PSMD5 Inactivation Promotes 26S Proteasome Assembly during Colorectal Tumor Progression

Protein synthesis and degradation are coordinated to allow adaptation to a rapidly changing microenvironment (1–3). Elevated protein synthesis can generate significant amounts of superfluous and damaged proteins, which are potentially toxic. However, cells can increase their rates of protein degradation to maintain protein homeostasis (4, 5). Selective protein degradation plays central roles in multiple biological processes including removal of misfolded and potentially toxic proteins, control of cell-cycle progression, regulation of gene expression, and changes in cell size and morphology (6, 7). The selective degradation of most intracellular proteins is carried out by the ubiquitin-proteasome system (UPS; refs. 8, 9). Proteins tagged with polyubiquitin chains are hydrolyzed into small peptides by the 26S proteasome in an energy-dependent manner. The 26S proteasome is assembled from a cylinder-shaped 20S proteolytic complex and a PA700 (19S) ATPase-regulatory complex (10). The assembly and activity of the 26S proteasome is tightly regulated by a large number of cofactors and chaperones (8). Cancer cells are more exposed to proteotoxic stress due to high metabolic rates, reprogrammed energy metabolism, and aneuploidy (5, 11–13). Despite this pressure, they successfully proliferate and efficiently withstand this challenge by fully exploiting the cells' defense mechanisms against proteotoxic stress. For instance, constitutive activation of the heat shock factor protein 1 (HSF1)-regulated heat shock response (HSR) pathway is among the recurring features of malignant cells, and inhibition of molecular chaperones represents a promising therapeutic strategy for cancer treatment (14). In addition, increased proteasome activity can suppress aneuploidy-associated proteotoxic stress (12). Activation of proteasomal degradation by deletion of the proteasome-associated deubiquitylating enzyme UBP6 attenuates the aneuploidy-induced changes in cellular protein composition and improves cell fitness (15). The mechanisms by which cancer cells overcome proteotoxic stress are of considerable scientific and clinical interest as advances in this area may identify new therapeutic targets. Notably, proteasome inhibitors have antitumor activity and are clinically used for the treatment of multiple myeloma and mantle cell lymphoma (16, 17).

The rate-limiting step in proteolysis by the UPS is considered to be substrate ubiquitination. However, recent findings indicate that proteolysis rate is significantly changed by regulatory pathways that affect the activity of the 26S proteasome itself, for example, by posttranslational modifications of proteasome subunits, or by facilitating the assembly of the 26S proteasome from the core proteasome and the regulatory complexes (18, 19). It has been previously shown that colon cancer cells increase expression of proteasome subunits in a concerted manner (20). However, upregulating the levels of proteasome subunits, while important, is not sufficient to significantly increase proteasome activity. Indeed, the levels of functional 26S proteasomes depend not only on the expression of its subunits, but also on their precise assembly. Work in yeast clearly demonstrates that assembly of 26S can vary significantly to meet the changing proteolytic needs of a cell (21).

Here we investigated a possible role of 26S proteasome assembly for protein homeostasis during tumorigenesis of the mouse gut epithelium. We show that 26S proteasome assembly is highly increased upon tumorigenic transformation of the gut epithelium. Importantly, this increase in 26S assembly is cell-autonomous and occurs exclusively in transformed cells, but not in other rapidly dividing cells. In addition, levels of the proteasome assembly inhibitor PSMD5/S5B are significantly decreased in tumors, and reexpression of PSMD5 blocks 26S proteasome assembly. Collectively, these results suggest that intestinal cancer cells escape proteotoxic stress by reducing PSMD5 to stimulate 26S proteasome assembly.

