Stress-resistant Translation of Cathepsin L mRNA in Breast Cancer Progression*

Background: Translational regulation might underlie the high expression levels of the protease cathepsin L (CTSL) associated with poor breast cancer prognosis. Results: Translation of CTSL mRNA is highly stress-resistant and promotes metastasis of murine breast cancer. Conclusion: CTSL mRNA circumvents translational shutdown in cancer-associated stress conditions. Significance: High expression of a metastasis promoting protease is maintained by translational regulation. The cysteine protease cathepsin L (CTSL) is often thought to act as a tumor promoter by enhancing tumor progression and metastasis. This goes along with increased CTSL activity in various tumor entities; however, the mechanisms leading to high CTSL levels are incompletely understood. With the help of the polyoma middle T oncogene driven breast cancer mouse model expressing a human CTSL genomic transgene, we show that CTSL indeed promotes breast cancer metastasis to the lung. During tumor formation and progression high expression levels of CTSL are maintained by enduring translation of CTSL mRNA. Interestingly, human breast cancer specimens expressed the same pattern of 5′ untranslated region (UTR) splice variants as the transgenic mice and the human cancer cell line MDA-MB 321. By polyribosome profiling of tumor tissues and human breast cancer cells, we observe an intrinsic resistance of CTSL to stress-induced shutdown of translation. This ability can be attributed to all 5′ UTR variants of CTSL and is not dependent on a previously described internal ribosomal entry site motif. In conclusion, we provide in vivo functional evidence for overexpressed CTSL as a promoter of lung metastasis, whereas high CTSL levels are maintained during tumor progression due to stress-resistant mRNA translation.

The cysteine protease cathepsin L (CTSL) is often thought to act as a tumor promoter by enhancing tumor progression and metastasis. This goes along with increased CTSL activity in various tumor entities; however, the mechanisms leading to high CTSL levels are incompletely understood. With the help of the polyoma middle T oncogene driven breast cancer mouse model expressing a human CTSL genomic transgene, we show that CTSL indeed promotes breast cancer metastasis to the lung. During tumor formation and progression high expression levels of CTSL are maintained by enduring translation of CTSL mRNA. Interestingly, human breast cancer specimens expressed the same pattern of 5 untranslated region (UTR) splice variants as the transgenic mice and the human cancer cell line MDA-MB 321. By polyribosome profiling of tumor tissues and human breast cancer cells, we observe an intrinsic resistance of CTSL to stress-induced shutdown of translation. This ability can be attributed to all 5 UTR variants of CTSL and is not dependent on a previously described internal ribosomal entry site motif. In conclusion, we provide in vivo functional evidence for overexpressed CTSL as a promoter of lung metastasis, whereas high CTSL levels are maintained during tumor progression due to stress-resistant mRNA translation.
One hallmark of tumor progression is the switch from benign hyperplasia to invasive growth and subsequent formation of metastases. Several steps to invasive tumor growth are facilitated by proteolysis (1). Among other protease families lysosomal cysteine cathepsins, especially cathepsin B and L, are often found to be highly expressed in aggressive tumors (2)(3)(4).
Tumor mouse models for pancreatic islet carcinoma showed retarded tumor growth upon cathepsin B or L deficiency (5). Furthermore, high activity of these proteases positively correlates with poor clinical outcome in several cancer entities. Especially the prognostic value of high cathepsin L (CTSL) 2 activity in breast cancer is appreciated (6 -9).
Gain of CTSL activity might be explained on several levels of regulation. First, CTSL mRNA can be elevated in tumor cells either by enhanced transcription as observed in ErbB2 positive breast cancer (10) or by epigenetic regulation (11,12). Second, CTSL protein stability and activity may be modulated by protein turnover and expression of endogenous cathepsin inhibitors (13). Interestingly, regulation on the level of mRNA translation into protein has been brought up as another aspect of CTSL regulation. The CTSL transcript is produced in different 5Ј UTR variants and it has been proposed that these variants differ in their efficiency to be translated into protein. Most of them vary in their length of the 5Ј UTR caused by different splice acceptor sites of exon 1 that are joined to the 5Ј end of exon 2 as depicted in Fig. 3A (14,15). The CTSL open reading frame starts in exon 2 so all splice variants encode for the same functional protein. In previous studies contradictory findings about translation efficiencies have been reported. Some reports assign the highest translation efficiency to the shortest variant (15), whereas others state that the longest variant is favored (16). Only some of the previous reports take into account that translation has to be assessed upon conditions that prevail within the cancer tissue.
