Metabolic Engineering 20 (2013) 157–166

Contents lists available at ScienceDirect

Metabolic Engineering journal homepage: www.elsevier.com/locate/ymben

Stable microRNA expression enhances therapeutic antibody productivity of Chinese hamster ovary cells$ Michaela Strotbek a, Lore Florin b, Jennifer Koenitzer b, Anne Tolstrup b, Hitto Kaufmann b, Angelika Hausser a,n, Monilola A. Olayioye a,n a b

Institute of Cell Biology and Immunology, University of Stuttgart, Allmandring 31, 70569 Stuttgart, Germany Boehringer Ingelheim, Birkendorfer Str. 65, Biberach/Riss, Germany

art ic l e i nf o

a b s t r a c t

Article history: Received 4 July 2013 Received in revised form 24 September 2013 Accepted 11 October 2013 Available online 18 October 2013

MicroRNAs (miRNAs) are short non-coding RNAs that post-transcriptionally regulate the expression of different target genes and, thus, enable engineered gene networks to achieve complex phenotypic changes in mammalian cells. We hypothesized that exploiting this feature of miRNAs could improve therapeutic protein production processes by increasing viable cell densities and/or productivity of the mammalian cells used for manufacturing. To identify miRNAs that increase the productivity of producer cells, we performed a genome wide functional miRNA screen by transient transfection of Chinese hamster ovary (CHO) cells stably expressing an IgG1 antibody (CHO-IgG1). Using this approach, we identified nine human miRNAs that improved the productivities not only of the CHO-IgG1 cells but also of CHO cells expressing recombinant human serum albumin (HSA), demonstrating that the miRNAs act in a product-independent manner. We selected two miRNAs (miR-557 and miR-1287) positively impacting the viable cell density and the specific productivity, respectively, and then stably co-expressed them in IgG1 expressing CHO cells. In these cells, higher IgG1 titers were observed in fed-batch cultures whilst product quality was conserved, demonstrating that miRNA-based cell line engineering provides an attractive approach toward the genetic optimization of CHO producer cells for industrial applications. & 2013 The Authors. Published by Elsevier Inc. All rights reserved.

Keywords: Functional genomic screen miRNA library Recombinant therapeutic protein production Cell line engineering Fed-batch culture

1. Introduction CHO cells are the most widely used mammalian cell line for the manufacture of recombinant protein therapeutics. These cells can easily be adapted to growth in suspension under serum-free conditions and engineered to stably secrete recombinant proteins of interest that are correctly processed and post-translationally modified (Jayapal et al., 2007; Wurm, 2004). Despite optimization of the bioprocess, manufacturing of biologics for therapeutic use is expensive, and therefore, the industry seeks means to lower costs. In recent years, cell line engineering efforts have improved the productivity of CHO cells. These approaches included the stable expression of genes that impact cell performance at different

Abbreviations: miRNA, microRNA; CHO, Chinese hamster ovary; CGE, capillary gel electrophoresis; FAM, 6-FAM (6-carboxyfluorescein); HSA, human serum albumin; hsa-miR, Homo sapiens-microRNA ☆ This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-No Derivative Works License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and source are credited. n Corresponding authors. Fax: þ 49 711 685 67484. E-mail addresses: [email protected] (A. Hausser), [email protected] (M.A. Olayioye).

levels by, for example, increasing cell proliferation (Dreesen and Fussenegger, 2011), protecting from apoptosis (Becker et al., 2010; Lim et al., 2006), or enhancing their secretory capacity (Becker et al., 2008; Florin et al., 2009; Ku et al., 2008; Le Fourn et al., 2013; Peng and Fussenegger, 2009; Peng et al., 2010; Tigges and Fussenegger, 2006). Switching the phenotype of a cell from a low to a high producer state may be difficult to achieve by a single genetic alteration. Here, the class of RNA regulatory molecules, so-called microRNAs or miRNAs, are promising as they can regulate entire networks of genes and thus contribute to cell fate determination (Bartel, 2009; Fabian et al., 2010). miRNAs are small non-coding RNAs between 19 and 25 nucleotides in length that control the expression of target mRNAs at the post-transcriptional level. Because miRNAs bind to imperfect sequence matches within the 3′ UTR of target mRNAs, a single miRNA generally has multiple mRNA targets. miRNA binding results in translational repression and/or degradation of the target mRNA, thereby fine-tuning protein expression (Bartel, 2009; Fabian et al., 2010). Imposing no extra burden on the cell's translational machinery is an additional advantage of using miRNAs in cell line engineering (Hackl et al., 2012a). miRNAs have been well characterized in the context of development and cell transformation, where the expression of specific

1096-7176/$ - see front matter & 2013 The Authors. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ymben.2013.10.005

158

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

miRNAs correlates with different cell lineages, cancer types or tumor progression stages (Iorio and Croce, 2012; Lee and Ambros, 2001; Reinhart et al., 2000). In CHO cells, several miRNA profiling studies have also revealed that miRNAs are differentially expressed in response to specific culture conditions, such as nutrient depletion (Druz et al., 2012) or temperature shift (Barron et al., 2011; Gammell et al., 2007), and during the different growth phases of batch cultivation (Hernandez Bort et al., 2012). While this has led to the identification of distinct miRNA expression patterns, these data alone cannot predict which miRNA is functionally involved in a specific response and should be manipulated to confer an advantage to CHO cells in the bioprocess. To identify potential miRNAs for CHO cell line engineering, we performed a functional genome-wide screen using the titer of a therapeutic antibody model protein in the cell culture supernatant as a read-out. By transient transfection of an IgG1 producing CHO cell line (CHO-IgG1) with a human miRNA library, we were able to identify nine miRNAs that improved the titer and the specific productivity of CHO-IgG1 cells. These miRNAs also increased the productivity of CHO cells stably secreting recombinant HSA. We further demonstrated that the combined stable expression of two miRNAs selected from the CHO-IgG1 cell screening gave rise to IgG1 producing cell pools with increased productivity in fed batch cultures. To our knowledge this is the first report describing the improved performance of CHO cells by stable miRNA expression in a cell culture format that closely mimics the production process of a therapeutic protein.