APC^min/+ mice were purchased from JAX Mice. We induced colonic tumors using the azoxymethane/dextran sulfate sodium (AOM-DSS) protocol as described previously (22). Briefly, we injected AOM (10 mg/kg body weight) into 8- to 10-week-old mice. Six days after injection, we treated mice with 2% DSS solution for 5 days. We repeated the DSS treatment on days 26–31 and on days 46–51. We assessed tumors 6–8 months after AOM injection. All experiments involving mice were approved by Institutional Animal Care and Use Committee of The Rockefeller University (protocol number 14703-H).

Intestinal crypts or tumors from WT, APC^min/+ and AOM-DSS–treated mice were isolated as described previously (23, 24), with several modifications: large bowels with and without tumors were flushed with cold PBS and opened longitudinally, washed with 70% ethanol. Large-bowel tumors were dissected from normal large bowel. Normal large bowels and tumors were agitated in a solution of 2 mmol/L EDTA in PBS for 30 minutes at 4°C. Tumors were dissociated using 70-μm strainer.

Crypts were collected as described and embedded onto Matrigel (BD Biosciences) with 50 μL/well in 24-well plates. DMEM/F12 culture medium (Gibco), containing Glutamax (Gibco) and penicillin/streptomycin was supplemented with B27 (Gibco, 1:50), 100 ng/mL murine Noggin (Peprotech), 20 ng/mL murine EGF (Peprotech), 10 ng/mL human basic-FGF (Peprotech) and 250 ng/mL human R-Spondin1 (Peprotech), 100 ng/mL mouse Wnt-3a (Peprotech) and 10 mmol/L nicotinamide (Sigma) for large-bowel organoids. Tumor organoids were cultured in medium containing only B27 (Gibco, 1:50), 100 ng/mL murine Noggin (Peprotech) and 20 ng/mL murine EGF (Peprotech) supplements.

Lentiviral infection was used to transduce genes into advanced tumor organoids. We generated recombinant lentiviral particles expressing mouse PSMD5 in HEK293T. Because lentivirus particles cannot penetrate the Matrigel layer, we incubated dissociated organoids with viral particles in liquid media for 6 hours 37°C under 5% CO₂ (v/v). Next, cells were embedded in Matrigel and grown into organoids. Two to 3 days after plating of cells, puromycin was added (5 μg/mL) to the medium for selection of successfully transduced cells.

The rate of protein translation was measured using the nonradioactive surface Sensing of translation (SUnSET) labeling method (25). Briefly, puromycin (4 μmol/L) was added to culture medium of organoids for 30 minutes, then organoids were harvested and puromycin incorporation was detected by Western blot analysis using an anti-puromycin antibody, clone 12D10 (1:5,000, Sigma). To label new synthesized proteins in vivo, we injected into peritoneum puromycin (0.04 μmol/gr of body weight of mice). We euthanized animals 30 minutes after puromycin injection and analyzed tissues by Western blot analysis using an anti-puromycin antibody, clone 12D10 (1:5,000, Sigma).

HEK293T and HT-29 cells (purchased from ATCC) were grown in DMEM containing 10% FBS and penicillin/streptomycin at 37°C under 5% CO₂ (v/v). For lentiviral production, HEK293T cells were transfected with the appropriate vectors using X-tremeGENE 9 DNA Transfection Reagent (Roche) according to the manufacturer's instructions.

Protein extracts of tissues were prepared in buffer [50 mmol/L Tris‐HCl (pH 8.0), 5 mmol/L MgCl₂, 0.5 mmol/L EDTA, 2 mmol/L ATP, 0.2% NP4, protease inhibitor, phosphatase inhibitor (Pfizer)], using glass dounce or polytron homogenizers. Extracts were clarified by centrifugation at 14,000 rpm for 30 minutes to remove nuclei and cell debris, and protein concentrations were measured by Bradford assay. To resolve proteasomes, 26-well 3%–8% triacetate gels (Criterion) were used. Samples were mixed with 2× native loading buffer (Bio-Rad) just before loading. Electrophoresis were carried out at room temperature (1 hour at 50 V) and then at 4°C for an additional 5 hours at 120 V, in running buffer (0.45 mol/L Tris, 0.45 mol/L boric acid, 5 mmol/L EDTA, 12.5 mmol/L MgCl₂, supplemented with 0.5 mmol/L DTT, 0.5 mmol/L ATP). For immunoblotting, proteins in native gels were transferred to polyvinylidene difluoride membranes. 26S and 20S proteasomes were detected with anti-RPT1 (19S specific) and anti-Alpha7 (20S specific) antibodies (Enzo). Purified bovine 20S, and 26S proteasomes (UBPBio), were used as standards.

