Transcription of pro-inflammatory cytokines is induced early by high glucose concentrations in rMC1
We preliminarily noticed that immunostaining of the 17 kDa IL-1β fragment (i.e., the biologically active species) was very robust in whole lysates of rMC1 challenged with 25 mmol/L glucose (referred to as high glucose throughout the text) for up to 2 h (Fig. 1A). The intensity of cytokine immunostaining was significantly lower in lysates of cells stimulated with 5 mmol/L glucose or mannitol, used as hyperosmolar control (referred to as “controls”). Conversely, immunostaining of IL-12 (Fig. 1A) and IL-10 (undetectable, data not shown), which is not regulated by NF-kB, was unaltered by treatments over the same time-interval.
To address whether the IL-1β increase reflected the transcriptional upregulation and not a post-synthetic mechanism (e.g., increased processing of pre-IL-1β or impaired secretion), a RT-PCR analysis of rMC1 challenged with 25 mmol/L glucose and controls for 40 min was undertaken. Compared to control cells, cells challenged with high glucose displayed a ~ twofold increase of IL-1β transcript (Fig. 1B). Thus, additional cytokines linked to NF-kB activation were enrolled in the study: IL-8 (~ fivefold increase) and MCP-1 (~ threefold increase) turned out to be induced to the greatest extent in high glucose challenged cells. Conversely, transcription of IL-6 and TNFα was not induced (Fig. 1B).
An atypical NF-kB pathway is activated within minutes after rMC1 challenge with high glucose
To investigate whether high glucose may actually trigger an early activation of NF-kB, a time-course analysis of IkBα phosphorylation at Serine 32 [pIkBα(Ser32)] in whole cell lysates of rMC1 challenged for 10, 20 and 40 min with 25 mmol/L glucose and low glucose or mannitol was performed by Wb [11,12,13]. Compared to control cells, pIkBα(Ser32) immunostaining dropped after 10 min of stimulation with high-glucose, and was persistently decreased within the time-interval analysed (Fig. 2A). Immunostaining of unphosphorylated IkBα dropped as well, but, with respect to pIkBα(Ser32), it showed a 10 min delay: in fact, compared to control cells, the greatest decrease of total IkBα was documented 20 min after high glucose delivery (Fig. 2A; Additional file 1: Fig. S1). Thus, the pIkBα(Ser32)/IkBα ratio, usually very high during canonical NF-kB signalling, was markedly decreased at 10 min of high glucose challenge, but unaltered at 20 and 40 min (Fig. 2A).
To rule out the existence of unprecedented IkBα phospho-sites, a phos-Tag analysis was set up. Separation of lysates from control and high glucose treated cells allowed to detect a band pair compatible with unphosphorylated and phosphorylated IkBα (see lower and upper band in overexposed blot, Fig. 2B). However, high glucose did not cause any obvious increase of delayed-mobility species at any time-point investigated (Fig. 2B).
To verify whether pIkBα(Ser32) and IkBα drop was followed by nuclear translocation of p65-p50, the nuclear and cytosolic fractions of rMC1 stimulated with 25 mmol/L glucose and controls for 30 min were isolated and analysed by Wb. Immunostaining of p65 and p50 highlighted a significant increase of the two proteins in the nuclear fraction of high glucose vs control cells, mirrored by a drop in the corresponding cytosolic fraction (Fig. 2C). Lamin-D2 and β-tubulin immunostaining confirmed the identity of the two fractions and a negligible reciprocal contamination (Fig. 2C).
To further validate NF-kB activation the DNA binding activity of p65 was assayed in the nuclear extracts of cells challenged with high glucose and controls for 30 min by using a commercially available ELISA-immunoassay (Fig. 2D). The NF-kB DNA binding ability, measured by the absorbance at 470 nm (O.D.), was about twofold higher in high glucose than in control cells nuclear extracts (Fig. 2D).
A second cycle of canonical NF-kB activation is observed at later time points in rMC1 exposed to high glucose
Whole cell lysates were collected from rMC1 challenged with high glucose and low glucose or mannitol for 1, 2 and 3 h to figure out whether cytokines, once released, induced a canonical NF-kB signalling following a pro-inflammatory pathway. Compared to control cells, immunodetection of pIkBα(Ser32) progressively increased 2-3 h after high glucose delivery, whilst IkBα dropped significantly (Fig. 3). Hence, this time, the pIkBα(Ser32)/IkBα ratio was greater in high glucose than in control cells.