2. Material and methods 2.1. miRNAs, siRNAs, DNA oligos, primers, and probes The human microRNA library (CS-001010 mimic microRNA library, lot 09167, ThermoFisher scientific, Waltham, MA, USA), miRNAs (see Table S1) and the mimic miRNA negative control #1 (CN-001000-01) were obtained from Dharmacon. The siRNAs siLacZ-FAM (5′-FAM-GCGGCUGCCGGAAUUUACCTT-3′), siLC (specific siRNA targeting the light chain of the IgG1) and siHSA (5′-AUUCCAGAAUGCGCUAUUATT-3′) were purchased from MWG (Ebersberg, Germany). Taqmans microRNA Assays included reverse transcription primers (for hsa-miR-557: RT001525, for hsa-miR-1287: RT002828 and for RNU6B: RT001093), qPCR primers and Taqman probes (for hsa-miR-557: TM001525, for hsamiR-1287: TM002828 and for RNU6B: TM001093) and were ordered from Life Technologies (Darmstadt, Germany). DNA oligos for cloning were from Biomers (Ulm, Germany) and are listed in Table S2. 2.2. Cell culture CHO-DG44 cells (Urlaub et al., 1986) stably secreting human serum albumin (CHO-HSA) or a monoclonal IgG1 antibody (CHOIgG1) and stable transfectants thereof were cultivated in suspension in a BI proprietary, serum-free media (Boehringer-Ingelheim, Biberach, Germany) supplemented with 400 nM methotrexate (MTX, Sigma-Aldrich, Taufkirchen, Germany) and 500 mg/mL G418 (Life Technologies). Seed stock cultures were sub-cultivated every 2–3 days with seeding densities of 2–3  105 cells/mL, respectively. Cells were grown in T-flasks (Greiner, Frickenhausen, Germany) in humidified incubators at 37 1C and 5% CO2. The cell concentration and viability was determined by trypan blue (Sigma-Aldrich) exclusion using a Neubauer counting chamber. MDA-MB468 cells (CLS, Eppelheim, Germany) were grown at 37 1C and 5% CO2 in humidified incubators in DMEM/F-12 medium (Life Technologies) supplemented with 10% fetal calf serum (FCS, PAN

Biotech, Aidenbach, Germany). Cells were sub-cultured every 3 to 4 days. 2.3. Fed-batch cultivation A 3  105 cells/mL cells were seeded into 125 mL shake flasks (Corning, Wiesbaden, Germany) in 30 mL of BI-proprietary production medium (Boehringer-Ingelheim) without antibiotics or MTX. The cultures were agitated at 120 rpm at 37 1C and 5% CO2 in a minitron incubator (Infors, Einsbach, Germany). On day 3, CO2 was reduced to 2%. BI-proprietary feed solution (BoehringerIngelheim) was added daily and the pH was adjusted to pH 7.0 using NaCO3. At concentrations below 2 g/L, glucose (#G8769, Sigma-Aldrich) was adjusted to 4 g/L. Cell densities and viabilities were determined by trypan blue exclusion using an automated counting chamber TC10 (Biorad, Munich, Germany). Cumulative specific productivity was calculated by dividing the product concentration at the respective day by the “integral of viable cells” (IVC). 2.4. Transient miRNA screen CHO-IgG1 cells were transfected with 1 mM RNA via nucleofection one day after passaging (4  105 cells/sample) in SG Cell Line 96-well Nucleofector™ Kit solution (#V4SC3096, Lonza, Verviers, Belgium) using the Amaxa 96-well Shuttle Device (Lonza) and program 96-DT-133 according to the manufacturer´s instructions. Subsequently, 12.5% of the cells were seeded in 100 mL medium into a well of a 96-well U-bottom plate (Greiner). In total, four plates were prepared. One day after transfection, the volume of the medium was doubled by addition of fresh medium without antibiotics or MTX. Supernatants were collected on days 1–4 post transfection by centrifugation of one of the 96-well plates (290  g, 5 min). Supernatants were transferred into a new 96-well plate and stored at 20 1C. A FAM-coupled non-targeting siRNA (siLacZFAM) and mock-transfected cells were used as negative controls. Transfection efficiency was determined by flow cytometry analysis (Cytomics FC-500, Beckman Coulter, Krefeld, Germany) of siLacZFAM transfected cells. The siLacZ-FAM, siLC and a mock control were included on each screening plate in duplicate. Two independent biological replicates of the screen were performed. For statistical analyses, the antibody concentration of each sample was first normalized to the median antibody concentration of the respective screening plate, referred to as the fold change. Then, the mean fold change of the two independent biological replicates was calculated. p-Values were determined by comparing antibody concentrations of the two biological replicates to the antibody concentrations of all siLacZ and mock controls using Student's t-test (two-tailed, unpaired). 2.5. Validation screen CHO-IgG1 and CHO-HSA cells were transfected with miRNAs via nucleofection as described above and seeded into 12-well plates (Greiner). Cell densities and viability were determined by trypan blue exclusion using a CEDEX cell quantification system (Roche, Mannheim, Germany). Product concentrations in the supernatant were measured by ELISA. siLacZ-FAM, and siLC or siHSA served as controls. Statistical analysis of antibody concentrations and specific productivities obtained on day 1–4 was performed using a two-way ANOVA followed by a Bonferroni post-test. For the co-transfection of hsa-miR-557 and hsa-miR1287, 0.5 mM of each miRNA was used, and the amount of total RNA in the single control transfections was adjusted to 1 mM by adding mimic miRNA negative control #1.