Protein extracts were prepared as for Native gel electrophoresis. For Western blotting, proteins were resolved by SDS-PAGE and transferred onto a 0.45-μm PVDF membrane. Membranes were blocked overnight at 4°C with 5% milk in PBS and 0.5% Tween-20 (PBST). Membranes were incubated for 60 minutes with one of the following antibodies: polyubiquitin-FK2 (Enzo, 1:2,000), Actin HRP conjugated (Cell Signaling Technology, 1:10,000), p-S6 (Cell Signaling Technology, 1:1,000), PSMD 5 (Thermo Fisher Scientific, 1:4,000), ECM 29 (Abcam, 1:1,000), PSMD 11 (Cell Signaling Technology, 1:4,000), RPT1 (Enzo, 1:1,000), RPT3 (Enzo, 1:1,000), Alpha2 (Enzo, 1:1,000), Alpha4 (Enzo, 1:1,000), Alpha7 (Enzo, 1:1,000), and tubulin (Sigma, 1:1,000). Primary antibodies were detected by secondary species–specific HRP-conjugated antibodies (Jackson ImmunoResearch, 1:5,000). Detection was performed with Amersham ECL Western Blotting Detection Reagent (GE Healthcare).

ImageJ software was used for densitometric quantitation. Protein bands were selected with rectangle tool, picks were isolated, and area of the pick was calculated. Background calculation was done using same rectangle placed just below protein bands. Background pick values were subtracted from corresponding protein band picks values.

Proteasome activity was measured using Proteasome-Glo (Promega), which detects chymotrypsin-like activity in cell-based assays. For proteasome activity assays, tissue/cell lysates were prepared in PIPES buffer (50 mmol/L PIPES, 1 mmol/L MgCl₂, 50 mmol/L NaCl, 2 mmol/L EGTA, and 2 mmol/L ATP) using glass dounce. Fifty micrograms of protein extracts were aliquoted in duplicates to a black 96-well plate. Samples were assayed for proteasome activity with the Proteasome-Glo Chymotrypsin-like Cell-Based Assay (Promega) in a Spectramax M2 reader (Molecular Devices).

For PSMD5 KD experiments, HT-29 (5 × 10³ cells per well) were seeded in a white 96-well plate. Twenty-four hours later, cells were transfected with control or PSMD5 siRNA (Dharmacon OnTarget pull of 5 different targeting RNAis) using Dharmafect1 transfection reagent as recommended by the manufacturer. Cells were analyzed 5–6 days posttransfection using Proteasome-Glo assay (Promega). Luciferase values were normalized to cell viability as measured by PrestoBlue (Thermo Fisher Scientific) in a parallel plate. Samples were analyzed in a Spectramax M2 reader (Molecular Devices).

Normal gut epithelium and tumors of APC^min/+ mice were lysed in 300 μL of buffer [50 mmol/L Tris, pH 7.5, 150 mmol/L NaCl, 10% glycerol, 5 mmol/L MgCl₂, 0.22-μm–filtered and supplemented with 0.5% NP-40, 1 mmol/L DTT, 1 mmol/L ATP, phosphatase inhibitor (Roche)] using glass dounce. Cell lysates were cleared twice by centrifugation, 0.22-μm filtered, and quantified by Bradford protein assay. Equal amounts of each lysate (1.0 mg in 500 μL) were sequentially injected into a Superose 6 10/300 GL column (GE Healthcare) controlled by the ÄKTA-FPLC system (GE Healthcare). The column was preequilibrated and run with the above buffer without NP-40 at a rate of 0.2 mL/minute. Fractions (0.25-mL each) were collected and 15 μL from each fraction was used in Western blot analysis.