Tumor-associated stress conditions coin the state of translation as described in the following. Especially in rather large solid tumors, e.g. tumors of the mammary gland, cellular stress due to reduced oxygen and nutrient supply is common. It is known that such conditions cause a general decline in translation of mRNAs into protein (17). Translation is enabled and regulated by at least 12 eukaryotic translation initiation factors (eIFs) (18). Under stress conditions a general shutdown of translation is mediated by reduced phosphorylation of eIF2␣, which abrogates formation of the pre-initiation complex of the 40S ribosomal subunit, the initiating methionyl tRNA, and eIFs. Under normal conditions this complex is recruited to the 5Ј cap of the mRNA. Stress signaling interferes in this process by activation of 4E-BP, a factor that hinders cap recognition. The key pathway to mediate translational shutdown is the mTOR pathway (19). Active mTOR inactivates 4E-BP by phosphorylation and keeps up activity of other eIFs to maintain cap-dependent translation. Consequently, pharmacological inhibition of mTOR by rapamycin or Torin-1 is a way to induce translational shutdown. Under such conditions mechanisms of cap-independent translation come into play. This can be facilitated by the use of internal ribosomal entry sites (IRES), a concept known from viral polycistronic mRNAs. Several eukaryotic mRNAs encoding for proteins that are essential for survival of the cell contain potential IRES domains in their 5Ј UTR (20). The longest CTSL splice variant is thought to form an IRES structure that enables favored translation under stress conditions (21). The basic functionality of the IRES structure has been shown by experiments with bicistronic reporter vectors (16,21). However, the functionality and actual impact of IRES structures on cellular mRNAs is still under debate (22)(23)(24).
In this study we address if one of the CTSL splice variants does indeed represent a stress-resistant source for CTSL in tumor tissue. Similar to previous reports we observed a discrepancy between CTSL mRNA and protein levels. However, to investigate whether this phenomenon is due to increased CTSL translation we choose a different approach than previous studies. Polyribosome profiling allowed us to analyze efficiency of translation of single splice variants transcribed from the genuine gene locus. We observed that all CTSL splice variants were recruited to the polyribosome with high efficiency in a stressresistant manner. This stress resistance was further confirmed by expression of single splice variants under hypoxia as well as mTOR inhibition. The circumvention of translational shutdown might be due to escape from translationally silent mRNA accumulations like stress granules or P-bodies rather than the predominant use of an IRES structure. Furthermore, expression of a human genomic CTSL transgene in the MMTV-PyMT mouse model of metastasizing breast cancer revealed increased metastasis, which might be fostered by the stress resistance of CTSL biosynthesis.
Patient Material-Patient material was obtained and worked with according to guidelines set by the Ethics Committee Freiburg (ethics approval 324/09_120807).
Tumor Phenotype-MMTV-PyMT mice were sacrificed at 10 or 14 weeks, and tumor tissue was resected from all mammary glands to measure the total tumor weight. Histopathological grading of hematoxylin and eosin (HE)-stained tumor sections of left thoracic mammary glands was performed in a blinded setting by an experienced histopathologist as described previously (27). For histomorphometric analysis, the area and number of metastatic foci on lung sections was measured using Axiovision LE 4.4 software (Zeiss, Oberkochen, Germany). For analysis of hypoxic areas, 1.5 mg of pimonidazole-HCl (Hypoxyprobe) (Hypoxyprobe, Inc., Burlington, MA) was injected intraperitoneally 30 min prior to sacrifice of tumor bearing mice. Binding of pimonidazole was visualized on paraffin sections using a FITC-labeled ␣-hypoxyprobe antibody and subsequent usage of a POD-labeled ␣-FITC antibody (catalog number MAB045P; Millipore, Darmstadt, Germany). Peroxidase activity was detected by incubation with diaminobenzidine (Sigma). Adobe Photoshop software was used for moderate contrast enhancement.