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

2.6. Cloning of miRNA expression plasmids To stably express miRNAs the BLOCK-iTTM Pol II miR RNAi Expression Vector Kit (pcDNA6.2-GW/emGFP-miRNA expression system kit, K4936-00, Life Technologies) was used. DNA oligonucleotides encoding hsa-miR-557 and hsa-miR-1287 were designed as shown in Table S2 and hairpin structures were analyzed using the online software mfold (Zuker, 2003). The DNA oligos were hybridized and then cloned into the 3′-UTR of the emerald GFP reporter as described by the manufacturer. To generate a vector containing two different miRNAs in tandem the manufacturer's instructions were followed. The controls were a negative control miRNA (pcDNA6.2-GW/emGFP-neg. control miR) and LacZ (pcDNA6.2-GW/emGFP-control). Sequence integrity was confirmed by DNA sequencing using the primers EmGFP forward and miRNA reverse provided in the BLOCK-iTTM Pol II miR RNAi Expression Vector Kit. 2.7. Stable overexpression of miRNAs CHO-IgG1 cells were transfected with pcDNA6.2-GW/emGFPmiRNA vectors using Lipofectamines 2000 and PlusTM Reagent (Life Technologies). Cells were selected with 10 mg/mL Blasticidin S (Life Technologies). miRNA-positive populations were enriched by fluorescent activated sorting (FACSDiVa, BD Bioscience, Heidelberg, Germany) for GFP-positive cells and miRNA overexpression was verified by qRT-PCR. 2.8. RNA extraction and quantitative RT-PCR analysis Total RNA was extracted from 2  105 to 2  106 CHO cells using the mirVANATM miRNA Isolation Kit (Life Technologies) according to the manufacturer´s protocol. RNA samples were quantified using a Nanophotometer (Implen, Munich, Germany) at OD 260/ 280 nm. Total RNA (10 ng) was reverse transcribed into cDNA using Taqmans microRNA Assay and TaqMans MicroRNA Reverse Transcription Kit (Life Technologies) according to the manufacturer's instructions. Q-PCR was performed with Taqmans microRNA Assays and DyNAmo ColorFlash Probe qPCR Kit (Biozym, Hess. Oldendorf, Germany) using a Cfx96 device (Biorad). RNU6B (U6 small nuclear 2) was used as reference. Calculation of ΔCq values was done with the single threshold method (Biorad CFX manager software 2.1). 2.9. Antibody and HSA quantification by enzyme-linked immunosorbent assay (ELISA) For the quantification of antibody concentrations, ELISA plates (Greiner) were coated with a goat anti-human Fc fragment (#109005-008, Dianova, Hamburg, Germany) (diluted 1:500 in coating buffer: 0.1 M sodiumcarbonate buffer, pH 9.5) at 4 1C overnight. Plates were washed 3 times with washing buffer (0.15% Tween-20 in PBS) and then blocked for 1 h at RT with PBS supplemented with 1% human albumin (#T844.2, Carl Roth, Karlsruhe, Germany). After washing, the samples and purified IgG1 antibody as a standard diluted in assay diluent (PBS supplemented with 0.5% human albumin and 0.01% Tween-80) were added and incubated for 1.5 h at RT. Detection was performed with an anti-human kappa light chain alkaline phosphatase conjugated antibody (#A3813, Sigma-Aldrich) at a 1:5000 dilution in assay diluent. After 1 h incubation at RT, plates were washed and 4-nitrophenyl phosphate disodium salt hexahydrate (Sigma-Aldrich) dissolved in substrate buffer (0.1 M glycine, 1 mM ZnCl2, 1 mM MgCl2, pH 10.4) was added. The enzyme reaction was stopped with 3 M sodium hydroxide after 15–20 min and plates were measured at 405 nm (reference 492 nm) with a multiskan FC reader (ThermoFisher

159

Scientific). For the quantification of albumin concentrations in the cell culture supernatants, the Human Albumin ELISA Quantitation Set (#E80-129, Bethyl Labs, Montgomery, TX, USA) was used according to manufacturer's instructions. A goat anti-human albumin antibody was used for coating and a HRP-conjugated goat anti-human albumin antibody for detection. 3,3′,5,5′-Tetramethylbenzidine (TMB, BD Biosciences) was added, the reaction was stopped with 1 M H2SO4 and the absorbance was measured at 450 nm with a multiskan FC reader.

2.10. Antibody purification Cell culture supernatants obtained from fed-batch cultivation of CHO-IgG1 cells were concentrated using 50 kDa Amicon centrifugal filter units (Millipore, Schwalbach, Germany). The concentrate was purified with Protein A HP spin trap columns (GE Healthcare, Dornstadt, Germany) according to the manufacturer's instructions. The antibody was eluted with 0.1 M glycine (pH 2.7) and neutralized with 1 M Tris–HCl to pH 7. The buffer was exchanged to PBS using 50 kDa Amicon centrifugal filter units. Protein concentrations were determined photometrically by measuring absorbance at 280 nm using a NanoDrop spectrometer (ThermoFisher Scientific) and the IgG1 specific molar extinction coefficient calculated according to (Pace et al., 1995).

2.11. SDS-PAGE To assess antibody integrity, 2.5 mg purified protein was subjected to a 6% Tris-glycine gel under non-reducing conditions or boiled in sample buffer and subjected to a 12% Tris-glycine gel under reducing conditions. Protein bands were stained with Coomassie brilliant blue (Carl Roth).

2.12. Flow cytometry analysis A 5  105 MDA-MB468 cells were incubated with 10 ng IgG1 antibody purified from the supernatant of parental or miRNA expressing cells for 1 h on ice in FACS buffer (PBS containing 2% FCS and 0.02% sodium azide). Cells were washed two times with FACS buffer, followed by incubation with a goat anti-human IgG1 coupled to PE (#P9170, Sigma-Aldrich) for 45 min on ice. After two washes with FACS buffer, cells were analyzed by flow cytometry. As a negative control unstained cells and cells incubated with a purified human IgG isotype control (#027102, Life Technologies) were used.

2.13. Glycan pattern analysis The antibody glycosylation pattern was analyzed with the ProfilerPro Glycan Profiling Kit Ver 2 on a LabChip GXII capillary gel electrophoresis (CGE) instrument (Caliper Life Sciences, Hopkinton, Massachusetts, USA) according to the manufacturer's protocol. Antibodies were deglycosylated using PNGase F. The released glycans were labeled with a fluorescent dye in a hydrazide reaction. Samples then underwent microchip-based separation by CGE. Electropherograms were analyzed by the LabChip GX software package to identify and quantify the individual sugar structures. All values were normalized to 100% total sugar structures per sample and comprise the N-linked bi-antennary high mannose (Man5) and the complex structures in the presence (A2FG0; A2FG1; A2FG2) and absence (A2G0) of fucose.

