The BAG-1 isoform BAG-1M regulates keratin-associated Hsp70 chaperoning of aPKC in intestinal cells during activation of inflammatory signaling

ABSTRACT

Atypical PKC (ι/λ and ζ; hereafter referred to as aPKC) is a key player in the acquisition of epithelial polarity and participates in other signaling cascades including the control of NF-κB signaling. This kinase is post-translationally regulated through Hsp70-mediated refolding. Previous work has shown that such a chaperoning activity is specifically localized to keratin intermediate filaments. Our work was performed with the goal of identifying the molecule(s) that block Hsp70 activity on keratin filaments during inflammation. A transcriptional screen allowed us to focus on BAG-1, a multi-functional protein that assists Hsp70 in nucleotide exchange but also blocks its activity at higher concentrations. We found the BAG-1 isoform BAG-1M upregulated threefold in human Caco-2 cells following stimulation with tumor necrosis factor receptor α (TNFα) to induce a pro-inflammatory response, and up to sixfold in mouse enterocytes following treatment with dextran sodium sulfate (DSS) to induce colitis. BAG-1M, but no other isoform, was found to co-purify with intermediate filaments and block Hsp70 activity in the keratin fraction but not in the soluble fraction within the range of concentrations found in epithelial cells cultured under control and inflammation conditions. Constitutive expression of BAG-1M decreased levels of phosphorylated aPKC. By contrast, knockdown of BAG-1, blocked the TNFα-induced decrease of phosphorylated aPKC. We conclude that BAG-1M mediates Hsp70 inhibition downstream of NF-κB.

INTRODUCTION

ABC kinases (i.e. PKA, PKB or Akt, and PKC) share a conserved turn motif that enables enzymatic activity when phosphorylated (Newton, 2003). This motif is also a binding domain for the Hsc/Hsp70 chaperones (Gao and Newton, 2006). Binding of these chaperones is crucially important, because ABC kinases expose their turn motif to phosphatases when they are catalytically active. Dephosphorylation of the turn motif results in loss of kinase activity, ubiquitylation and degradation, unless the molecule is refolded by Hsc/Hsp70 and then re-phosphorylated (‘rescued’) (Gould and Newton, 2008). For example, for atypical PKC (isoforms ι/λ and ζ; hereafter referred to as aPKC) inhibition of the rescue mechanism results in a 70–90% decrease of its steady-state levels, together with a reduction of the half-life of the molecule from over 24 hours to 4 hours. The levels of phosphorylated protein generally parallel those of total protein. However, whereas the former are a direct result of Hsp70 chaperoning followed by PDK1-mediated phosphorylation and then autophosphorylation, the total protein reports on the ubiquitylation and proteasomal degradation (Mashukova et al., 2009; Mashukova et al., 2011). For simplicity, the levels of active conformation, phosphorylated turn domain aPKC are referred to as aPKC levels.

The aPKC rescue mechanism in epithelial cells occurs on a keratin intermediate filament scaffold. In fact, Hsp70, Hsp40 and co-chaperones including BAG-1 (Bcl-2-associated athanogene) are known to bind keratin filaments (Izawa et al., 2000; Liao et al., 1995; Naishiro et al., 1999; Perng et al., 1999). Hsp70 binds the head domain of Krt5 (and possibly Krt18) (Planko et al., 2007). Hsp40 binds domain II of the Krt18 rod (Izawa et al., 2000). In epithelial cells, this interaction with intermediate filaments is necessary for chaperoning function, because keratin knockdown prevents the rescue of a subset of Hsp70 client kinases (e.g. aPKC and Akt), whereas other chaperone client proteins are only slightly affected or not affected at all (e.g. Chk1 and PKA). In keratin 8 (Krt8) knockout mice levels of aPKC are severely decreased in cells that lack intermediate filaments, whereas in Krt8-overexpressing enterocytes it is upregulated and delocalized (Mashukova et al., 2009).

In epithelial cells, aPKC is a crucial regulator in the establishment of apico-basal polarity and assembly of tight junctions (Suzuki and Ohno, 2006). In addition, it is involved in the regulation of the innate immunity NF-κB pathway (Forteza et al., 2013). Our laboratory has shown that aPKC is post-translationally downregulated through activation of NF-κB, downstream of tumor necrosis factor receptor (TNFR), in intestinal epithelial cells. This effect was observed in the tissue culture cell line (Caco-2) as well as in an animal model of chemical colitis induced by dextran sodium sulfate (DSS), and affected both total protein levels and the phosphorylated active form of aPKC (Mashukova et al., 2011). Downregulation of aPKC also correlated with inflammation in human tissue samples of patients suffering from inflammatory bowel disease (Wald et al., 2011). These results are clinically important because integrity of the epithelial barrier determined by tight junction permeability is a crucial determinant of intestinal inflammation (Edelblum and Turner, 2009).