Cell Culture and Stress Treatments-MDA-MB 231 cells (catalog number 300275; CLS, Eppelheim, Germany) and BT-474 cells (ATCC, Manassas, VA) were cultured in Dulbecco's modified Eagle's medium (Gibco/Invitrogen, Paisley, UK) supplemented with 10% fetal calf serum (PAN, Aidenbach, Germany), 1% penicillin/streptomycin, and 2 mM L-glutamine (both from Gibco/Invitrogen) in humidified air containing 5% CO 2 . For stress treatments 1 ϫ 10 6 cells were seeded in a 100-mm dish and left overnight for cells to attach. For hypoxia cells were incubated in Anaerocult A bags (Merck, Darmstadt, Germany) for 16 h. In these sealed bags a hypoxic milieu of Ͻ0.5% is created within the first 90 min of incubation. During the experiment maintenance of the hypoxic milieu was monitored using Anaerotest strips (Merck). For mTOR inhibition equivalent cell numbers were treated with 250 nM Torin-1 (Tocris, Bristol, UK). For polyribosome gradient preparations ϳ6 ϫ 10 6 cells were seeded. In the case of cells carrying an inducible expression vector, CTSL expression was induced by treatment with 2 g/ml of doxycycline (Sigma). For blockage of stress granules formation of cells were treated with 10 M nocodazole (Sigma).
Cloning of Inducible Expression Vectors-The different CTSL transcript variants (CTSL-A: catalog number SC124186; CTSL-A3: SC100267; Origene, Rockville, MD) were amplified by PCR with mutagenesis primers, generating two new restriction sites at the ends of the PCR product. The purified PCR product was cloned into a lentiviral pTRIPZ shRNAmir vector (Thermo Fisher Scientific) via AgeI and MluI restriction sites. For cloning of the CTSL-CDS variant, primers for amplification of the insert were flanking the coding sequence only.
Preparation of PyMT Cells with Inducible CTSL Expression-Primary cells were prepared from tumors of MMTV-PyMT mice by mechanical disruption and enzymatic digestion of the tissue as previously described (27). Cells were spontaneously immortalized by repeated passaging for at least 25 passages. For the expression of different CTSL cDNAs cells were stably transduced by lentiviral transduction, employing the pMISSION system (Sigma) as previously described (28).
Polyribosome Profiles-Ten to 50% sucrose gradients containing 40 mM Tris, pH 8, 20 mM KCl, 10 mM MgCl 2 , 100 g/ml of cycloheximide were prepared in 12-ml polyallomer ultracentrifuge tubes (Beckman Coulter). Cells were incubated for 5 min with 1ϫ PBS, 100 g/ml of cycloheximide and detached from the cell culture plate by trypsination. Cells were pelleted and washed one time with PBS/cycloheximide. The cell pellet was resuspended in 500 l of hypotonic polyribosome extraction buffer (5 mM Tris-HCl, pH 7.5, 0.5% (w/v) sodium deoxycholate, 2.5 mM MgCl 2 , 1.5 mM KCl, 100 g/ml of cycloheximide, 2 mM DTT, 0.5% (v/v) Triton X-100). Lysates were passed through 27-gauge needles twice to ensure break up of cell membranes. For dissociation of tissue samples ϳ300 mg of frozen tumor tissue was ground into fine powder in liquid nitrogen. Only after addition of 600 l of hypotonic polysome extraction buffer were samples allowed to thaw on ice. Lysates from cells or tissues were centrifuged at 13,000 ϫ g for 7 min to pellet cellular membranes. Optical density values at 260 nm (A 260 ) were determined and 10 A 260 units were loaded onto sucrose gradients. The gradients were centrifuged at 275,000 ϫ g (TH-641 rotor (Sorvall), Thermo Scientific) for 2 h and 30 min at 4°C. Twenty fractions with each 600 l were collected while measuring the absorbance at 254 nm using a density gradient fractionator system (Teledyne ISCO, Lincoln, NE). The polyribosome profile obtained shows an ribonucleoprotein peak, the 40S and 60S subunits, and a single assembled ribosome appearing as the 80S peak followed by peaks representing mRNA with increasing numbers of assembled ribosomes. Profiles are presented under "Results" with the 80S peak, 0.5 M lithium chloride and 3 volumes of 100% ethanol were added to gradient fractions (both solutions were pre-cooled). RNA was precipitated for 2 h and 30 min at Ϫ80°C before tubes were centrifuged for 40 min at 16,500 rpm at 4°C. Supernatants were removed and pellets were briefly dried before adding 300 l of RNA Lysis buffer T. RNA was then purified using the "RNeasy Mini Kit" columns and buffers according to the manufacturer's instructions (Qiagen).

