest that in vivo, RPE in AMD may lose the ability to initiate the apoptotic program in response to accumulated misfolded proteins in the ER, possibly as a protective measure to ensure cellular survival of the post-mitotic RPE. However, this must be interpreted with caution as unlike the RPE in vivo, cultured ARPE-19 cells can be highly proliferative and are accompanied by a phenotype associated with de-differentiation 35 . This may explain why our findings using the immortalised ARPE-19 cell line differ from those reported by Dunaief et al., who described upregulated apoptosis from AMD RPE in vivo 36 . Therefore, while we suggest that the upregulation of apoptosis in AMD RPE is not directly or solely mediated by ER stress signalling, further experiments in primary donor tissue are required to confirm this effect. The RPE is subject to oxidative damage from both endogenous and exogenous pro-oxidant sources 28 . Nrf2 is a principal regulator of phase II antioxidant enzymes and detoxifying gene expression and displays decreased gene expression in AMD RPE 28,37 . PERK has been established as a direct kinase of Nrf2, facilitating Nrf2 phosphorylation and phase-II antioxidant NQO1 expression in response to ER stress 9,20 . As the UPR via PERK pathway signalling has been shown to increase oxidative stress, we reasoned that ARPE-19 cells are unable to induce an antioxidant response via Nrf2 signalling following PERK downregulation in a setting of BFA-induced ER stress 38 . However, in response to BFA-induced ER stress, we observed a time-dependent decrease in Nrf2 phosphorylation and NQO1 protein expression. Studies in RPE and MEF cells have shown an inverse regulatory relationship between Nrf2 and CHOP protein expression 20,39 . Hence it is possible that BFA-induced and time-dependent CHOP upregulation in this study may explain the decrease in Nrf2 signalling and resultant decrease in NQO1 expression at the same timepoints. While previous studies in ARPE-19 cells have demonstrated that the Nrf2 signalling axis is regulated in an XBP1 dependent manner, we sought to establish whether this was true for PERK 40 . Importantly, in PERK knockdown ARPE-19 cells, we observed a decrease in both phosphorylated Nrf2 and NQO1 expression in the absence of ER stress, which was further reduced following ER-stress induction using BFA. Our data demonstrates that loss of PERK impairs the Nrf2-dependent antioxidant response to ER stress in ARPE-19 cells. While our findings are in line with previous evidence of PERK-dependent Nrf2 phosphorylation in MEF cells 9 16,37 . PERK and its downstream signalling effectors have been shown to be essential in regulating the induction of autophagy in response to ER stress 41 . In ARPE-19 cells, autophagy can be induced by photooxidative or highglucose environments that result in ER stress 11,42 . The impairment of autophagy has been extensively implicated in the ageing RPE and in the pathogenesis of AMD 24,43,44 . Our data show that combined BFA-induced ER stress and lysosomal inhibition using CQ result in a synergistic increase in LC3B-II and P62 expression compared to BFA treatment alone. This suggests that ER stress and induction of the UPR regulates the rate of autophagic induction and results in autophagosome formation exceeding that of lysosomal degradation, thus impairing autophagic flux in ARPE-19 cells 31 . Furthermore, we observed that PERK knockdown ARPE-19 cells displayed even greater increases in LC3B-II and P62 expression in the presence of BFA alone and BFA + CQ treatment compared to non-targeting control siRNA cells. This indicates a key role for PERK in the regulating autophagy induction. Early studies established that classical ER stress inducing agents such as thapsigargin and tunicamycin upregulate autophagosome accumulation in an ER stress-specific manner, playing a protective role in settings of ER stress 25,[45][46][47] . As such, autophagy may represent a para-adaptive mechanism supportive of the UPR 25,[45][46][47] . It has been previously established in MEF cells that during ER stress, protein influx into the ER continues following PERK knockdown 8 , which could explain why autophagy induction is dramatically increased and autophagy flux is impaired in PERK knockdown ARPE-19 cells. Thus, the requirement for autophagy as a supportive mechanism of the UPR becomes apparent, aiding in the clearance of a large amounts of accumulated protein from the ER 25,[45][46][47] . Therefore, we suggest that BFA-induced ER stress regulates the induction of autophagy and that PERK regulates the rate of autophagy induction in ARPE-19 cells.
A recent study demonstrated that following nutrient starvation, primary RPE cells cultured from AMD donors were unable to induce autophagy 24 . As PERK pathway signalling was shown to be essential in regulating the induction of autophagy in MEF cells 41 , we sought to investigate whether PERK knockdown ARPE-19 cells were able to induce autophagy when challenged with serum starvation in a setting of BFA-induced ER stress. Under these conditions, we observed an increase in autophagosome formation, indicated by LC3B-II and P62 protein expression; however, following BFA-induced ER stress in the presence of CQ-induced lysosomal inhibition, both LC3B-II and P62 protein expression decreased compared to BFA treatment alone. If PERK knockdown ARPE-19 cells were competent in autophagy induction, a synergistic effect on the increase of LC3B-II would be expected following lysosomal inhibition 48,49 . Our findings show that when lysosomal degradation is inhibited using CQ, autophagosome formation is occurring at a lower rate, suggesting that PERK knockdown ARPE-19 cells are unable to induce autophagy in response to nutrient starvation. Furthermore, while an increase in P62 expression may indicate autophagosome accumulation and impairment 49 , a decrease could possibly reflect lower number of total autophagosomes formed as a result of impaired autophagy induction. These findings are in line with the observations of Mitter et al. who described significant reduction of LC3 in RPE in early and late-stage AMD compared to age-matched controls 43 .