The effect of NF-κB on steady-state aPKC levels is mediated by an inhibition of the Hsc/Hsp70-mediated rescue (Mashukova et al., 2011). The molecule(s) mediating such an inhibition remains so-far unknown. Using standard procedures to purify cytoskeletal components, epithelial cells have been fractionated into a detergent-soluble fraction (that comprises membrane and cytosolic proteins), an actin-rich fraction and a keratin intermediate filament fraction. Hsp70 proteins are found in all three fractions. It is remarkable that, whereas most of the chaperone is present in the soluble fraction, the keratin fraction is capable of refolding denatured luciferase (Hsp70 chaperoning activity) when supplemented with ATP (Mashukova et al., 2011). By contrast, the soluble fraction cannot refold or rescue aPKC, unless supplemented with purified native filamentous keratins (Mashukova et al., 2009). Intriguingly, the chaperoning rescue of aPKC associated with intermediate filaments in vitro can be inhibited by adding the soluble fraction from cells treated with tumor necrosis factor α (TNFα) (Mashukova et al., 2011). This result suggested that a soluble (cytosolic) factor is responsible for inhibiting Hsc/Hsp70 chaperoning. In addition, the keratin fraction from TNFα-treated cells is intrinsically blocked to rescue aPKC, suggesting that the inhibitor factor remains bound to the intermediate filaments throughout the isolation process (Mashukova et al., 2011). Multiple pieces of evidence indicate that intermediate filaments act as a scaffold that can bind and, possibly, be in equilibrium with cytosolic chaperones. When the keratins are purified through various cycles of urea solubilization and reassembly by dialysis (Steinert et al., 1982), they do not sustain aPKC refolding (Mashukova et al., 2009). Yet, they regain refolding function when supplemented with the soluble fraction. In these experiments, immunodepletion of Hsp70 in the soluble fraction abrogated the rescue of the filament refolding, showing the role of this particular chaperone (Mashukova et al., 2009). However, it is likely that other chaperones, such as Hsp40, and co-chaperones also associate to the filament compartment from the cytosol.

The identification of the Hsp70 inhibitor in the intermediate filaments following activation of NF-κB was the reason to undertake this work. There was no detectable change in the levels of Hsp70 or Hsp40 during pro-inflammatory TNFα signaling (Mashukova et al., 2011). Accordingly, we hypothesize that the molecule that mediates NF-κB-dependent inhibition of aPKC rescue is one of the small chaperones or co-chaperones that normally assist the Hsp70 or Hsp90-mediated mechanisms.

RESULTS

BAG-1 is upregulated in epithelial cells during activation of pro-inflammatory signaling

To identify molecules that modulate chaperoning in intestinal epithelial cells we first conducted a transcriptome PCR screen focusing in chaperones and co-chaperones. We compared mRNAs from human colon carcinoma (Caco-2) control cells and Caco-2 cells stimulated with TNFα at the stage of full epithelial differentiation and polarization (15 days of culture on filters) (Grasset et al., 1984). Out of 89 genes tested, 11 genes were upregulated threefold or more, whereas one was downregulated more than threefold while still showing robust detection in quantitative PCR (qPCR) (threshold cycle was <30 for both control and test samples). (Table 1, see full data set in supplementary material Table S1). One of the upregulated genes was a transcription factor (ATF6), which we ruled out because the inhibition of Hsp70 chaperoning can be mimicked by soluble cytosolic proteins in the absence of transcription. Five of the upregulated genes were members of the Hsp40 family and one was Hsp70 itself. None of these proteins are known to block their own chaperoning activity. That left four mRNAs of small chaperones that were upregulated 3- to 11-fold in the PCR screen: αB-crystallin (CRYAB), heat shock protein β-8 (HSPB8, hereafter referred to as Hsp22), heat shock protein β-1 (HSPB1, hereafter referred to as Hsp27), and BAG family molecular chaperone regulator 1 (BAG1, hereafter referred to as BAG-1) as possible candidates.

Table 1. Summary of quantitative PCR results from screens comparing transcriptional changes of chaperone and cochaperone genes in Caco-2 cells stimulated with TNFα.

The values were summarized from the full screen report (supplementary material Table S1). The hits were selected according to the following criteria: (i) Fold change >3 and (ii) threshold cycle < 30 for both control and test samples.

αB-crystallin is known to act synergistically with Hsp70 for protein refolding (Peschek et al., 2013; Wang and Spector, 2000). It is often upregulated in the gastrointestinal tract and in muscle (Nefti et al., 2005; Paulsen et al., 2009) but, to our knowledge, is not known to inhibit Hsp70 chaperoning. The upregulation of Hsp22 could not be validated by qPCR (data not shown). Nor could we independently validate the upregulation of Hsp27 at protein level (Fig. 1A). Accordingly, we dismissed those genes as false positives in the screen. BAG-1, however, is a known inhibitor of Hsp70 activity (Gassler et al., 2001; Nollen et al., 2001). Four isoforms can be produced by BAG1 through alternative initiation, that share a C-terminal sequence: BAG-1L (large), BAG-1M (medium), BAG-1 and BAG-1S (small) (Alberti et al., 2003). In Caco-2 cells, we found trace amounts of BAG-1L (Fig. 1A). The protein expression of BAG-1M and BAG-1S was upregulated threefold following stimulation with TNFα (Fig. 1A,B), which is fully consistent with the mRNA upregulation following the same treatment (Table 1). The effectiveness of treatment with TNFα and activation of NF-κB was monitored through the decrease in IκB steady-state levels (Fig. 1A,B).