Transgenic Expression of Human CTSL in a Murine Breast
Cancer Model-To examine the influence of human CTSL expression on breast cancer progression we generated a breast cancer mouse model (expressing the PyMT oncogene under the control of the MMTV long terminal repeat/promoter (26,29)) with an additional genomic transgene for CTSL (25,30). Importantly, the CTSL transgene consists of a genomic construct, which requires RNA splicing and enables transcriptional control via the genuine human CTSL promoter allowing for experimental in vivo studies of human CTSL regulation (25). Expression of CTSL in tumor cells prepared from breast tumors out of this model was shown by immunoblotting (Fig. 1A). The CTSL protein presents in three distinct bands: a 37-kDa inactive proform that is further processed in the endolysosomal compartment into the active single chain form of 27 kDa and the active single chain form of which the heavy chain of 21 kDa is visible. The expression of the CTSL transgene induces about a 4-fold increase of CTSL proteolytic activity as compared with wild type mice (30). Tumor growth in the MMTV-PyMT model starts around 4 weeks along with progressing puberty in the mice (29). At around 10 weeks tumors are palpable in most of the 10 mammary glands and first micrometastases can be found in the lungs. When tissue sections of the primary tumor in this stage are stained for hypoxic conditions using a pimonidazol hydrochloride-based probe only a few areas stain positive as expected for rather small tumor volume (marked by arrow, Fig.  1B). At 14 weeks single tumors reach sizes around 1 cm 3 and most animals show metastases in the lung. In this stage tumors often show necrosis and extended areas of hypoxia (marked by arrow, Fig. 1B). Furthermore, tumor and metastasis burden were measured at these two time points. Transgenic expression of CTSL did not influence the weight and grading of the primary tumors formed at 10 or 14 weeks of age (Fig. 1, C and D). When the lungs of 14-week-old MMTV-PyMT mice were analyzed, animals carrying the CTSL transgene showed a higher metastatic burden compared with human CTSL negative tumor mice (Fig. 1E). Representative pictures of lungs from 14-week-old MMTV-PyMT mice with and without CTSL transgene are shown in Fig. 1F.
Translation of CTSL mRNA in Murine Breast Cancer Tissue-High expression of CTSL in the tumor increased the ability of metastases formation. This is in line with clinical findings of high CTSL levels as a poor prognostic factor (6). However, how tumor cells maintain high levels of CTSL activity needs to be further elucidated. In the tumor tissue of MMTV-PyMT mice, abundance of the CTSL mRNA derived from the genomic transgene was not increased with progression of the tumor from 10 to 14 weeks ( Fig. 2A). In contrast, the abundance of CTSL protein in tumor tissue lysates increased over this period of time (Fig. 2B). We wondered if this discrepancy in

Stress-resistant Translation of CTSL
CTSL mRNA and protein levels might be due to high efficiency of CTSL mRNA translation under tumor-associated stress conditions like hypoxia. To analyze translation in vivo, polyribosome profiles of tissue lysates from 14-week-old mice were prepared and fractions were analyzed by qRT-PCR. In these profiles the 80S peak is followed by peaks representing mRNAs with increasing numbers of associated ribosomes, i.e. the polyribosomes (example in Fig. 2C). mRNAs found in the fractions with high numbers of ribosomes are thought to be strongly translated (31). The translation status of CTSL was assessed in tumor samples of three different mice (Fig. 2D). The presence of CTSL mRNA in different fractions of the polyribosome gradient was normalized to ␤-actin. In addition the distribution of eukaryotic elongation factor 2 (eEF2) mRNA over the gradient was measured to assure that distribution in the profiles is not due to changes in ␤-actin ribosome association. Around 50% of CTSL mRNAs found in ribosome fractions range within high polyribosomes (shown by the bars below the graphs in Fig. 2D). The eEF2 mRNA is less efficiently translated as it can be found mainly in the 80S peaks and within the single ribosome fractions. In conclusion, high CTSL protein levels in the late stage tumors of MMTV-PyMT mice might indeed be caused by high efficiency in mRNA translation.