Protein extracts of tissues were prepared and resolved on native gel as described above. For the in-gel peptidase activity assay, we carefully dislodged the resolving gel from the glass plate into a tray with buffer [50 mmol/L Tris, pH 7.5, 150 mmol/L NaCl, 5 mmol/L MgCl₂, 5 mmol/L ATP, 200 μmol/L of suc-LLVY-AMC substrate (Bachem) with or without 0.02% SDS] and incubated it at 30°C for 30 minutes in a shaker with slow agitation (∼30 rpm). After that, the gel was transferred from the tray to a UV trans-illuminator for analysis.

Bulk population samples were processed by extracting RNA with RNeasy Plus Micro Kit (Qiagen) as per the manufacturer's recommendations, and then 10 ng of RNA from each sample was used to prepare RNAseq libraries using a modified SMART-Seq2 protocol as reported previously (26). The libraries were sequenced on an Illumina NextSeq 500. Because of different complexities between the early and advanced tumor organoid RNAseq libraries, the data was quantile normalized using the betweenLaneNormalization function of the EDASeq package in R. We then performed differential expression analysis on two biological repeats using the R package DESeq2 (27), which fits a negative binomial generalized linear model (GLM).

The rate of protein translation is higher in tumors than normal tissue, although to date the exact timing of this switch has not been determined (28). First, we directly measured protein translation in two different mouse models of colon cancer, APC^min/+ and AOM-DSS treatment, with the nonradioactive SUnSET method (19). In the SUnSET method, the antibiotic puromycin, a structural analogue of aminoacyl tRNAs, is incorporated into the nascent polypeptide chain and prevents further elongation (25). We observed highly increased puromycin uptake in large bowel tumors of APC^min/+ mice compared with normal tissue 30 minutes after puromycin injection (Fig. 1A). As puromycin prevents elongation, the resulting puromycin-tagged proteins are prematurely truncated, resulting in damaged proteins. The level of polyubiquitination in tumors and normal tissue was compared to determine whether the increase in the amount of damaged proteins corresponds to increased stimulation of the ubiquitin–proteasome system. Indeed, as expected, a significant increase of polyubiquitination staining in tumors 30 minutes after puromycin injection was observed in large bowel tumors of APC^min/+ mice compared with normal tissue (Fig. 1B). To establish whether the observed increase in translation derives from transformed epithelial cells or other cells residing in tumors, we generated organoids from normal large bowel epithelium and from tumors of DSS-AOM mice. Tumor organoids had significantly higher puromycin uptake than normal tissue organoids (Fig. 1C). Taken together, these results show that protein translation is rapidly increased upon transformation of the intestinal epithelium.