Fig. 1.

Inflammation upregulates BAG-1 expression in intestinal epithelial cells. (A) Caco-2 cells were grown on Transwell filters and some cultures were incubated with 20 ng/ml TNFα from the basolateral side, while others were left untreated (contr). SDS extracts of the cells were analyzed in immunoblot with the antibodies listed on the left. BAG-1M and BAG-1S were recognized with the same antibody, but the exposure was shorter for the small isoform. (B) Average and STD of the relative values of BAG-1 and IκB were normalized to the control (untreated) values. n = 5; *P<0.05; **P<0.001. (C) C57BL/6 mice received 3% DSS in the drinking water (4–6) (or not, control, 1–3) until the disease activity index reached 3, upon which the animals were sacrificed, and epithelial cells harvested from the small intestine were analyzed by immunoblot (35 µg total protein/lane). (D) Ratios of BAG-1 signal relative to GAPDH in the same blots from 7 animals per condition were normalized to the average of the control signals. *P<0.05.

Because BAG-1 is upregulated in cancer cells (Sharp et al., 2004), we sought to confirm whether the upregulation of BAG-1 following induction of a pro-inflammatory response is also observed in normal intestinal epithelial cells. To that end, we treated normal mice with DSS, an extensively used method to induce chemical colitis in animals, which results in overactivation of the NF-κB pathway in epithelial cells (DSS colitis) (Bhattacharyya et al., 2009). Although DSS colitis effects are usually studied in the colon, we used small intestine for the following reasons. (1) There are minimal morphological changes in the small intestine following DSS treatment, as compared with the colon (Yazbeck et al., 2011). (2) Levels of TNFα in the small intestine are still increased following DSS stimulation, which is the goal of the experiment (Kawauchi et al., 2014). (3) There is less interference of bacterial products that cause cellular infiltration in the small intestine following treatment with DSS (Hans et al., 2000), which results in less potential contamination with non-epithelial inflammatory cells. (4) When using the small intestine, cell and protein yields are bigger and more consistent in control vs DSS-treated animals, which facilitated the detection of BAG-1, especially in controls. The latter was crucial, because BAG-1 was barely detectable in control cells of enriched enterocytes from small intestine. However, in DSS-treated animals, we observed a sixfold upregulation of BAG-1M (Fig. 1C,D) in similar preparations. We conclude that in the colitis animal model, BAG-1M is also upregulated, but to a larger extent than in Caco-2 cells.

BAG-1M co-purifies with the keratin intermediate filament fraction following stimulation with TNFα

On the basis of previous publications, the Hsp70 inhibitor – which is downstream of activated NF-κB – is expected to be soluble but capable of binding to and co-purifying with intermediate filaments. To test whether BAG-1 fits this prediction we used a standard method to purify keratin intermediate filaments originally devised by Steinert and co-workers (Steinert et al., 1982), and adapted it to intestinal cells (Mashukova et al., 2009). This method separates the cells into a Triton-X-100-soluble fraction that comprises all cytosolic and most membrane proteins, an actin-rich pellet and a keratin pellet. As expected, BAG-1 was present mostly in the soluble fraction. However, bands of modest signal intensity were also observed in the actin-rich pellet and the keratin pellet fractions (Fig. 2A). This was not surprising because BAG-1 has been reported to colocalize with actin and keratin filaments (Naishiro et al., 1999). BAG-1M and BAG-1S co-purified with the actin filaments, but the intensity of the bands did not change in response to stimulation with TNFα (Fig. 2B,C), despite the fact that the total cellular content increased threefold (Fig. 1). In contrast, only BAG-1M was detected in the keratin pellet fraction and was found to increase twofold following stimulation with TNFα (Fig. 2B,C). The stability of BAG-1M binding to the keratin fraction makes it a suitable candidate for the role of Hsp70 inhibitor following activation of pro-inflammatory signaling.

Fig. 2.

BAG-1M co-purifies with the intermediate filament fraction in a TNFα-dependent manner. (A) Keratin intermediate filaments (keratin pellet, kP) were purified by a standard procedure and compared with the Triton-X100-soluble fraction (S) and the rest of the Triton-insoluble cytoskeletal fraction (actin-rich pellet, aP). Identical amounts of protein from each fraction (40 µg/lane) were blotted and analyzed by using Ponceau S Red staining (molecular mass standards in kDa), and by immunoblot with the antibodies against the proteins given on the right-hand side of the blot. (B) BAG-1 was analyzed by immunoblotting of the actin-rich pellet (aP) and keratin pellet (kP), in the presence (+) and absence (−) of stimulation with TNFα. Actin and keratin 8 were used as loading controls for aP and kP respectively. (C) Relative BAG-1M signal/loading control, normalized to the values for untreated cultures. n = 3, *P<0.05.