Polyribosome Profiling of CTSL 5Ј UTR Splice Variants in Murine Breast Cancer-Next, we addressed if high translation efficiency of CTSL can be assigned to one of the 5Ј UTR splice variants in particular. The CTSL splice variants vary in their length of exon 1 caused by different splice acceptor sites that are joined to exon 2 (Fig. 3A). The longest variant with an exon 1 of 340 bp named CTSL-A (NCBI entry NM_001912.4) has been proposed to build a three-dimensional structure that functions as an internal ribosomal entry site (predicted IRES is marked by the dotted line box in Fig. 3A (21)). The segment in position 250 to 340 bp has been described as critical for the formation of the IRES structure (21). The shorter variants CTSL-A1 (NCBI entry NM_125791.1), CTSL-A2 (described in Ref. 15), and CTSL-A3 (NCBI entry NM_145918.2) are lacking gradually longer parts of the 3Ј part of exon 1. The size of exon 1 in these transcripts is 315 bp for CTSL-A1, 250 bp for CTSL-A2, and 195 bp for CTSL-A3. If primers span from the constant 5Ј part of exon 1 to exon 2 a ladder of PCR products can be amplified (scheme in Fig. 3B). In tumor samples of 14-week-old mice all 5Ј UTR splice variants were produced from the CTSL transgene, although the A1 variant was detectable only in very small abundance and, therefore, not further analyzed (Fig. 3B). Splice site-specific primers were used for quantification of splice variants by qRT-PCR (scheme in Fig. 3C). The CTSL-A, -A2, and -A3 variants were quantified by taking the specific efficiencies of the different PCRs into account. Ratios were calculated by defining the sum of the three variants as 100%. In tumor samples of 14-week-old MMTV-PyMT mice the CTSL-A variant represents about 49%, CTSL-A3 accounts for ϳ45%, and the CTSL-A2 mRNA makes up to ϳ6% of the total CTSL mRNA. The percentage distribution of splice variants over the polyribosome profile remained fairly constant without increased polyribosome association of individual CTSL splice variants (Fig. 3C). In summary, our in vivo analysis of polyribosomeassociated 5Ј UTR CTSL splice variants did not provide evidence for one of the mRNA variants being superior to the others in terms of translation initiation.
Expression of CTSL Splice Variants in Clinical Samples-Although our mouse tumor samples do carry the human genomic construct of CTSL, its transcript is expressed in murine cells. Translational co-factors might be species specific and thereby translational regulation of human CTSL may differ in murine cells. To ensure that the distribution of splice variants is comparable in MMTV-PyMT mouse cancers and clinical breast cancer samples, we analyzed four different mammary cancer biopsies all classified as ductal carcinomas in the premetastatic stage (overview in Fig. 4A). To first assess abundance of total CTSL mRNA in these samples qRT-PCRs were performed. The mRNA abundance varied by about 50% (Fig. 4B). However, in this rather small cohort we did not see a correlation of grading or classification to the quantity of CTSL mRNA. The CTSL-A, -A2, and -A3 splice variants were found by RT-PCR in the biopsy samples, whereas the CTSL-A1 variant was not detectable (Fig. 4C). Corroborating the results in the MMTV-PyMT mouse tumor samples, CTSL-A and -A3 mRNAs are the major transcripts among the splice variants each accounting for ϳ45% of the total transcripts, whereas the remaining 10% are represented by the CTSL-A2 variant (Fig.  4D). The percentage distribution of CTSL splice variants is similar between the four cancer biopsies, only the sample hBC 2 showed somewhat increased CTSL-A mainly on the cost of reduced CTSL-A3 levels. However, due to the lack of biopsy material, the correlation of this splice variant distribution to the CTSL protein could not be addressed. Therefore, we sought to analyze translation of CTSL splice variants in human breast cancer cells in the next step.