Next, we investigated whether increased protein synthesis in intestinal tumors is counterbalanced by increased protein degradation. The results shown in Fig. 1B demonstrate that the first step of UPS-mediated protein degradation, the conjugation of substrates with ubiquitin, occurs efficiently. To assess effects on the 26S proteasome, we first assayed levels of several individual proteasome subunits (Fig. 2A). We did not detect substantial differences in the levels of proteasome subunits between normal tissue and tumors (Fig. 2A). To assess the capability of the proteasome to degrade polyubiquitinated proteins, we measured total proteasome proteolytic activity using a Proteasome-Glo Assay (Promega) on cell lysates from large- and small-bowel tumors and the corresponding normal gut epthelium of APC^min/+ mice. Tumors from the small and large bowel of APC^min/+ mice displayed three times higher proteasomal activity compared with corresponding normal tissues (Fig. 2B). Subsequently, we used native gel and size-exclusion chromatography to compare the proteasome content of normal gut epithelium and tumors. To detect all proteasomal subcomplexes, we used native gel electrophoresis followed by in-gel suc-LLVY-AMC activity assay with SDS. Treatment with SDS allows detection of 20S proteasomes activity, which is otherwise not detected due to the closed gate of the 20S proteasome. We found that large-bowel tumors of AOM-DSS mice had higher levels of 26S proteasomes and lower levels of 20S compared with normal tissue, indicating enhanced assembly of 26S in tumors (Fig. 2C). Next, we evaluated 26S proteasome assembly in large-bowel epithelium and tumors of APC^min/+ mice using native gel electrophoresis (Fig. 2D). We quantitated Western blots from three independent experiments and present data as 26S/20S ratio. Again, this revealed a significant shift in the 20S/26S ratio to a higher portion of 26S proteasomes in tumors. As 26S proteasomes are the active particles responsible for degradation of polyubiquitinated proteins, these findings indicate that tumors have increased proteolytic potential. To further confirm these findings, we analyzed endogenous proteasome complexes from small-bowel tumors and normal tissues of APC^min/+ mice using size-exclusion chromatography. These experiments revealed a considerable shift toward larger proteasomal complexes without affecting the total amount of 19S complexes in tumors, as evaluated with an anti-RPT1 antibody (Fig. 2E). We also performed quantitative analyses and calculated the total amount of 26S (first peak) and 20S (second peak) proteasomes in tumors and normal tissue. (Fig. 2E, table). This analysis revealed again a reduction of unassembled 20S and elevated levels of 26S particles in tumors, indicating enhanced 26S assembly (Fig. 2E). Collectively, these results show that malignant transformation of intestinal epithelial cells is associated with increased 26S proteasome assembly.

One of the most important histopathologic features of intestinal neoplasia is tumor-associated inflammation. Inflammatory cells not only reside in the tumor, but also play important pathophysiologic roles in the development of a tumor by secreting various cytokines. As inflammatory cells are rapidly dividing and share many metabolic features with neoplastic cells, we wanted to clarify their potential contribution to increased 26S assembly in tumors. For this purpose, we induced colitis in C57BL/6 mice using 2.5% DSS in drinking water for 7 days. Next, we determined the proteasome levels of large-bowel epithelium and mesenteric lymph nodes of healthy and DSS-treated animals using native gel electrophoresis. We observed no change in proteasome assembly in the colonic tissue as a result of inflammation, and no differences between mesenteric lymph nodes and large bowel epithelium were observed irrespective of the inflammation (Fig. 3A). In contrast, 26S assembly was increased in large-bowel tumors (Fig. 3A). To further validate that the change in 26S proteasome assembly is caused by transformation of epithelial cells themselves, and not from other cell types, we grew organoids from large-bowel tumors and normal epithelium. Next, we performed native gel electrophoresis analysis of normal tissue and large-bowel organoids and found an increase in the assembly of 26S proteasomes in tumor organoids compared with normal tissue (Fig. 3B).