160

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

Fig. 1. MicroRNA screen in IgG1-producing CHO cells. (A) Workflow of the screening procedure. CHO-DG44 cells secreting a human IgG1 antibody were transiently transfected with a human mimic microRNA library comprising 879 human microRNAs. Antibody concentrations in the supernatant of the transfected cells were determined by ELISA on day 1–4 post transfection. (B) Correlation plot of antibody concentrations measured on day 1 (light grey), 2 (dark grey), 3 (red) and 4 (blue) for the two biological replicates. Circles highlight the data points corresponding to hsa-miR-557 (grey) and hsa-miR-1287 (black). (C) Workflow of the data analyses. miRNAs hits were required to induce a significant change (po 0.01) in IgG titers of at least 1.5-fold. (D) List of miRNA screen hits that positively affected antibody titers with their fold changes on day 3 and 4 and p values. (E) For each miRNA sample, the fold change in antibody concentration based on the median of the respective plate is plotted for day 3 (red) and 4 (blue). The grey dotted lines indicate the variance region. Screen hits are highlighted in bold and numbered according to the list shown in (D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

3. Results 3.1. Transient miRNA productivity screen in CHO host cells To investigate whether miRNAs could impact the productivity of industrially relevant producer cell lines we chose CHO-DG44 cells stably secreting an IgG1 antibody (CHO-IgG1) as a model. Using these cells, a genome-wide screen was performed in a 96-well format by transient transfection with a human mimic miRNA library comprising 879 sequences that correspond to mature double-stranded miRNAs (see Fig. 1A for the detailed workflow of the screen). Transient transfection conditions of CHO-IgG1 cells in the 96-well format were optimized with a FAM-labeled non-targeting control siRNA (siLacZFAM) and a siRNA targeting the light chain of the IgG1 antibody (siLC). Using an optimized nucleofection protocol we achieved reproducible transfection efficiencies close to 100%, as estimated by quantification of FAM-positive cells by FACS, and transfection of siLC strongly reduced the antibody concentrations in the supernatant of CHO-IgG1 cells as measured by ELISA (see Supplemental Fig. S1). In the primary screen, we measured the IgG1 antibody concentration in the supernatant on days 1–4 post transfection by ELISA analysis. For each plate the transfection efficiency was monitored by FACS analysis of the siLacZFAM transfected cells and ELISA analysis of supernatants after siLC transfection. At this stage, it was not differentiated whether effects observed resulted from a change in specific productivity and/or cell growth. The screen was repeated and the data of the two independent biological replicates (replicates #1 and #2) obtained for the four consecutive days post transfection were then analyzed as described in (Birmingham et al., 2009) (see Fig. 1C for a summary of the data analysis). The raw data are shown in Supplemental data Table S1. To assess the quality of the screen, data from replicate #1 were plotted against replicate #2 data (Fig. 1B), resulting in a correlation coefficient of 0.87. In this plot, each data point corresponds to the antibody concentration in the supernatant after transfection with one specific miRNA. In general, the IgG1 titer in the supernatant accumulated over time in both replicates, but a subset of miRNAs increased or decreased antibody yields, which was most obvious on days 3 and 4. To identify these miRNAs, the fold change induced by each miRNA on days 3 and 4 was calculated by dividing the measured antibody concentrations by the median antibody concentration of all miRNA samples on the respective screening plate, thereby enabling the direct comparison of samples on different plates. Most miRNAs had no effect on antibody concentrations. As shown in Fig. 1E, most data points clustered around 1 within a variance range of 20% (marked by grey dotted lines). This is in line with the variance determined in control experiments performed during the establishment and optimization of the screening procedure. Criteria to qualify as a screening hit were set as follows: miRNAs were required to induce a significant change in antibody titer greater than 1.5-fold on days 3 and/or 4 (po0.01). Based on these criteria, the expression of 14 miRNAs significantly decreased antibody concentrations in the supernatant of CHO-IgG1 cells, whereas nine miRNAs were found to lead to a significant increase. The latter group of miRNAs, which is of particular interest to miR-engineering approaches of manufacturing cell lines, are highlighted in Fig. 1D and E and listed in Supplemental data Table S2. 3.2. Validation of miRNA screen hits To validate the nine miRNAs positively affecting IgG1 antibody yields as screen hits, CHO-IgG1 cells were transiently transfected in a 12-well format with each of these miRNAs. Cell densities were measured to determine not only total antibody concentrations in the supernatant but also the specific productivity of miRNAexpressing cells. Furthermore, to exclude the possibility that miRNA-mediated product titer changes are specific for the IgG1 antibody or the CHO-IgG1 cells, a stable pool of CHO-DG44 cells

161

secreting human serum albumin (CHO-HSA) was used as a second model cell system. Three days post transfection, product concentrations, specific productivities, and cell densities were determined (Fig. 2A–C). Indeed, transfection of all nine miRNAs increased the IgG1 concentrations, albeit in most cases to a lower extent than observed in the primary screen, with hsa-miR557, hsa-miR-612, and hsa-miR-1287 resulting in significantly increased IgG1 concentrations greater than 1.5 fold. Interestingly, all miRNAs except hsa-miR-1271 also improved HSA titers, with hsa-miR-1287 expression inducing the highest product titers in both CHO-IgG1 and CHO-HSA cells (see Fig. 2A). In all cases, the increased titers were reflected at the level of the specific productivity of CHO-IgG1 and CHO-HSA cells (Fig. 2B). Taken together, of the nine primary hits, three miRNAs were validated as top hits of the screening (hsa-miR-557, hsa-miR-612, and hsa-miR-1287), increasing both the antibody titers and the specific productivity of IgG expressing CHO producer cells. Expression of most of the miRNAs negatively affected cell densities (see Fig. 2C). In both model cell lines the cell density of miRNA-transfected cells was decreased on day 3 by approximately 30–50% compared to control transfected cells. Of interest, only hsa-miR-557 did not negatively impact cell growth and even had a positive effect on the proliferation of CHO-IgG1 cells (Fig. 2C). Because hsa-miR-1287 had the strongest product-independent effect on the specific productivity, we speculated that its coexpression together with hsa-miR-557 might further boost the productivity of producer cells. To test this, we transiently transfected CHO-IgG1 cells with these two miRNAs individually or in combination and determined the cell density, specific productivity, and antibody concentration on day three post transfection (Fig. 2D). Under these conditions, the growth stimulation by hsa-miR-557 was not seen, which may be due to the fact that only half of the miRNA amount was transfected in comparison to the validation experiment (see Fig. 2C). Likewise, hsa-miR-1287 did not reduce cell proliferation as strongly when expressed at a lower concentration. Nevertheless, productivity in singly transfected cells was still improved compared to cells transfected with the control miRNA (Fig. 2D). Most importantly, co-expression of miR-557 and miR-1287 led to enhanced antibody concentrations and specific productivity compared to the single expressions (Fig. 2D), showing that miRNAs can positively impact CHO cell productivity in an additive manner. 3.3. Stable expression of productivity-enhancing miRNAs in CHOIgG1 cells For industrial purposes, the stable expression of miRNAs in producer cells is desirable. We therefore made use of the BLOCK-IT POLII miR expression system that was recently reported to be functional in CHO cells in transient experiments (Jadhav et al., 2012). Here, the sequence of the respective mature miRNA is integrated into the genetic backbone of the mouse miRNA mmumiR-155 consisting of an optimized loop, and the 3′- and 5′flanking regions. To stably express hsa-miR-557, hsa-miR-1287, or hsa-miR-557 together with hsa-miR-1287 in CHO cells we cloned the respective mature human miRNA sequences (Table S3) into the BLOCK-IT POLII miR vector, generating the constructs depicted in Fig. 3A. In addition, a control vector containing a negative control RNA was included (pcDNA6.2-GW/emGFP-neg. control miR). To verify the expression and correct processing of the mature human miRNAs, CHO-IgG1 were transiently transfected with the indicated plasmids and quantitative real-time PCR (qRT-PCR) analysis was performed. As a positive control the mature miRNA was transfected. A signal was obtained for cells transfected with the plasmids encoding hsa-miR-557 and hsa-miR-1287 whereas in the negative controls almost no signal was detectable (see Fig. 3B