Stress-resistant Translation of CTSL-mRNA in Human Breast Cancer Cells-To address translation of CTSL mRNA in human breast cancer cells under defined stress conditions, MDA-MB231 cells were exposed to hypoxia or chemical mTOR inhibition by Torin-1 for 16 h. Hypoxia was monitored by accumulation of hypoxia inducible factor-1␣ in whole cell lysates (Fig. 5A). Chemical inhibition of mTORC1 is a more defined way to suppress translation (19) and might give us further insight into stress-resistant translation of CTSL. Torin-1 is superior over rapamycin regarding inhibition of mTORC1 especially with regard to 4E-BP1 regulation (32). Hypoxia as well as Torin-1 treatment led to a general decline in translation indicated by the increase of the 80S peak and the lowering of the high polyribosome peaks (Fig. 6A, upper panel). Additionally, stress was monitored by an increase in phosphorylation of eIF2A (Fig. 5B). Levels of total eIF2A decreased upon hypoxic conditions (Fig. 5B), whereas protein levels of CTSL were not influenced by hypoxia and even increased during Torin-1 treatment (Fig. 5C). CTSL mRNA abundance remained stable under hypoxic or Torin-1 stress (Fig. 5D). In line with the protein levels, abundance of CTSL mRNA in high polyribosome fractions stayed constant during hypoxia, whereas eEF2 mRNA shifted into the 80S or single ribosome fractions (Fig. 6A). Notably, the percentage of CTSL mRNA in high polyribosome fractions increased to ϳ80% upon mTOR inhibition (Fig. 6A), which is in line with the higher abundance of CTSL protein under these conditions (Fig. 5C). The shift of eEF2 mRNA toward low polyribosome fractions was not that pronounced upon Torin-1 treatment compared with hypoxia (Fig. 6A). MDA-MB 231 cells have been described previously to express high CTSL protein levels, which correspond with high invasiveness (33). They are derived from pleural effusion and kept their potential to form metastases in the lung (34). Their molecular subtype corresponds to basal-like breast cancers, which differ  from the more luminal B-like subtype of tumors formed in the MMTV-PyMT model. To test if the observed stress resistance is dependent on the breast cancer subtype, we analyzed the polyribosome association of CTSL under Torin-1 treatment in BT-474 cells, classified as luminal B mammary carcinomas (Fig.  6B). In BT-474 cells, Torin-1 treatment increased the ratio of CTSL in high polyribosome fractions comparable with the results in MDA-MB 231 cells.
Furthermore, the MDA-MB 231 cells were used to assess translation of the 5Ј UTR splice variants by analysis of polyribosome association under defined stress conditions. The ratios of splice variants did not vary dependent on localization in the polyribosome gradient (Fig. 6C). Here, we corroborate the find-ing from the mouse tumors that none of the splice variants is superior in translation over the others. In summary, we observe that translation of CTSL mRNA is stress-resistant especially upon direct inhibition of mTOR. However, all CTSL splice variants are translated with comparable efficiencies.