To confirm the occurrence of a canonical NF-kB signalling, filters were probed for unphosphorylated and phosphorylated IKKβ [at Serine 180, referred to as pIKKβ(Ser180)], and unphosphorylated and phosphorylated p65 (at Serine 536, referred to as p65(Ser536] (Fig. 3). After 2–3 h of stimulation, pIKKβ(Ser180) turned out to be detectable, in the absence of any variation of total IKKβ. Similarly, phosphorylation of p65(Ser536) significantly increased after 2–3 h of high glucose stimulation with respect to control cells. Also in this case, no modulation of the unphosphorylated p65 was observed (Fig. 3). Thus, the p65(Ser536)/p65 ratio was increased in high glucose treated (in particular at 2 h) compared to control cells.
Proteasome bulk activity is stimulated by 25 mmol/L glucose in rMC1
Since IkBα is a prototypical proteasome substrate, to shed light on the NF-kB early activation, we verified whether high glucose delivery for 10–40 min actually induced a drop of ubiquitylated proteins by Wb. Compared to control cells, high glucose induced a robust decrease of ubiquitinylated proteins and a slight drop of p21 and p53, two reporter substrates of proteasome, over this time interval (Fig. 4A). To rule out alterations of the ubiquitylation pathway, cells were pre-treated with 500 nM epoxomicin (or DMSO vehicle) for 1 h before administering the high glucose challenge for 40 min. The proteasome inhibitor increased the basal content of ubiquitylated proteins in control cells and prevented their drop in high glucose treated cells (Fig. 4B).
To address whether bulk proteasome activity was induced by glucose uptake, crude cell extracts of rMC1 challenged with high glucose and low glucose or mannitol for 10, 20 and 40 min were assayed for the kinetics of Suc-LLVY-amc hydrolysis, a synthetic substrate of chymotrypsin-like activity (Fig. 4C). The cleavage rate of Suc-LLVY-amc was faster by extracts of cells challenged with high glucose at each time-point analysed (Fig. 4C). To further validate this finding, proteasome particles were separated by native-gel electrophoresis and probed with 75 µM Suc-LLVY-amc (Fig. 4D). Compared to control cells, high glucose-stimulated rMC1 displayed a robust hyper-activation of the capped assemblies (i.e., 30S–26S) which reached a peak around 20 min after stimulus delivery (Fig. 4D).
Immunostaining with an anti-Rpt5 (i.e., a 19S subunit) antibody clarified that the capped particles content was actually increased in high glucose stimulated cells with respect to control cells, in association with a drop of free 19S (Fig. 4D, bottom panel). To rule out uneven gel loading, crude cell extracts were further run by canonical Wb and probed with antibodies raised against α4 (i.e., 20S), Rpt5 and Rpt6 (i.e., 19S) (Fig. 4D, bottom panel).
Thereafter, proteasome proteolytic activity in living cells was further probed with TED, a quenched fluorogenic peptide of proteasome, according to a method validated elsewhere [32].
To meet all the chronological criteria (TED properties and timing of proteasome activation by glucose), rMC1 were first pre-treated with 500 nM epoxomicin for 1 h. Thereafter, high glucose and controls were administered and, after additional 5 min of incubation, 17 µM TED peptide was added to the culture medium and fluorescence release recorded for a total of 20 min after high glucose delivery. Following TED administration, fluorescence intensity quickly increased under all experimental conditions tested. However, after 15 min from TED addition (that is 20 min from glucose administration), whilst TED fluorescence reached a plateau in low-glucose or mannitol-treated cells, it displayed an additional steep increase in high glucose-stimulated cells (Fig. 4E).
High glucose induces the phosphorylation of the Rpt6 subunit of the 19S
The rapidity through which proteasome activation took place was suggestive of a post-synthetic modification, such as phosphorylation, on some proteasome subunits.
To test this possibility, a phos-Tag assay was set up to screen the electrophoretic mobility of a panel of 19S subunits for which phospho-sites have been described, such as Rpn6, Rpt3, Rpt5 and Rpt6 [33,34,35,36,37], in rMC1 stimulated with 25 mmol/L glucose and controls for 20 and 40 min. Rpn6 and Rpt5 immunostaining did not reveal any delayed mobility of species others than the unphosphorylated one (Fig. 5). Rpt6 immunostaining highlighted a delayed mobility of two bands (black arrow), which, indeed, were nicely detectable under all the experimental conditions. However, these phosphorylated species stained much stronger in high glucose-treated than in controls-treated cells, and this was paralleled by a drop of the intensity of the unphosphorylated species. Conversely, Rpt3 showed delayed-mobility species but no altered ratio between phospho- and unphosphorylated bands under any of the experimental conditions tested (Fig. 5).
Whole cell lysates were then probed with a phospho-specific antibody raised against Rpt6 phosphorylated at Serine 120, a major phospho-site. Immunodetection of phospho- and unphospho-Rpt6 confirmed the presence of a significant basal level of phosphorylation of this subunit under all experimental conditions, but, compared to untreated rMC1, further highlighted a robust increase of the pRpt6(Ser120)/Rpt6) ratio in cell lysates harvested after 20 and 40 min of high glucose stimulation (Fig. 5).