162

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

and C). We thus conclude that CHO-IgG1 cells successfully transcribed and processed the plasmid-encoded human mir-557 and mir-1287 hairpins. Noteworthy, in the vectors containing hsa-miR557 and hsa-miR-1287 in tandem the order miR557–miR1287 resulted in higher expression levels of both miRNAs than the order miR1287–miR557 did (2.5 and 1.4 fold higher for hsa-miR-557 and hsa-miR-1287, respectively; see Fig. 3B and C). It is thus conceivable that processing of the hairpins is dependent on the order of the miRNA sequence due to secondary structure formation. Next, stable cell pools were generated by Blasticidin S selection of CHO-IgG1 cells transfected with the miR557–miR1287 dual expression construct. Stable cell pools expressing the miRNA negative control (neg. control miR) were generated as a control. The stable pools were sorted based on GFP fluorescence to enrich for miRNA expressing cells, yielding a cell population in which the mean GFP fluorescence intensity remained stable for at least 50 days in culture (see Fig. 3D). The stable cell pools were then

analyzed for miRNA expression by qRT-PCR. Indeed, expression of the mature miRNA was readily detected in the stable pools, the level of which was comparable to that seen in the transient experiments (see Fig. 3E and F), indicating that the cells did not activate any mechanisms to suppress long-term ectopic miRNA expression. 3.4. Improved performance of stable miR-engineered CHO-IgG1 cells in fed batch cultures Finally, we were interested in determining whether miRNA overexpression improves the performance of CHO cells in a production process and therefore subjected the stable CHO-IgG1 pools co-expressing hsa-miR-557 and hsa-miR-1287 to fed-batch culture. To exclude clonal artefacts three independent CHO-IgG1 pools stably expressing the miR557–miR1287 combination and four independent CHO-IgG1 pools stably expressing the negative

Fig. 2. Validation of the productivity-enhancing effects of the miRNA screen hits. CHO-DG44 cells stably secreting IgG1 (grey bars) or HSA (black bars) were transiently transfected with the indicated miRNAs and cultured in 12-well plates in quadruplicates. Product concentration in the supernatant (A), specific productivity (B) and cell density (C) were monitored over 4 days. Results were normalized to the control (siLacZ). As additional controls, siRNAs targeting the respective product were used (siLC in the case of IgG1 and siHSA in the case of HSA). Graphs show the mean (error bars represent SEM) of four (CHO-IgG1) and two (CHO-HSA) independent experiments for day 3 post transfection. Statistical analysis of the data obtained on day 1 to 4 from CHO-IgG1 cells was performed using two-way ANOVA followed by a Bonferroni post-test. (D) CHO-IgG1 cells were transfected with a negative control miRNA, hsa-miR-557, hsa-miR-1287, and hsa-miR-557 plus hsa-miR-1287. Shown are the results obtained on day 3 for product concentration (light grey bars), specific productivity (dark grey bars), and cell density (black bars) normalized to control (neg. control miRNA). Bars represent the mean of 5 independent experiments each performed in duplicates in a 12-well format. Error bars represent SEM, and statistical analysis of data obtained on day 1 to 4 was performed using two-way ANOVA followed by a Bonferroni post-test (n: p o 0.05, nn: p o 0.01, nnn: po 0.001).

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

163

Fig. 3. Vector-based miRNA expression in CHO-IgG1 cells. (A) Schematic representation of the constructs used to express miRNAs along with a GFP reporter. (B) and (C) CHOIgG1 were transiently transfected with the miRNA expression vectors, control vector (pcDNA6.2-GW/emGFP-control) or the respective mature miRNA (hsa-miR-557 or hsamiR-1287). Two days post transfection (for vector transfection) or one day post transfection (for mature miRNA transfection), RNA was extracted and levels of the miRNAs hsa-miR-557 (B) and hsa-miR-1287 (C) were measured by qRT-PCR analysis. (D)–(F) CHO-IgG1 cells were stably transfected with the indicated miRNA expression constructs and sorted for GFP. (D) The sorted pools were cultured for 8 and 51 days, respectively, and analyzed by flow cytometry. As a control parental cells were used (filled peak). miRNA expression in GFP-sorted pools was analyzed by qRT-PCR for hsa-miR-557 (E) and hsa-miR-1287 (F). As negative controls, parental cells or cells stably expressing the neg. control miRNA were used. In all cases, the relative expression was calculated by normalization to RNU6B using the ΔCq method (n¼ 3, error bars represent SEM).

control miRNA were generated. These pools were sorted for GFP as described and miRNA overexpression was verified by qRT-PCR (data not shown). Stable miRNA overexpressing CHO-IgG1 pools as well as the parental cells were cultivated in shake flasks with daily feeding to mimic conditions of fed-batch cultivation. Samples were taken from day 3 on to determine antibody concentrations in the supernatant, cell growth, viability, and specific productivity (see Fig. 4A). The negative control miR-expressing CHO-IgG1 cells behaved similarly to the parental cells with respect to cell growth and viability. However, slightly decreased product titers and specific productivities compared to parental cells were observed. Strikingly, miR557–miR1287 overexpression resulted in a significant increase in product concentration of approximately 1.3 fold on day 7 compared to the negative control and the parental cells. Growth and viability of miR557–miR1287 cells were similar to parental and negative control miRNA expressing cells with a marginal tendency to accelerated cell growth. This could be due to expression of hsa-miR-557 as observed in the transient experiments (Fig. 2C). Finally, specific productivity was also improved compared to the negative control miRNA expressing cells and the parental cells. To compare product quality of antibodies produced in miRNA engineered producer cells during fed-batch cultivation, the IgG1 antibody was purified from the supernatant of the stable cell pools. The integrity of the antibody was first analyzed by reducing and nonreducing SDS-PAGE. In all samples, the Coomassie-stained gel showed