Stress-resistant Translation Is Mediated by the CTSL UTRs Independently of the CTSL-A "IRES" Element-To further prove the translational advantage of CTSL mRNAs, we expressed CTSL transcript variants by stable transduction of murine breast cancer cells with doxycyline-inducible cDNA expression vectors (overview in Fig. 7A). Doxycycline treatment and exposure to stress conditions was started simultaneously to minimize the bias by effects of protein turnover (scheme in Fig. 7B). When the CTSL-A variant was expressed under hypoxia or Torin-1 treatment, stress did not affect biosynthesis of the CTSL protein (Fig. 7, C and D, left panel). The same outcome was observed when we expressed the shortest splice variant, i.e. CTSL-A3, which does not contain the pre-dicted IRES element (Fig. 7, C and D, middle panel). In contrast, translation of a CTSL transcript without flanking UTRs (only the coding sequence (CDS) of CTSL) is strongly impaired by stress-induced shutdown of translation (Fig. 7, C and D, right  panel). Apparently, the 5Ј UTRs of all investigated CTSL splice variants are able to mediate stress-resistant translation, although the 5Ј UTR of the CTSL-A3 variant does not contain a potential IRES or alternative regulatory motifs that may enhance translation initiation. This has been checked by scanning the sequence using RegRNA 2.0 (35) (data not shown). Therefore, we thought about alternative ways of how mRNA translation is regulated. mRNAs can be silenced by deposition in translationally inactive mRNA/protein accumulations, called stress granules or P-bodies (36). The inhibition of microtubule association by nocodazole is a well established method to block formation of stress granules (37). Cells that express either the CTSL-A transcript or the CTSL coding sequence devoid of UTRs were treated with stress and nocodazole from the time point of doxycycline induction (see Fig. 7E). As long exposure with nocodazole is cytotoxic, protein levels were analyzed after 6 h by immunoblotting. Due to this short period of doxycycline induction the cathepsin L protein presents mainly as the still unprocessed proform of 37 kDa. Nocodazole treatment did not affect CTSL protein production in cells expressing the UTR-bearing full-length cDNA (Fig. 7F). Importantly, nocodazole treatment fully rescued biosynthesis of CTSL protein under stress conditions in cells expressing the UTR-free cDNA. In line when Torin-1-stressed cells were treated with nocodazole, the UTR-free mRNA variant partially shifted from the 80S and low polyribosome fractions into the high polyribosome fraction (Fig. 7G). Thus, nocodazole prevented dislocation of UTR-free CTSL mRNA from polyribosomes, whereas CTSL biosynthesis from the full-length mRNA was not affected (Fig. 7F). Taken together, this series of experiments suggests that full-length CTSL mRNA intrinsically avoids its targeting to translationally inactive cytoplasmic mRNA deposits thereby enabling its persistent polyribosome association and CTSL protein biosynthesis upon cancer-associated cell stress conditions.

Discussion
Often the up-regulation of cancer-associated genes can be directly attributed to the level of gene transcription. The cysteine protease CTSL has been frequently reported to be highly abundant and prognostic in tumors, however, the mechanisms leading to these high CTSL levels are incompletely understood. Here, we report an intrinsic resistance of CTSL to stress-induced shutdown of translation in tumor tissues and provide in vivo functional evidence for overexpressed CTSL as promoter of metastasis.
In the tumor context the stable translation yielding constantly high levels of proteolysis by CTSL is very likely to contribute to the progression and metastasis of carcinomas. In our analysis, the transgenic expression of human CTSL in mouse mammary cancer led to an increase in lung metastatic burden in agreement with clinical studies that established positive correlations of CTSL with metastasis and poor prognosis in mammary cancer patients (6,7,9). It needs to be further explored at which stage of tumor progression high CTSL levels are especially critical. However, cancer cells are repeatedly exposed to stress conditions during the complex metastatic cascade from primary tumor to colonization of a secondary organ. Thus maintaining high CTSL levels during cell stress might actively promote several steps of metastasis (38).