Unlike humans, APC^min/+ mice do not live long enough to develop adenocarcinomas as they die due to bleeding from gut adenomas (29). Growing these tumors as organoids enabled us to study tumor progression ex vivo. After 20 passages, large-bowel tumor organoids started to change their shape and form less symmetric and organized spheres (Fig. 4A). We hypothesized that the change in the shape represents ex vivo progression of large-bowel adenomas of APC^min/+ mice to carcinomas. To validate our hypothesis, we subcutaneously injected 5 × 10⁵ cells into NSG mice. Six weeks later, we detected subcutaneous tumors only in mice injected with the advanced organoids (Fig. 4B; Supplementary Fig. S1A). Histologic examination of these tumors revealed cells with different nuclear size and shape, prominent nucleoli and almost no gland formation, resembling the histology of poorly differentiated human adenocarcinoma (Fig. 4C). Next, we measured protein synthesis in early and advanced tumor organoids and found a dramatic increase in protein synthesis in the advanced tumor organoids (Fig. 4D). As we hypothesized that increased translation of advanced tumor organoids must be counterbalanced by enhanced degradation, we measured selected individual proteasome subunits and the proteasomal sub-complex content of early and advanced tumor organoids using PAGE-SDS and native gel electrophoresis, respectively (Fig. 4E; Supplementary Fig. S1B). Similar to what we saw in tumors and normal epithelium, we did not observe substantial differences in the levels of individual subunits. However, we detected an increase in 26S proteasomes in advanced tumor organoids compared with early tumor organoids (Fig. 4E). To validate that these 26S proteasomes are active, we performed in-gel suc-LLVY-AMC activity assays without SDS, which allows specific detection of active 26S proteasomes. Advanced tumor organoids contained more active single and double-capped 26S proteasomes (Fig. 4F, top blot) at the expense of free 20S proteasomes detected by anti-20S immunobloting (Fig. 4F, bottom blot). Recent studies showed that the rapamycin complex 1 (mTORC1) is linked to the regulation of protein degradation by proteasomes (30, 31). Therefore, we checked early and advanced organoids and found a strong activation of mTORC1 in advanced organoids as judged by the levels of pS6, which probably contributes to increased translation (Fig. 4G). To investigate whether the mTORC1 pathway is responsible for the enhanced assembly of 26S in the advanced organoids, we treated these organoids with rapamycin. Although rapamycin treatment resulted in the suppression of the mTORC1 pathway, we did not detect a substantial decrease in the assembly of 26S, as amounts of both 26S and 20S declined under these conditions (Fig 4G and H), indicating that mTORC1 pathway, though activated, is not responsible for the enhanced 26S assembly in advanced tumor organoids.

To identify potential genes that can explain the difference in 26S assembly between early and advanced tumor organoids, we performed RNA-seq analysis on the two lines of organoids. Using this strategy, we identified 50 proteasome-related genes, of which the expression of 17 changed more than 2-fold (Supplementary Table S2A; Table 1). As we were interested in factors that may increase assembly of the 26S proteasome in the advanced organoids, we looked for previously reported enhancers and inhibitors of assembly in the list of up- and downregulated genes, respectively. This revealed significantly decreased expression of one proteasome assembly inhibitor, PSMD5, and increased expression of two activators, ECM29 and PSMD11 (32–34). Western blot analysis of early and advanced tumor organoids demonstrated significant reduction of PSMD5 and elevation of ECM29 in advanced organoids, with no change in PSMD11 (Fig. 5A). Subsequently, we checked expression of these candidate proteins in the large-bowel tumors compared with normal epithelium. While no increase in the levels of ECM-29 and PSMD11 proteins was observed, there was a subtle reduction in the levels of PSMD5 in tumors compared with normal tissue (Figs. 2A and 5B; Supplementary Fig. S2B). Next, we plated xenografts originating from advanced tumor organoids and overexpressed PSMD5. Overexpression of PSMD5 caused a decrease in 26S levels and an increase in the amount of unassembled 20S, demonstrating reduced proteasome assembly as expressed in 26S/20S ratio (Fig. 5C). In addition, PSMD5 overexpression resulted in the reduction of total proteasome activity (Fig. 5D). Moreover, the decrease in 26S proteasomes led to accumulation of polyubiquitinated proteins (Ub K48; Fig. 5C). To check whether the suppression of PSMD5 may boost proteasome assembly and activity in human colorectal cancer, we knocked down PSMD5 in a human colorectal cancer cell line, HT-29 (Fig. 5E). Indeed, downregulation of PSMD5 led to enhanced assembly of 26S and increased total proteasome activity similarly to what we found in the murine model (Fig. 5E and F).