CamKII but not PKA phosphorylates Rpt6 and stimulates bulk proteasome activity in rMC1
To identify the kinase that actually phosphorylated Rpt6, we first focused on the calcium-dependent calmodulin Kinase II (CamKII) and protein kinase A (PKA) by using two specific inhibitors called Inhibitor XII (Inh. XII) and KT5720, respectively. First, we ruled out the presence of major effects of the inhibitors on proteasome activity under basal growth condition (Additional file 1: Fig. S2). Then, after a 2 h pre-treatment with the two inhibitors (each delivered at 10 µM) (or DMSO vehicle), rMC1 cells were stimulated with 25 mmol/L glucose or controls for 20 min and crude cell extracts probed for the kinetics of Suc-LLVY-amc hydrolysis (Fig. 6A). Cells treated with DMSO or KT5720 displayed an increase of Suc-LLVY-amc hydrolysis by high glucose comparable to that reported previously (see Fig. 4B) (Fig. 6A). Conversely, Inh. XII robustly opposed proteasome activation at 20 min of the high glucose challenge (Fig. 6A).
This finding was confirmed also by native-gel electrophoresis (Fig. 6B). Again, whilst DMSO and KT5720 did not alter the proteolytic activation of the capped assemblies by 25 mmol/L glucose (see Fig. 4D), Inh. XII prevented particle activation by sugar intake, both at 20 and 40 min after challenge induction (Fig. 6B). Interestingly, unlike DMSO and KT5720, Inh. XII further reduced the engagement of the free 19S in assembling the capped particles in the presence of 25 mmol/L glucose (Fig. 6B).
Thereafter, the phos-Tag approach confirmed that Inh. XII blocked the phosphorylation of Rpt6, whilst DMSO and KT5720 did not, as the delayed-mobility band staining positive for Rpt6 was increased only in the combined presence of the two latter compounds and of high glucose (Fig. 6C).
Moreover, in cells challenged with high glucose, the immunodetection of pRpt6(Ser120), as well as the pRpt6(Ser120)/Rpt6 ratio were markedly more robust in DMSO and KT5720 treated-cells than in Inh. XII-treated cells which further displayed a drop of basal phosphorylated Rpt6 (Fig. 6D).
To further verify that Rpt6 is actually phosphorylated at Serine 120 by CaMKII in rMC1, we assayed the immunodetection of this phospho-site after silencing the CaMKIIα gene, an isoform already studied in this cell model for its involvement in the signalling cascade of high glucose stimulation, by delivery of 27-mer siRNAs [9]. To this aim, rMC1 grown in standard low glucose medium were challenged with two individual siRNAs (referred to as siRNA#1-#2) or a pool of non-targeting siRNA (referred to as Pool) for 72 h. As internal control, rMC1 were also left untreated (referred to as Ctrl). Whole cell lysates were then harvested and analysed by Wb to verify effective silencing of CaMKIIα and pRpt6(Ser120) content (Additional file 1: Fig. S3).
With respect to Ctrl cells and cells challenged with the non-targeting pool, both siRNA#1 and siRNA#2 induced a significant downregulation of kinase content, although to a slightly different extent (siRNA#2 > siRNA#1) (Additional file 1: Fig. S3).
Interestingly, pRpt6(Ser120) immunodetection paralleled that of CaMKIIα under all experimental conditions tested. Thus, with respect of Ctrl cells, a significant drop of the basal phosphorylation of Rpt6 was documented in the presence of siRNA#1 and siRNA#2, but not in the presence of the non-targeting Pool (Additional file 1: Fig. S3). Conversely, basal level of unphosphorylated Rpt6 were unaltered by treatment. Hence, with respect to Ctrl and Pool-treated cells, the pRpt6(Ser120)/Rpt6 ratio did significantly drop in siRNA#1 and siRNA#2 treated cells.
Inhibition of CamKII and proteasome blocks NF-kB activation by high glucose in rMC1
We then tried to address whether proteasome and NF-kB activation were actually part of the same pathway.
First, we investigated the effects of Inh. XII, KT5720 and epoxomicin on the canonical NF-kB signalling and on CamKII content over 2 h of stimulation under basal growth conditions (Additional file 1: Fig. S4). Whole lysates were analysed by Wb revealing no alteration of IkBα and IKKβ and CamKII content (Additional file 1: Fig. S4). Immunostaining of pIkBα(Ser32), was unaltered by Inh. XII and KT5720 but displayed a robust increase at 2 h of epoxomicin stimulation.