the expected protein bands migrating at molecular weights corresponding to the intact IgG, and the light and heavy chains, respectively (Fig. 4B). Next, we analyzed binding to the IgG1 target expressed on MDA-MB468 cells by flow cytometry. As a control, antibody purified from the parental CHO-IgG1 cells was used. As shown in Fig. 4C, using non-saturating concentrations antibody binding to the cells was similar in all samples, indicating that the stable expression of miRNAs does not affect the binding properties of the IgG1 antibody. The isotype-matched IgG negative control did not bind to the MDA-MB468 cells proving specificity of the signal. Finally, we analyzed the glycosylation pattern of the purified antibody to confirm that the glycoprofile compared with the parental host cell line. The glycans were enzymatically released from the antibody and analyzed with microchip based separation by CGE. The glycosylation pattern of the antibody purified from CHO-IgG1 stably expressing miRNAs was comparable to the pattern observed in IgG1 produced in parental cells (Fig. 4D), showing that no aberrant glycoforms had been introduced. Taken together, we successfully established a functional miRNA screen in a CHO producer cell line that resulted in the identification of nine miRNAs of potential interest to cell line engineering approaches aiming at boosting the productivity of mammalian production cell lines. Specifically, we could show that the stable co-expression of hsa-miR-557 and hsa-miR-1287 in an IgG producing CHO cell line significantly improved the specific productivity

164

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

Fig. 4. Stable miRNA expression enhances the productivity of CHO-IgG1 cells in fed-batch cultures. CHO-IgG1 cells stably co-expressing miR-557 and miR-1287, the negative control miRNA or parental CHO-IgG1 were subjected to fed-batch cultivation. (A) Shown is the antibody concentration in the supernatant, the cell density, cell viability and specific productivity. The data are from one representative experiment performed in duplicate in shake flasks using three independent stable CHO-IgG1 pools co-expressing miR-557 and miR-1287 and four independent pools expressing the negative control miRNA. Statistical analysis was done using two-way ANOVA (po 0.01; negative control vs. miR557–miR1287) followed by a Bonferroni post-test (nn: p o0.01 on day 7). Error bars represent SEM. The experiment was repeated three times with similar results. (B) Antibodies purified from the indicated cell supernatants were analyzed by Coomassie blue staining after separation by reducing (left) and non-reducing (right) SDS-PAGE. Arrows indicate the antibody-specific protein band(s). (C) Antibodies purified from the indicated cell supernatants were incubated with surface antigen expressing MDAMB468 cells and antigen binding was analyzed by flow cytometry. Unstained cells (grey) and an IgG isotype control were used as negative controls. (D) Glycans were released from the purified antibodies and analyzed by microchip-based CGE separation. Shown is the relative amount of the different glycans normalized to 100%.

and product yields in fed-batch culture whilst maintaining product quality, providing proof-of-concept for improved performance of miR-engineered cells in biopharmaceutical applications.

4. Discussion In the past years miRNAs have been discussed as an attractive tool for cell line engineering approaches of mammalian producer cells such as the widely used CHO cells. Despite intense sequencing efforts and

various profiling studies performed in CHO cells (Hackl et al., 2011; Hammond et al., 2012; Johnson et al., 2011; Lin et al., 2011) thus far, only a few examples link the expression of a specific miRNA to a desired cellular phenotype. For example, in response to nutrient depletion, miR-466h was found to be strongly upregulated (Druz et al., 2012). The transient inhibition of miR-466h enhanced the expression of several anti-apoptotic genes, leading to increased cell viability and decreased caspase activation in nutrient-depleted medium, implicating a function for this miRNA in the apoptotic response. In a second study, miR-7 was found to be down-regulated in response

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

to low temperature, hinting at a role for miR-7 in cell cycle regulation (Barron et al., 2011). However, inhibition of endogenous miR-7 using an antagomir failed to affect cell growth. This highlights the difficulty of selecting specific miRNAs for the genetic manipulation of host cells merely based on expression data. Here we describe a functional screen aiming at the identification of miRNAs that impact the productivity of an IgG secreting CHO producer cell line. Due to the lack of a commercially available hamster miRNA library, we used a human miRNA library in our screen. Recently, next-generation sequencing of small RNAs isolated from CHO cells enabled the annotation of 387 mature hamster miRNAs of which nearly 60% have human orthologs, revealing a high degree of conservation between the species (Hackl et al., 2011). The human miRNAs expressed in CHO cells are thus likely to function by utilizing the endogenous machinery to regulate physiologically relevant targets. Our transient miRNA screen identified miRNAs that significantly increased or decreased the productivity of CHO-IgG1 cells by at least 1.5-fold on days 3 and/or 4 post transfection. Five of the 14 miRNAs that reduced antibody yields are members of two miRNA families (miR-302 and miR-518n), suggesting that these miRNAs suppress the productivity of cells via common underlying mechanisms. Importantly, nine miRNAs significantly increased productivity in the primary screen, all of which were confirmed to have a positive impact on CHOIgG1 antibody titers and specific productivity in the secondary screen. The highest yields were observed for hsa-miR-557, hsamiR-612, and hsa-miR-1287. Interestingly, all of the nine miRNAs also positively impacted the specific productivity of a CHO cell pool stably secreting recombinant HSA, indicating that these miRNAs act both clone and product independently. While the miRNA concentrations used in the validation screen negatively impacted cell growth (see Fig. 2C), half the concentration had negligible effects on cell growth (see Fig. 2D). Furthermore, the growth rate of cells stably expressing miR-557 and miR-1287 was similar to that of the parental and vector control cells, indicating that the levels of ectopically expressed miRNA do not impose a metabolic burden. Although the levels of stably expressed miR-557 and miR-1287 were 100-fold lower compared to transient transfections, the performance of these cells improved in fed batch cultures. This indicates that relatively modest miRNA expression levels may be sufficient to confer alterations of cellular phenotypes. For industrial applications, stable cell clones expressing the recombinant protein of interest are routinely generated. Because clonal variations are very common, it is necessary to screen a large number of clones to verify the correlation of any phenotypic changes with the genetic modification introduced. Here, we used independently generated stable pools enriched for miRNA expression by FACS sorting, based on the presence of a GFP cassette contained within the transcript. The productivity of these cells therefore reflects the mean productivity of numerous clones. It is thus conceivable that a subset of clones isolated from these populations will present with even higher productivity. From the validated miRNAs, we selected miR-557 due to its positive effects on the viable cell number and miR-1287, which had the strongest impact on the specific productivity of cells. By qPCR, no endogenous expression of miR-557 or miR-1287 was detected in CHO cells under basal conditions, in line with a previous report, which did not identify the corresponding hamster miRNAs (Hackl et al., 2012b). While the single stable expression of miR-557 or miR1287 did not improve the performance of CHO pools in fed batch cultures (data not shown), the combined stable expression of both miRNAs increased the specific productivity of cells and resulted in a higher overall yield. Although there are a few reports on altered expression of miR-557 (Chen et al., 2013; Katayama et al., 2012; Mosakhani et al., 2012) and miR-1287 (Guo et al., 2012; Wang et al., 2012) in different cancer types, no specific targets have been