BAG-1M but not BAG-1S inhibits Hsp70 chaperoning activity in the keratin filament fraction

In Caco-2 intestinal cells, the keratin pellet fraction is sufficient to refold aPKC and requires only to be supplemented with the appropriate kinase (PDK1) to complete the rephosphorylation of aPKC (Mashukova et al., 2012). Hsp70, however, is present in both the soluble and the keratin pellet fractions. Both fractions are independently sufficient to sustain luciferase refolding in the presence of ATP (Mashukova et al., 2011), a landmark of Hsp70 chaperoning activity. To study the possible effect of BAG-1M inhibiting this activity, it was first essential to determine the endogenous BAG-1M concentrations in the cytosol. To this end, we titrated, by immunoblot, endogenous BAG-1M in confluent, differentiated, non-stimulated Caco-2 cells and compared its amount with known amounts of purified recombinant BAG-1M or BAG-1S. By using densitometry we determined the amount of endogenous BAG-1M to be ∼5 ng/35 µg of total cell protein (Fig. 3A). Converting that to the corresponding cell volume for 200 mg/ml (total protein per cell volume), it represents ∼1 µM BAG-1M. This basal concentration is relatively low as compared with other cells (2 µM) (Nollen et al., 2001). In carcinomas, BAG-1 overexpression of 10- to 20-fold the basal levels, even under gene amplification, has been reported (Mäki et al., 2007). Indeed, it was surprising that Caco-2 cells, being derived from a carcinoma, express moderate levels of BAG-1, a fact that makes them useful in order to study BAG-1 functions representative of non-cancer cells.

Fig. 3.

BAG-1M levels expressed during activation of pro-inflammatory signaling inhibit Hsp70 in the keratin fraction. (A) BAG-1M signal in SDS Caco-2 cell extracts (total protein/lane in µg) was compared with known amounts of recombinant BAG-1M (values expressed in ng) by immunoblot. (B) Hsp70-mediated refolding of luciferase was measured by rescue of luminescence in the presence of ATP and either the Caco-2 Triton-X-100-soluble fraction (S) or the intermediate filament fraction (kP), as a function of time. Luminescence data are presented as the maximum luminescence before luciferase denaturation in percent. Control assay (white bars) contained only the endogenous BAG-1. Recombinant BAG-1M was added to a final concentration of 1 µM (light grey bars), 2 µM (dark grey bars) or 6 µM (black bars), mimicking the range of upregulation observed in the cytosol during activation of the pro-inflammatory signaling (Fig. 1). n = 3, *P<0.05, **P<0.001. (C) Using the averages in B, inhibition in percent was calculated by subtracting the corresponding 0 hour background luminescence first. Then, the inhibition was obtained as the result of 1−(value at×BAG-1M/control value)×100. ** indicate values obtained from averages statistically significant from control (from B).

To analyze the role of BAG-1M in the cytosolic compartment and the intermediate filament scaffold, the soluble and keratin pellet cell fractions were supplemented with 1 µM, 2 µM or 6 µM recombinant BAG-1M to mimic the two- or threefold upregulation induced by TNFα (Fig. 1). For the intermediate filament fraction, ‘concentration’ of BAG-1M is difficult to estimate. However, the data in Fig. 1 and Fig. 2 indicates a proportionality between the total cellular content of BAG-1 and the content in the intermediate filaments. We reasoned that the cytosol is the normal environment for the intermediate filaments and that, accordingly, the filament BAG-1M concentration must be in equilibrium with the cytosolic concentration. Accordingly, the latter was used to mimic the cellular environment for isolated intermediate filaments. At these concentrations, the nominally basal 1 µM BAG-1M caused modest, non-statistically significant, decreases of luciferase refolding in the Triton X-100 soluble fraction, whereas after 4 hours 2 µM and 6 µM caused 49% and 70% inhibition, respectively, in the chaperoning activity. In contrast, all three BAG-1M concentrations resulted in significant inhibition of Hsp70 chaperoning activity in the keratin intermediate filament fraction at 1 and 4 hours. In fact, the rate of refolding in the keratin pellet fraction was almost flat between 1 and 4 hours of incubation as compared with the untreated control at 6 µM Bag-1M. Furthermore, the inhibition of chaperoning induced by 6 µM Bag-1M at 1 and 4 hours, was significantly greater in the keratin pellet than in the soluble fraction (Fig. 3B).

A similar set of experiments was conducted by using recombinant BAG-1S. The total cellular concentration of BAG-1S was calculated as described for Fig. 3 and found to be 0.85 µM in non-stimulated Caco-2 cells (Fig. 4A). Accordingly, the same range of concentration as in the case of BAG-1M was used. In this case, a modest but significant inhibition of luciferase refolding was observed only in the soluble fraction at 6 µM BAG-1S. By contrast, BAG-1S completely failed to inhibit Hsp70 chaperoning activity in the intermediate filament fraction (Fig. 4B).

Fig. 4.

The amount of BAG-1S expressed during activation of pro-inflammatory signaling fails to block Hsp70 function in the keratin fraction. (A) BAG-1S signal in SDS Caco-2 cell extracts (total protein/lane in µg) was compared with known amounts of recombinant BAG-1S (values expressed in ng) by immunoblot. (B) Hsp70-mediated refolding of luciferase was measured by rescue of luminescence as in Fig. 3. Luminescence data are presented as maximum luminescence in percent before luciferase chemical denaturation. Control assay (untreated, white bars) contained only the endogenous BAG-1. Recombinant BAG-1S was added to a final concentration of 1 µM (light gray bars), 2 µM (dark gray bars) or 6 µM (black bars) final concentration, mimicking the range of upregulation observed during activation of pro-inflammatory signaling (Fig. 1). n = 3, *P<0.05, **P<0.001.