Then, the turnover of IkBα after 20 min of high glucose challenge was investigated after having pre-treated rMC1 with each inhibitor. BAPTA-AM (10 µM), a validated cell-permeable Ca2+ chelator was also tested in this study. With respect to control cells, induction of IkBα decrease by high glucose was documented only in the presence of DMSO and of KT5720. Addition of either Inh. XII, epoxomicin or BAPTA-AM blocked the degradation of IkBα (Fig. 7A). To figure out whether inhibition of CamKII by Inh. XII and of proteasome by epoxomicin was effective also in inhibiting the transcriptional upregulation of pro-inflammatory cytokines, IL-8, IL-1β and MCP-1 transcripts were assayed by RT-PCR in rMC1 cells exposed or not to 25 mmol/L glucose for 40 min (Fig. 7B). Pre-treatment with Inh. XII or epoxomicin almost abolished the upregulation of the cytokines by high glucose (Fig. 7B) while DMSO had no effect (see comparison with Fig. 1B).
To test whether CamKII, but not PKA inhibition was effective also in halting pIkBα phosphorylation at Serine 32 during the second cycle of NF-kB activation (see Fig. 2), rMC1 cells, pre-treated with Inh. XII and KT5720 were challenged with high glucose for 2 and 3 h (Fig. 8). Compared to control cells, pIkB(Ser32) immunostaining in the presence of high glucose was increased in cells treated with DMSO or KT5720, whereas Inh. XII drastically opposed the accumulation of this phosphorylated protein (Fig. 8).
Influence of Rpt6 mutagenesis on early NF-kB activation in rMC1 exposed to high glucose
To clarify the role of Rpt6 phosphorylation in driving the early NF-kB activation, rMC1 were stably transfected with a plasmid encoding a phospho-dead mutant Rpt6 in which Serine 120 was replaced by alanine (hereafter referred to as Rpt6_S120A) and with the empty vector (hereafter referred to as Vector) as internal control. Two independent clones for Vector and Rpt6_S120A were generated.
Following transfection, the Rpt6 immunostaining was greater in Rpt6_S120A than Vector cells, whilst that of additional Rpt(s) subunits was not, and pRpt(Ser120) was significantly reduced in mutated clones (Additional file 1: Fig. S5A).
Surprisingly, Rpt6_S120A clones displayed a twofold increase of chymotrypsin-like activity on Suc-LLVY-amc with respect to Vector clones (Additional file 1: Fig. S5B). Furthermore, the capped assemblies and, most notably, the free 20S of Rpt6_S120A clones were significantly more active than those isolated from Vector cells, as from native-gel studies (Additional file 1: Fig. S5C).
Probing of filters with an anti-Rpt3 antibody made it clear that the immunostaining of capped particles was markedly increased, especially that of the 30S, whereas the free 19S content was significantly reduced in Rpt6_S120A vs Vector clones. Conversely, the 20S immunostaining by an anti-α4 antibody was unaltered between clones, implying that the free 20S of Rpt6_S120A clones was hyperactive (Additional file 1: Fig. S5D).
Since the mass/charge electrophoretic pattern of the free 20S argued against the binding of an alternative regulatory particle [38, 39], to interpret the 20S hyperactivation, whole cell lysates of all clones were run in parallel by phos-Tag and Wb. Filters were probed with an anti-α4 and an anti-pan-α20S antibody, that targets 6 out of 7 α-subunits except for α4, producing a typical smear. The phos-Tag approach highlighted (Additional file 1: Fig. S5E) that a delayed-mobility species of α4 was missing out in Rpt6_S120A clones in the absence of variation of total α4 (Additional file 1: Fig. S5E). By probing the same filters with the pan-α20S antibody it emerged that the pattern of the six subunits was unaltered, although in the former case separation of individual subunits was unsuccessful probably due to the presence of constitutive phosphorylation of some subunits which altered the migration pattern superimposing with that of at least some of them (Additional file 1: Fig. S5E).
To address the consequences of proteasome activation on the dynamics of NF-kB, a panel of proteins involved in the pathway were assayed by Wb. Surprisingly, IkBα displayed a robust drop in Rpt6_S120A clones (Additional file 1: Fig. S6A). Conversely, pIkBα(Ser32) was significantly increased and p65 unaltered (Additional file 1: Fig. S6A). Yet, the poly-ubiquitinylated proteins were slightly and not significantly increased in the Rpt6_S120A clones together with p21, whilst p53 was significantly increased in these cells (Additional file 1: Fig. S6B).
Based on the IkBα results, the transcription of pro-inflammatory cytokines was then assayed by RT-PCR. Surprisingly, the IL-8 mRNA was found to be nearly fivefold higher in Rpt6_S120A vs Vector clones, whereas MCP-1 did show a nearly 3.5-fold increase (Additional file 1: Fig. S6C). IL-1β (data not shown), instead, was characterized by a significant inter-experimental variability which did not allow to draw unambiguous conclusions.