We assessed CTSL translation by monitoring the association of multiple ribosomes to mRNA, a method called polyribosome profiling, which represents the gold standard for translation analysis (31). It allowed for the first time to analyze translation of intrinsically expressed human CTSL mRNA either directly from cancer tissue or under distinct stress conditions. With help of this method we established that association of CTSL mRNA to the ribosome is consistently strong in tumor tissues with prevalent hypoxia when compared with mRNAs of housekeeping genes. Even the intense exposure of tumor cells to hypoxia, which led to a general decline in polyribosomes, did not reduce recruitment of CTSL mRNA to the remaining polyribosomes. Interestingly, inhibition of mTORC1 even led to a further shift of CTSL mRNA to high polyribosome fractions. This might imply that upon complete mTORC1 inhibition the translation of housekeeping mRNAs might be reduced to a state allowing for a more pronounced CTSL translation. Inhibition of mTOR is a common tool to induce autophagy in cells. The advantage of CTSL mRNA translation under such conditions might mirror its functional role in autophagy as it is needed for the turnover of autophagolysosomes (39).
A major starting point of this study was the question if certain 5Ј UTR variants of CTSL serve as preferred sources for CTSL protein upon stress. The longest splice variant, variant CTSL-A, is able to form a secondary mRNA structure that functioned as an IRES element in bicistronic reporter assays (21). This suggested that CTSL-A would be preferentially translated upon stress-induced shutdown of cap-dependent translation. However, when we analyzed the translation of different CTSL splice variants, the ratios of the different transcripts in total cell and tissue lysates reflected the ratios found in the polyribosome fractions. This means all 5Ј UTR splice variants of CTSL are recruited to ribosomes with the same efficiency. Indeed, the biological relevance of IRES-mediated translation on cellular mRNAs has recently been doubted for some classical cellular "IRES genes" like hypoxia inducible factor and vascular endothelial growth factor (24,40). Thus, although IRES sequences proved to function in bicistronic assays, they might play a minor role in translational regulation in cancers or upon cell stress, at least in case of CTSL.
The above notion gained further support by expressing individual CTSL splice variants under stress conditions and quantification of the protein produced from these variants. Here, we found that the CTSL-A variant as well as the CTSL-A3 variant, which does not contain the putative IRES sequence, are both able to maintain efficient CTSL translation under stress. At the same time a CTSL variant devoid of UTRs was sensitive to stress-induced translational shutdown. This sensitivity was suspended when microtubule association was blocked by nocodazole treatment showing that this regulation is dependent on trafficking of mRNA within the cell. Nocodazole treatment blocks the formation of translationally inactive stress granules. However, the exact fate of the stress-sensitive CTSL mRNA variant still needs to be further defined. At the same time nocodazole treatment had no influence on the translation of the UTR-bearing CTSL variants supporting that the UTRs mediate the escape from stress-induced deposit in translationally inactive granules. How this is facilitated has to be further explored, especially as there are no annotated features proposing any translational regulation in the mRNA of the CTSL-A3 splice variant. The escape of certain mRNAs from stress granules has been observed before (41). For instance, mRNAs translated at the endoplasmic reticulum have been suggested to be protected from deposit in stress granules (42,43). However, the CTSL variant without UTRs does contain an endoplasmic reticulum-import signal and should thereby be normally recruited to the endoplasmic reticulum. Mechanisms of selective translational control of distinct mRNAs are emerging as regulators of gene expression. These are based on RNA-binding proteins in combination with micro-RNAs binding mainly to UTRs thereby regulating the initiation step. As examples, selective regulations serve diminished translation of mRNAs with 5Ј terminal oligopyrimidine tracts (44), the usage of iron responsive elements (45), and more recently described the function of cytosolic metadherin as a regulatory RNA-binding protein (46). Especially the regulation by metadherin might be interesting for our observation as it is highly expressed in solid tumors and even further up-regulated by hypoxia (47). The analysis of different breast cancer cell lines revealed that the observed resistance of CTSL mRNA to translational shutdown is not dependent on the molecular subtype, suggesting a more general mechanism. However, the critical factors that mediate persistent translation of CTSL need to be further defined.
In summary, breast cancer cells that face stress conditions during tumor progression maintain or even increase CTSL levels unaffected by the stress-induced general shutdown of protein biosynthesis. This stress resistance of CTSL translation is mediated by its UTR that might enable the CTSL mRNA to escape from deposit in translationally inactive stress granules. This posttranscriptional mechanism is likely to promote CTSLmediated tumor progression and metastasis.