165

identified to date. Target prediction algorithms give rise to a high number of theoretical targets, which generally only poorly overlap with the targets identified experimentally. Therefore, to elucidate the molecular mechanism by which these miRNAs impact cell productivity, analysis of transcriptome changes will be required. However, due to the promiscuous nature of miRNAs, it is very likely that the down-regulation of a single target mRNA will not be sufficient to account for the enhanced viable cell number and specific productivity induced by the miRNA. It is rather the simultaneous modulation of several targets that act in concert (Selbach et al., 2008; Serva et al., 2012) to support a hyperproductivity phenotype of cells. This first functional miRNA screen in CHO producer cells was performed in a model cell line that was selected based on its good productivity. Our results thus demonstrate that cellular productivity is not at its limit but can be boosted even further. We have previously shown that the stable expression of the lipid transfer protein CERT, which functions at the level of the Golgi complex to ensure constitutive vesicular trafficking to the plasma membrane, can further augment the specific productivity of CHO cells secreting an IgG1 or HSA (Florin et al., 2009). This indicates that high producer CHO cell lines may have a bottle-neck at the level of protein secretion. It is possible that miR-1287 acts in part by alleviating this bottle-neck, through the modulation of protein synthesis and transport, thereby increasing the specific productivity of cells. Based on the protocol established here, further screens can now be tailored to identify candidate miRNAs that positively impact additional parameters relevant to the bioprocess, such as stress tolerance or metabolic disorders of cells. The fact that different miRNAs can be encoded in tandem or more copies by a single vector, analogous to the clustered genomic organization of miRNAs, provides an unlimited tool-box that can be designed to meet the needs of the particular bioprocess. In summary, our data support the notion that the optimization of mammalian host cells based on miR-engineering is a promising strategy toward the more efficient and affordable production of recombinant proteins for future therapeutic applications.

Acknowledgments We are grateful to Peter Scheurich (University of Stuttgart) for FACS support. Monilola A. Olayioye is supported by a DFG Heisenberg grant.

Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.ymben.2013.10.005. References Barron, N., Kumar, N., Sanchez, N., Doolan, P., Clarke, C., Meleady, P., O'Sullivan, F., Clynes, M., 2011. Engineering CHO cell growth and recombinant protein productivity by overexpression of miR-7. J. Biotechnol 151, 204–211. Bartel, D.P., 2009. MicroRNAs: target recognition and regulatory functions. Cell 136, 215–233. Becker, E., Florin, L., Pfizenmaier, K., Kaufmann, H., 2008. An XBP-1 dependent bottle-neck in production of IgG subtype antibodies in chemically defined serum-free Chinese hamster ovary (CHO) fed-batch processes. J. Biotechnol 135, 217–223. Becker, E., Florin, L., Pfizenmaier, K., Kaufmann, H., 2010. Evaluation of a combinatorial cell engineering approach to overcome apoptotic effects in XBP-1 (s) expressing cells. J. Biotechnol 146, 198–206. Birmingham, A., Selfors, L.M., Forster, T., Wrobel, D., Kennedy, C.J., Shanks, E., SantoyoLopez, J., Dunican, D.J., Long, A., Kelleher, D., Smith, Q., Beijersbergen, R.L., Ghazal, P., Shamu, C.E., 2009. Statistical methods for analysis of high-throughput RNA interference screens. Nat. Methods 6, 569–575.