These observations indicate that inactivation of PSMD5 promotes 26S proteasome assembly during the malignant transformation of the intestinal epithelium and illustrate the importance of this mechanism for protein homeostasis in tumors.

It is well accepted that proteostasis in cancer cells changes dramatically upon malignant transformation (5, 11, 12, 14, 35). In this study, we utilized mouse models of colorectal cancer, APC^min/+ and DSS-AOM mice, to explore changes in proteostasis after malignant transformation. It is well documented that protein translation is enhanced upon malignant transformation (14, 28, 36). We show that an increase in protein synthesis is an intrinsic feature of transformed cells, as it occurs in organoids, outside the cellular context of the organism. Using a new model of ex vivo tumor progression from adenoma to carcinoma, we demonstrate that global protein synthesis further increases with tumor progression. These changes in protein synthesis of mouse colon cancer cells make organoids an attractive model to study adaptation responses of transformed cells to the proteotoxic stress.

Multiple mechanisms have been suggested to protect cancer cells from proteotoxic stress. The UPS is one of the two main protein degradation machineries and we show here that poly-ubiquitination of damaged proteins is very efficient in bowel tumors of mice (Fig. 1B). This increase in polyubiquitination is accompanied by increased proteasome activity. One means of proteasome activation in cancer is a global increase in the expression of genes encoding proteasome subunits. It was suggested that mTORC1 increases the amounts of proteasome subunits through induction of the transcription factor nuclear factor erythroid-derived 2-related factor 1 (NRF1; ref. 1), although this result remains controversial (30). In our model, rapamycin treatment resulted in the reduction of the levels of 26S, but not in the assembly of 26S. Another study showed an elevation in the number of proteasomal subunits through induction of nuclear factor E2-related factor 2 (Nrf2; ref. 20). While upregulation of individual proteasomal subunits is important to increase the amount of active 26S proteasomes, it is likely that additional mechanisms contribute to assure proper assembly of these individual subunits into active 26S proteasomes. Moreover, it was shown that there is no change in the levels of 18 individual proteasome subunits upon acute loss of APC despite an increase in translation (37). In this study, we show that assembly of 26S from 20S and 19S is increased in the intestinal tumors. This increase in assembly 26S is not solely due to upregulation of individual subunits as we saw a concomitant decrease in the amounts of 20S subcomplexes. Moreover, we show that the increase in assembly occurs only in transformed cells and not in other rapidly dividing cells, such as inflammatory cells. We also observed increased 26S proteasome assembly in intestinal tumors growing ex vivo as organoids, indicating that it is a cell-autonomous response independent of the in vivo tumor microenvironment. Finally, we found that 26S assembly increases with tumor progression. Collectively, these results suggest that 26S proteasome assembly increases as tumors progress to counterbalance elevated proteotoxic stress that is present under these conditions.