We conclude that the Hsp70 chaperoning complex associated with the keratin intermediate filaments is strongly inhibited by cytosolic BAG-1M concentrations observed during pro-inflammatory signaling. Importantly, Hsp70 that co-purifies with intermediate filaments is more sensitive to BAG-1M inhibition than the cytosolic soluble Hsp70 and insensitive to BAG-1S in the same range of concentration.

BAG-1M expressed at concentrations similar to those induced by TNFα is sufficient to decrease levels of active phosphorylated turn domain aPKC in vivo

To determine whether the results observed in vitro also apply to living cells, we sought to constitutively express BAG-1M by using lentiviral transduction followed by G-418 selection. The level of expression in cultures varied from 2- to 14-fold the endogenous levels, depending on the transduction and the number of passages under G-418 selection. Accordingly, we selected cultures expressing less than sevenfold the endogenous levels of BAG-1M (average of expression in Fig. 1B). Other isoforms of BAG-1 were not expressed by the vector (Fig. 5A,B). In these cultures, we determined the levels of aPKC in its normal, active configuration (determined by an antibody against the phosphorylated turn domain that recognizes PKCι and ζ). The rationale to use an antibody against the phosphorylated turn domain is twofold. (1) In the sequence of events, refolding leads to phosphorylation of the activation domain (by PDK1), leading to phosphorylation of the turn domain (autophosphorylation); pT555 antibody staining is a stable reliable readout of the first step. (2) This domain is identical in both aPKC isoforms. Finally, it should be noted that total protein may vary depending on the extent of ubiquitylation and/or proteasomal degradation. Because BAG-1 can potentially modify this other branch of the pathway leading to degradation (Demand et al., 2001), we did not use total protein as a measure of refolding, and will refer to ‘levels of active phosphorylated turn domain aPKC’ as ‘aPKC levels’ hereafter.

Fig. 5.

Expression of BAG-1M downregulates active aPKC. (A) Caco-2 cells were transduced with lentiviral particles expressing either GFP or BAG-1M and selected in G-418. BAG-1 was probed by immunoblot (L, M, S indicate the position of the large, medium and small isoforms, respectively). (B) Quantification of the BAG-1M signal relative to loading control normalized to the average expression in GFP-transduced cells. n = 3, *P<0.01. (C) Expression of aPKC, Akt and Chk1 in the same experiments was analyzed by immunoblot. (D) Quantification of the signals in experiments like the one shown in panel C, normalized to the average signal in GFP-expressing cells (black bars). n = 3, *P<0.02. (E) Immunofluorescence staining by using antibody pT555 PKCι (red) in GFP- or BAG-1-expressing cells is shown as a maximum projection of all confocal sections showing lateral (tight junction region, TJ) localization (∼ten per field) for the red signal and the maximum projection of all confocal sections for DAPI (blue) signal. (F) Quantification of pixel values (BKG, background subtracted) in the TJ region and in the apical cytoplasm shown for cells expressing GFP (black bars) or BAG-1M (white bars). n = 3 (81 cells for each culture), P<0.02. (G) Immunofluorescence staining by using antibody pT555 PKCι against aPKC (red) and ZO-1 (green) are shown as single confocal sections at apical region. Arrows point at regions of weak ZO-1 signal. Notice that TCA fixation abrogates GFP fluorescence. Scale bars: 10 µm.

The active aPKC levels were significantly decreased by expression of BAG-1M (Fig. 5C,D). As a reference, we used two other Hsp70 clients, Chk1 (Arlander et al., 2006) and Akt (Doong et al., 2003). Unlike active conformation aPKC and Akt (pT310) levels, which are dependent on the integrity of intermediate filaments, Chk1 has been found to be insensitive to keratin knockdown in Caco-2 cells (Mashukova et al., 2009). Therefore, Chk1 is considered to be independent of the keratin-based chaperoning system. As expected, the levels of aPKC and Akt in their active conformation signficantly decreased in BAG-1M expressing cells. By contrast, Chk1 levels modestly but significantly increased in the same cells (Fig. 5C,D). This result is consistent with the data in Fig. 3, indicating that, at levels of BAG-1M expression similar to those induced by TNFα, Hsp70 chaperoning is inhibited mostly in the intermediate filaments but not in the cytosol.

To independently corroborate the downregulation of active-conformation aPKC by BAG-1M expression, we used the same cultures in immunofluorescence experiments. As expected, aPKC was found mostly in the vicinity of tight junctions (at the apical end of the lateral domain) as well as in the apical cytoplasm in cells expressing GFP. BAG-1M expression decreased the aPKC signal at both locations by approximately the same proportion (shown by immunoblot; Fig. 5E,F). By using immunofluorescence, pT555 aPKC and ZO-1 were found to colocalize in similar experiments. Whereas ZO-1 was still present in some cells depleted of active aPKC, the ZO-1 signal was much weaker in many cells (Fig. 5G, arrows). We conclude that BAG-1M expression depletes the physiological, highly localized subcellular compartments of aPKC (i.e. along the tight junction and in the apical domain) in the active conformation, exactly like TNFα signaling or chemical colitis do in culture or animal models, respectively (Mashukova et al., 2011).