A previous study examining the regulation of gene expression by DNA methylation in early-stage human donor AMD RPE revealed that EIF2AK3 (encoding PERK) gene expression is downregulated, although not associated with differential DNA methylation 16 . Unanswered questions remain as to whether EIF2AK3 downregulation acts as an adaptive mechanism to ageing-related alterations which ultimately result in RPE dysfunction in late-stage disease. Indeed, inhibited apoptosis following PERK downregulation in a setting of BFA-induced ER stress observed in this study does not reflect upregulated apoptosis observed in later stages of disease 36 . Therefore, it is possible that PERK is downregulated in earlier stages of disease as an adaptive mechanism to prevent cell death in the post-mitotic RPE.
The UPR is comprised of a highly complex and interconnected network of signalling cascades that are yet to be fully delineated 23 . In this context, the next steps for fully characterising the regulatory role of PERK Figure 5. ER stressed ARPE-19 are unable to induce autophagy in response to serum starvation following PERK knockdown. ARPE-19 cells were subjected to 1 μg/ml BFA-induced ER stress with lysosome inhibition using 10 μM CQ for 24 h in the presence of PERK siRNA and non-targeting control siRNA to assess protein expression of LC3B-II (a,b) and P62 (c,d) in serum or serum starvation conditions, respectively. Consistent with previous results, there was a significant increase of LC3B-II protein level in BFA + CQ compared to both BFA and NT groups in cells transfected with non-targeting control siRNA, both in the presence and in the absence of serum. However, following PERK knockdown expression, while a significant increase in LC3B-II expression is observed in BFA + CQ compared to BFA groups in serum present conditions, there is a significant decrease in serum absent conditions. In (c) serum present conditions P62 displays a significant increase in expression with regards to the BFA + CQ and BFA alone groups compared to NT groups in cells transfected with PERK siRNA. No difference when comparing P62 expression between BFA + CQ and BFA alone. However, there was a significant decrease in P62 expression when comparing BFA + CQ and BFA groups in serum starvation conditions (d). Thus, in response to serum starvation, PERK knockdown ARPE-19 are unable to induce autophagy in conditions of ER stress. Protein levels were normalised to α-tubulin. Means ± standard deviations shown (n = 3). Statistical analysis was performed using a two-way ANOVA with TukeyHSD post-hoc test by comparing NT and BFA alone and combined BFA + CQ groups, within the non-targeting control siRNA and PERK siRNA group, respectively (*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001). NT Non-treated, BFA Brefeldin-A, CQ Chloroquine. www.nature.com/scientificreports/ downregulation in ARPE-19 and further in AMD RPE need to focus on examining the induction and downstream signalling pathways of the other two major components of UPR, the IRE1 and ATF6, that also regulate ER stress, apoptosis and autophagy following PERK knockdown 8,10,40 .

Conclusion
The findings of this study evidence a key role for PERK and components of the PERK signalling pathway in maintaining RPE homeostasis. Our results in the context of previous studies suggest that the PERK-encoding EIF2AK3 gene may be downregulated in early/intermediate AMD 16 to protect cells from apoptosis as a result of mounting ER stress early in the disease course. We propose that following PERK downregulation, the compensatory UPR and cellular antioxidant defence system is impaired. However upregulated ER chaperone expression alone is unable to maintain proteostasis, therefore an inhibition of apoptotic signalling and significant increase in the rate of protective mechanisms such as autophagy are required to support misfolded protein clearance. However, it is possible that alone these compensatory mechanisms are unable to mitigate mounting cellular dysfunction leading to RPE apoptosis observed in late-stage AMD 36 . With clear evidence of its role in dysregulated ER stress signalling pathways in RPE cells, the PERK signalling pathway emerges as a critical component of the complex regulatory mechanisms influencing AMD pathogenesis and as such may provide avenues for future therapeutic interventions.   SDS-PAGE and western immunoblotting. ARPE-19 cell lysates were prepared for western immunoblotting by washing wells in a 6-well plate twice with dPBS and scraping wells in lysis buffer (0.128 M β-mercaptoethanol, 40 mM tris, 10% glycerol, 1% SDS, 0.01% bromophenol blue). The cell lysate was homogenised on ice carefully using a 21G needle and incubated (95 °C, 5 min) to ensure sample denaturation. Western immunoblotting was performed as described previously 52,53 . Primary and secondary antibodies used in western immunoblotting are listed in Tables 1 and 2, respectively. Protein expression for experimental proteins of interest were normalised for loading using the loading control α-tubulin. For immunoblotting involving probing for P62 (SQSTM1) protein expression, membranes were stripped by incubating with stripping buffer (3 M Tris, 10% SDS, 0.128 M β-mercaptoethanol pH 6.7) (50 °C, 30 min) with gentle agitation and subsequently washed three times for 5 min in washing buffer (0.02 M Tris, 0.14 M NaCl, 0.1% Tween-20, pH 7.6) before being blocked in blocking buffer (0.02 M Tris, 0.14 M NaCl, 0.1% Tween-20, 5% (w/v) non-fat dry milk) and incubated in the respective primary antibody overnight for visualisation.

Methods
Statistical analysis. Data generated in this study was analysed to determine statistical significance using the GraphPad Prism 8 (GraphPad Software, USA). Sample means were compared to control groups using Oneway Analysis of Variance (ANOVA) with Dunnett's or Sidak's post hoc-test. When two or more treatments were measured alone or over a time-course, comparisons were made using a Two-way ANOVA with Sidak's or Tuk-eyHSD post-hoc test. All experiments consisted of n = 3 replicates, unless otherwise stated. All data values are presented as mean ± standard deviation and, when appropriate, as fold change of control groups for each experiment. p-values are shown in figure legends where *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.