To gain insight into the molecular mechanism underlying the increased 26S proteasome assembly in intestinal tumors, we performed RNA-seq analysis of early and advanced tumor organoids. This analysis did not reveal a global change in the expression of genes encoding proteasome subunits. On the other hand, we identified two genes, ECM-29 and PSMD5, as possible regulators of 26S assembly based on clear changes of their expression in advanced organoids. Expression of ECM-29, a well-known enhancer of 26S assembly, was significantly increased in advanced organoids, but not in the large bowel tumors. Interestingly, it has been suggested that ECM-29 is an oncogene in castration-resistant prostate cancer (33). Our data are consistent with an oncogenic function of ECM-29 in advanced intestinal tumors, but a clear demonstration will require further functional studies. Here, we focused on PSMD5, which had significantly reduced expression of both mRNA and protein in advanced tumors. PSMD5/S5b is a chaperone that promotes assembly of the base sub-complex of the 19S proteasomes and has complex and seemingly opposing roles for proteasome assembly and activity. One the one hand, PSMD5 is important for the formation of the PSMD5–Rpt1–Rpt2–Rpn1 complex during 19S assembly (38, 39). Hsm3, the yeast ortholog of PSMD5, can associate with 19S sub-complexes via a carboxy-terminal domain of the Rpt1 base subunit, and Hsm3Δ cells display defects in 19S base assembly (40). The crystal structure of yeast Hsm3 suggests that Hsm3 also binds to Rpt2 and serves as a scaffold protein for Rpt1–Rpt2–Rpn1 complex assembly (41). On the other hand, in Drosophila and mammalian cells, elevated PSMD5 can inhibit 26S proteasome assembly and activity (19, 32). It was also shown that interaction between PSMD5 and Rpt1 is essential for proteasome inhibition (32). Moreover, amino acids 180-188 of PSMD5 are indispensable for 26S proteasome inhibition, and a specific point mutation (R184E) can abolish this inhibitory effect (32). It is important to note that the association of PSMD5with 19S sub-complex proteins is transient, and removal of this chaperone promotes 26S proteasome assembly (19, 32). Therefore, it appears that PSMD5/S5b levels need to be very carefully regulated, and perturbations may have different effects depending on context. Here we show that PSMD5 is downregulated in intestinal tumors. Moreover, as the tumor progresses, PSMD5 is silenced at the transcriptional level and this allows for increased assembly of 26S proteasomes. Consistent with this model, we found that overexpression of PSMD5 in advanced intestinal tumor organoids reduced total proteasome activity, 26S proteasome levels, and caused the accumulation of polyubiquitinated proteins. In addition, suppression of PSMD5 in a human colorectal cell line boosts 26S proteasome assembly and total proteasome activity. Therefore, it appears that one of the mechanisms intestinal tumor cells utilize to maintain protein homeostasis in the face of increased protein synthesis is silencing of PSMD5 expression to stimulate 26S proteasome assembly (Fig. 5G). Interestingly, a recent study identified PSMD5 as the most frequently transcriptionally suppressed 19S proteasome subunit in a dataset of thousands of human cancer cell lines (42). This work indicates that PSMD5 transcriptional suppression is an epigenetic process regulated by DNA promoter methylation. Moreover, PSMD5 suppression was found to be associated with resistance to bortezomib, which could be explained by our finding of enhanced 26S assembly. However, in contrast to this study, we did not observe a general suppression of 19S, but rather found five different 19S subunits to be upregulated in advanced tumor organoids. We envision that this mechanism assures that proteasomes are not rate-limiting for protein breakdown in the context of increased metabolic rates of tumor cells. Proteasome inhibitors are clinically used for the treatment of multiple myeloma and mantle cell lymphoma (16). However, proteasome inhibitors have multiple side effects as they are nonselective agents that impair protein breakdown in both malignant and normal cells. We propose that targeting proteasomal assembly in cancer cells may be more selective and have less side-effect toxicity. Therefore, targeting cancer-associated proteasome assembly pathways may provide new opportunities for cancer therapy.

G. Lalazar is a medical monitor at Immuron Ltd, a consultant at Fortress Biotech Ltd, and a consultant/advisory board member for Immuron Ltd and Fortress Biotech Ltd. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Levin, H. Steller

Development of methodology: A. Levin, G. Lalazar, J. Rodriguez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Levin, A. Minis, G. Lalazar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Levin, H. Steller

Writing, review, and/or revision of the manuscript: A. Levin, A. Minis, H. Steller

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Levin, H. Steller

Study supervision: H. Steller

Other (PI on the project): H. Steller

We thank Sigi Benjamin and other members of our lab for valuable discussions and Sohail Tavazoi for feedback on the manuscript. This work was supported by the Iris and Jumming Le Foundation and The Rockefeller University Center for Clinical and Translational Science (to A. Levin), NCATS grant UL1 TR000043 (to A. Levin), a gift from the Leona M. and Harry B. Helmsley Charitable Trust (to A. Levin), and NIH grant RO1GM60124 (to H. Steller).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.