BAG-1 is necessary for the effect of TNFα on post-translational downregulation of aPKC

Next, we wanted to determine whether endogenous BAG-1 is responsible for the decrease in aPKC levels in intestinal epithelial cells upon stimulation with TNFα. To that end we knocked down BAG-1 in Caco-2 cells. We used two different short hairpin RNA (shRNA) sequences that were delivered and constitutively expressed by lentiviral vectors to minimize the chances of off-target effects. Both shRNAs knocked down BAG-1M levels to a similar extent (to below 50%) (Fig. 6A). In both cases, BAG-1M was still upregulated following stimulation with TNFα. Importantly, however, both shRNA-mediated knockdowns kept the levels of BAG-1M in TNFα-treated cells at or below the levels of untreated cells (Fig. 6A).

Fig. 6.

BAG-1 knockdown prevents downregulation of active aPKC following stimulation with TNFα. (A) Constitutive expression of two different anti-BAG-1 shRNAs by lentiviral transduction decreased BAG-1S and BAG-1M signals and prevented BAG-1 from reaching levels above control following stimulation with TNFα. Cells expressing a scrambled shRNA (scr) under the same antibiotic resistance were used as a control. (B) Caco-2 cells expressing scrambled (scr) or specific anti-BAG-1 shRNAs were incubated in the presence or absence of TNFα. Levels of phosphorylated aPKC (p-aPKC) decreased in scrambled shRNA cells stimulated with TNFα, but not in BAG-1-knockdown cells. (C) Quantification of aPKC signal in experiments like the one shown in panel C, using shRNA1, normalized to the average of the values in control (untreated, scrambled shRNA). n = 4, *P<0.01. (D) Quantification of aPKC signal independent experiments like the one shown in panel C, using shRNA2, normalized to the average of the values in control (untreated, scrambled shRNA). n = 3, *P<0.01.

As observed before in non-transduced cells, TNFα induced a significant decrease in active aPKC in cells that expressed a scrambled shRNA (Fig. 6B-D). This effect of TNFα was abolished by both anti-BAG-1 shRNAs (Fig. 6C,D). The levels of aPKC were not significantly different from control when using shRNA1 or 2 (Fig. 6C,D). In either case, there was no significant decrease in aPKC levels upon incubation with TNFα (Fig. 6C,D). It is important to note that downregulation of BAG-1M with shRNA resulted in greater variability of phosphorylated aPKC levels, as untreated knockdown cells often showed levels dissimilar to those of the scrambled-transduced cells. Importantly, however, in all experiments BAG-1M knockdown abolished the effect of TNFα. We conclude that BAG-1M is an essential component of the mechanism that post-translationally downregulates aPKC under the control of NF-κB.

DISCUSSION

BAG-1 was originally discovered as an inhibitor of apoptosis (Takayama et al., 1995). Work from many laboratories unveiled a number of other functions, including nucleotide exchange aiding the Hsp70 chaperoning cycle, direct binding and chaperoning of several proteins (Alberti et al., 2003; Mendes et al., 2012), and control of ubiquitylation and proteasomal degradation (Elliott et al., 2007). In addition, BAG-1L, the only isoform that contains a nuclear localization signal, has transcriptional control functions as well (Gehring, 2009). In the last few years, many publications have shown the role of BAG-1 in cancer and its potential use as a prognostic indicator (Ni et al., 2013; Wood et al., 2009). However, little is known regarding a role in innate immunity.

Here, we have shown that expression of BAG-1 is increased following TNFα stimulation of Caco-2 cells in culture and during inflammation in the DSS animal model. We also presented evidence that BAG-1M inhibits Hsp/Hsc70 chaperoning activity in vitro within the same range of cytosolic concentrations as in vivo, preferentially in the keratin intermediate-filament scafold. Also, we confirmed the hypothesis that BAG-1 mediates the TNFα-dependent post-translational downregulation of aPKC. Aside from the implication for the recently discovered cross-talk between aPKC signaling and NF-κB activity (pathway summarized in supplementary material Fig. S1) (Forteza et al., 2013; Mashukova et al., 2011), our results highlight functional differences between cytosolic Hsc/Hsp70 (which represents the bulk of these chaperones) and a small but functionally relevant pool of keratin-bound Hsc/Hsp70.

BAG-1 is expressed within a broad range of concentrations in different cell types (Mäki et al., 2007; Nollen et al., 2000). Accordingly, it is likely that each one of its multiple functions is exerted at a certain optimal concentration and, thus, is specific to cells expressing that level of BAG-1. In our current work, we focused on the range of concentrations of endogenous BAG-1 induced following stimulation Caco-2 cells with TNFα. Our calculation of total cellular concentration as mass of BAG-1M per cell volume is straightforward, has been used by others (Nollen et al., 2000) and allowed us to compare Caco-2 cells with other cell types. It was surprising that our cells, derived from a human carcinoma, display levels of BAG-1M similar to those of non-cancer cells. We used these BAG-1M concentrations in the in vitro assays. It could be argued that the true cytosolic concentration must be corrected for nuclear volume (BAG-1M is excluded) as well as other membranous compartments that may also exclude BAG-1M. These relative volumes were calculated recently for Caco-2 cells by using a dilution of specific fluorophores (Tan et al., 2012). The results indicate that the cytosolic volume is approximately half of the total cellular volume. Therefore, the results shown in Fig. 3 should be interpreted by using 1.7–2 µM as the basal (non-stimulated concentration) and 4–6 µM as the concentrations that mimic stimulation with TNFα. In that scenario, Hsp70 chaperoning on intermediate filaments is inhibited roughly 50%, but in the steep part of the curve (Fig. 3C), indicating that it is highly sensitive to small variations of BAG-1M. The soluble fraction, however, is still active at the highest levels of TNFα stimulation detected here. On this basis, it is tempting to speculate that Hsp70 chaperoning on intermediate filaments is operational in cells with low physiological levels of BAG-1M, but completely disabled in cells with high levels, such as cancer cells. Likewise, one would predict that upregulation of BAG-1S affects soluble Hsp70 chaperoning but not Hsp70 associated with intermediate filaments.

Whereas the soluble Hsp70 complex has been extensively studied, the structural features of the Hsc/Hsp70 complex on the keratin scaffold remain unknown. It has been recently reported that Hsp70 functional differences arise from protein partners in multi-protein complexes (Rauch and Gestwicki, 2013). It is noteworthy that the Hsp70-intermediate filament complex has a different client specificity than soluble Hsp70 (Mashukova et al., 2009). In this work, we have shown that it has specificity for the BAG-1M isoform and that the same isoform specifically inhibits the Hsp70/keratin complex at a concentration that does not significantly affect the soluble complex. Because all BAG-1 isoforms share the Hsp70-binding domain in their common C-terminal sequence, we must conclude that the binding of BAG-1M to keratin intermediate filaments is mediated through an independent interaction, possibly through a domain within the BAG-1M N-terminal region, between amino acid 1 and 55, that is, the amino acid sequence of BAG-1M that is not present in BAG-1S. Whereas the 230 amino acid BAG-1 isoform (with a size between the medium and small isoforms) was seldom expressed in Caco-2 cells, it was expressed in one experiment and co-purified with the keratin fraction. Accordingly, it is possible that the minimum sequence to mediate BAG-1 to intermediate filament binding lies within amino acids 44 and 55. However, we speculate that BAG-1L was not found bound to the keratin fraction because its nuclear localization signal removes it from the cytoplasm, making it unavailable to intermediate-filament binding. The BAG-1M binding partner in the keratin filaments is unknown. We and others have demonstrated the presence of at least 20 keratin-binding minor components in this fraction, including Hsc70 and Hsp40 proteins. Therefore, the nature of this interaction will require further studies. Importantly, the increase in cytosolic BAG-1M results in a corresponding increase of keratin-bound BAG-1M. Consequently, and unlike the actin-associated BAG-1 (Fig. 2), the binding sites in the keratin filaments are not saturated and seem to have an association rate within the normal level of BAG-1M concentrations in these cells. This, in turn, highlights a possible physiological role that controls intermediate-filament-based Hsp70 chaperones. Furthermore, this distinct chaperoning function of intermediate-filament-bound Hsp70 points to a specific non-mechanical function of the keratin scaffold. It may explain keratin-dependent cellular functions, such as the anti-apoptotic role of intermediate filaments, and functions in secretion and protection against chemical injury (Alam et al., 2013; Oriolo et al., 2007; Toivola et al., 2010). Finally, the results of the chaperoning assay imply that intermediate filament-associated chaperoning is functional in normal cells but may be disabled in cells that express higher levels of BAG-1M, such as under inflammatory conditions or in cancer.

Materials and Methods

Materials

Recombinant BAG-1M was obtained from Enzo Life Sciences (ADI-APR-400-0050) and BAG-1S was obtained from Prospec (PRO-817). TNFα was obtained from R&D. The antibodies used in this study were as follows: BAG-1 C-terminal domain (Abcam, ab32109). pT555 PKCι (also recognizes pT560 PKCζ, referred to as p-aPKC, GeneTex, GTX25813), pT308 Akt (Cell Signaling, 9275S), Chk1 (Abcam, ab47574), Hsp27 (Abcam, ab2790), IκB (Cell Signaling. 9246S), GAPDH (Sigma, G9545), tubulin (Sigma, T6199), actin (MP Biomed, 691001), keratin 8 (TROMA-1, Developmental Studies Hybridoma Bank), ZO-1 (InVitrogen, 339100).

Cell culture

Cells of the human intestinal Caco-2 cell line were originally obtained from ATCC and kept as described previously (Salas et al., 1997) except that medium was supplemented with 5% FCS in the absence of antibiotics. For experiments, the cells were usually cultured in 3 µm-pore Transwell filters (Corning) and kept confluent for 10 days. For inflammatory stimulation and NF-κB activation, the cells were incubated for 5 days in serum-free medium (Jumarie et al., 1996) (DMEM/F12 supplemented with 1.72 µM insulin, 68 µM transferrin and 38 nM selenite; all Gibco ITS-G). Stimulation was performed by supplementing the medium in the basolateral chamber with 20 ng/ml TNFα in the same serum-free medium for 18 hours.

Vectors

The lentiviral vectors expressing BAG-1M or GFP under G-418 selection were obtained from GeneCopoeia (EX-Z0205-Lv151). shRNA-expressing lentiviral vectors (anti-BAG-1 and scrambled) under puromycin selection were obtained from Sigma; shRNA1: 5′-TRCN0000007298, CCGGTGTCACCCACAGCAATGAGAACTCGAGTTCTCATTGCTGTGGGTGACATTTTT-3′; and shRNA2: 5′-TRCN0000007297, CCGGGAAACACCGTTGTCAGCACTTCTCGAGAAGTGCTGACAACGGTGTTTCTTTTT-3′. For transduction selection, 0.8 mg/ml G-418 or 5 µg/ml puromycin were added to the culture medium 24 hours after lentiviral transduction and maintained continuously thereafter.

Transcriptional screen

Heat Shock Proteins & Chaperones PCR Array Human Gene Expression (SABiosciences, PAHS-076Z) was used to obtain the data shown in Table 1. The screen was performed as instructed by the manufacturer.

Cell extraction

Triton X-100 extraction and cell fractionation in soluble, actin pellet and keratin pellet fractions was described before (Mashukova et al., 2009). To minimize actin contamination of the keratin pellet fraction, we supplemented the Triton X-100 extraction buffer with 1 mM ATP to further actin-myosin disassembly during extraction. For comparison among fractions by immunoblot, equivalent amounts of protein from each fraction were loaded on the gels. The ratio of total cell protein/cell volume (200 mg/ml) was obtained previously (Milo, 2013). We reconfirmed the value for Caco-2 cells by weighting cell pellets, measuring total protein and determining the density by flotation in sucrose gradients.

Hsp/Hsc70 activity assay

Hsp70 chaperoning assay was performed as described (Lu and Cyr, 1998) using the Caco-2 Triton X-100 soluble and insoluble fractions described above (Mashukova et al., 2011) and luminescence measured with Bright-Glo^TM Luciferase Assay System (Promega). Briefly, luciferase (13 mg/ml) is diluted 42-fold into denaturation buffer (25 mM HEPES, pH 7.4, 50 mM KCl, 5 mM MgCl₂, 6 M guanidinum-HCl and 5 mM DDT). This denaturation reaction is done for 1 hour at 25°C. 1-µl aliquots are then removed from denaturation and mixed with 125 µl of refolding buffer (25 mm HEPES, pH 7.4, 50 mm KCl, 5 mm MgCl₂, 1 mm ATP) that is supplemented with cellular fractions and incubated at 25°C. Aliquots of 1 µl are removed from the folding reactions at 0 hours 1 hour and 4 hours, as well as from equally diluted original non-denatured luciferase solution, and luciferase activity is measured.

Animal DSS colitis

Three- to five-month-old C57BL/6 mice received 3% dextran sodium sulfate (DSS) in the drinking water to induce colitis, which, in turn, activates NF-κB. The disease activity index (Cooper et al., 1993) was calculated daily from weight loss, consistency of the stools and fecal blood. DSS-treated animals were euthanized at a disease activity index of 3, together with untreated controls. The animal handling protocol was approved by the local IACUC and was consistent with PHS Guide for the Care and Use of Laboratory Animals.

Procedures for harvesting and enriching mouse small intestine (jejunum and ileum) enterocytes have been described before (McNicholas et al., 1994) with the following changes: Instead of everting the small intestine, it was filled with PBS supplemented with 20 mM EDTA and kept on ice for 20 minutes, with frequent gentle squeezing of the lumen to mechanically disrupt villi. The contents were then collected in an Eppendorf tube, spun (2 seconds, 5000 g), and resuspended in the same solution supplemented with 5 mM DTT. The cells were washed by 5 cycles of resuspension in the same solution, and the last pellet was frozen in PBS supplemented with antiproteases (Sigma). Epithelial enrichment was assessed from keratin/vimentin signals through immunoblotting. Quantification of immunoblot results was performed by using chemiluminescence densitometry using a VersaDoc (BioRad) gel scanner and Quantimet software.

Immunofluorescence, image analysis and quantification of data

Cell cultures grown on filters confluent for 10 days were fixed in 10% trichloroacetic acid (TCA) for phosphoepitope preservation (Hayashi et al., 1999). Fluorescence images were acquired using a Leica TCS SP5 confocal microscope using a 63×(1.4 NA) objective. Images were collected at 1024×1024 resolution at 0.5 µm intervals, thus generating a 3D confocal stack. For quantification of aPKC signals in the apical pole of the cells, we determined the top and bottom confocal sections showing signal at the cell–cell contact and obtained a maximum projection of all the confocal sections in that part of the stack using Leica LAS software. In those projection images, regions of interest (ROIs) were randomly selected: (i) on cell–cell contact images (tight junction ROIs) or (ii) on the cytoplasm between cell–cell contacts, the apical cytoplasm ROIs. Background fluorescence was calculated from cytoplasm in confocal images below the nucleus and subtracted from average pixel values of tight junction ROIs or apical cytoplasm ROIs. For statistical purposes, the resulting values from one apical cytoplasm ROI and one tight junction ROI per cell were used for cells in different images.