166

M. Strotbek et al. / Metabolic Engineering 20 (2013) 157–166

Chen, L., Li, Y., Fu, Y., Peng, J., Mo, M.H., Stamatakos, M., Teal, C.B., Brem, R.F., Stojadinovic, A., Grinkemeyer, M., McCaffrey, T.A., Man, Y.G., Fu, S.W., 2013. Role of deregulated microRNAs in breast cancer progression using FFPE tissue. PLoS One 8, e54213. Dreesen, I.A., Fussenegger, M., 2011. Ectopic expression of human mTOR increases viability, robustness, cell size, proliferation, and antibody production of chinese hamster ovary cells. Biotechnol. Bioeng 108, 853–866. Druz, A., Betenbaugh, M., Shiloach, J., 2012. Glucose depletion activates mmu-miR466h-5p expression through oxidative stress and inhibition of histone deacetylation. Nucleic Acids Res 40, 7291–7302. Fabian, M.R., Sonenberg, N., Filipowicz, W., 2010. Regulation of mRNA translation and stability by microRNAs. Annu. Rev. Biochem 79, 351–379. Florin, L., Pegel, A., Becker, E., Hausser, A., Olayioye, M.A., Kaufmann, H., 2009. Heterologous expression of the lipid transfer protein CERT increases therapeutic protein productivity of mammalian cells. J. Biotechnol 141, 84–90. Gammell, P., Barron, N., Kumar, N., Clynes, M., 2007. Initial identification of low temperature and culture stage induction of miRNA expression in suspension CHO-K1 cells. J. Biotechnol 130, 213–218. Guo, L., Zhao, Y., Yang, S., Cai, M., Wu, Q., Chen, F., 2012. Genome-wide screen for aberrantly expressed miRNAs reveals miRNA profile signature in breast cancer. Mol. Biol. Rep. Hackl, M., Borth, N., Grillari, J., 2012a. miRNAs—pathway engineering of CHO cell factories that avoids translational burdening. Trends Biotechnol 30, 405–406. Hackl, M., Jadhav, V., Jakobi, T., Rupp, O., Brinkrolf, K., Goesmann, A., Puhler, A., Noll, T., Borth, N., Grillari, J., 2012b. Computational identification of microRNA gene loci and precursor microRNA sequences in CHO cell lines. J. Biotechnol 158, 151–155. Hackl, M., Jakobi, T., Blom, J., Doppmeier, D., Brinkrolf, K., Szczepanowski, R., Bernhart, S.H., Honer Zu Siederdissen, C., Bort, J.A., Wieser, M., Kunert, R., Jeffs, S., Hofacker, I.L., Goesmann, A., Puhler, A., Borth, N., Grillari, J., 2011. Nextgeneration sequencing of the Chinese hamster ovary microRNA transcriptome: Identification, annotation and profiling of microRNAs as targets for cellular engineering. J. Biotechnol 153, 62–75. Hammond, S., Swanberg, J.C., Polson, S.W., Lee, K.H., 2012. Profiling conserved microRNA expression in recombinant CHO cell lines using Illumina sequencing. Biotechnol. Bioeng 109, 1371–1375. Hernandez Bort, J.A., Hackl, M., Hoflmayer, H., Jadhav, V., Harreither, E., Kumar, N., Ernst, W., Grillari, J., Borth, N., 2012. Dynamic mRNA and miRNA profiling of CHO-K1 suspension cell cultures. Biotechnol. J 7, 500–515. Iorio, M.V., Croce, C.M., 2012. MicroRNA dysregulation in cancer: diagnostics, monitoring and therapeutics. A comprehensive review. EMBO Mol. Med 4, 143–159. Jadhav, V., Hackl, M., Bort, J.A., Wieser, M., Harreither, E., Kunert, R., Borth, N., Grillari, J., 2012. A screening method to assess biological effects of microRNA overexpression in Chinese hamster ovary cells. Biotechnol. Bioeng 109, 1376–1385. Jayapal, K., Wlaschin, K., Hu, W.-S., Yap, M., 2007. Recombinant protein therapeutics from Cho cells—20 years and counting. CHO Consortium: SBE Spec. Ed, 40–47. Johnson, K.C., Jacob, N.M., Nissom, P.M., Hackl, M., Lee, L.H., Yap, M., Hu, W.S., 2011. Conserved microRNAs in Chinese hamster ovary cell lines. Biotechnol. Bioeng 108, 475–480. Katayama, Y., Maeda, M., Miyaguchi, K., Nemoto, S., Yasen, M., Tanaka, S., Mizushima, H., Fukuoka, Y., Arii, S., Tanaka, H., 2012. Identification of

pathogenesis-related microRNAs in hepatocellular carcinoma by expression profiling. Oncol. Lett 4, 817–823. Ku, S.C., Ng, D.T., Yap, M.G., Chao, S.H., 2008. Effects of overexpression of X-box binding protein 1 on recombinant protein production in Chinese hamster ovary and NS0 myeloma cells. Biotechnol. Bioeng 99, 155–164. Le Fourn, V., Girod, P.A., Buceta, M., Regamey, A., Mermod, N., 2013. CHO cell engineering to prevent polypeptide aggregation and improve therapeutic protein secretion. Metab. Eng. Lee, R.C., Ambros, V., 2001. An extensive class of small RNAs in Caenorhabditis elegans. Science 294, 862–864. Lim, S.F., Chuan, K.H., Liu, S., Loh, S.O., Chung, B.Y., Ong, C.C., Song, Z., 2006. RNAi suppression of Bax and Bak enhances viability in fed-batch cultures of CHO cells. Metab. Eng 8, 509–522. Lin, N., Davis, A., Bahr, S., Borgschulte, T., Achtien, K., Kayser, K., 2011. Profiling highly conserved microrna expression in recombinant IgG-producing and parental Chinese hamster ovary cells. Biotechnol. Progr 27, 1163–1171. Mosakhani, N., Guled, M., Leen, G., Calabuig-Farinas, S., Niini, T., Machado, I., Savola, S., Scotlandi, K., Lopez-Guerrero, J.A., Llombart-Bosch, A., Knuutila, S., 2012. An integrated analysis of miRNA and gene copy numbers in xenografts of Ewing's sarcoma. J. Exp. Clin. Cancer Res.: CR 31, 24. Pace, C.N., Vajdos, F., Fee, L., Grimsley, G., Gray, T., 1995. How to measure and predict the molar absorption coefficient of a protein. Protein Sci.: Publ. Protein Soc 4, 2411–2423. Peng, R.W., Fussenegger, M., 2009. Molecular engineering of exocytic vesicle traffic enhances the productivity of Chinese hamster ovary cells. Biotechnol. Bioeng 102, 1170–1181. Peng, R.W., Guetg, C., Tigges, M., Fussenegger, M., 2010. The vesicle-trafficking protein munc18b increases the secretory capacity of mammalian cells. Metab. Eng 12, 18–25. Reinhart, B.J., Slack, F.J., Basson, M., Pasquinelli, A.E., Bettinger, J.C., Rougvie, A.E., Horvitz, H.R., Ruvkun, G., 2000. The 21-nucleotide let-7 RNA regulates developmental timing in Caenorhabditis elegans. Nature 403, 901–906. Selbach, M., Schwanhausser, B., Thierfelder, N., Fang, Z., Khanin, R., Rajewsky, N., 2008. Widespread changes in protein synthesis induced by microRNAs. Nature 455, 58–63. Serva, A., Knapp, B., Tsai, Y.T., Claas, C., Lisauskas, T., Matula, P., Harder, N., Kaderali, L., Rohr, K., Erfle, H., Eils, R., Braga, V., Starkuviene, V., 2012. miR-17-5p regulates endocytic trafficking through targeting TBC1D2/Armus. PLoS One 7, e52555. Tigges, M., Fussenegger, M., 2006. Xbp1-based engineering of secretory capacity enhances the productivity of Chinese hamster ovary cells. Metab. Eng 8, 264–272. Urlaub, G., Mitchell, P.J., Kas, E., Chasin, L.A., Funanage, V.L., Myoda, T.T., Hamlin, J., 1986. Effect of gamma rays at the dihydrofolate reductase locus: deletions and inversions. Som. Cell Mol. Genet 12, 555–566. Wang, W., Corrigan-Cummins, M., Hudson, J., Maric, I., Simakova, O., Neelapu, S.S., Kwak, L.W., Janik, J.E., Gause, B., Jaffe, E.S., Calvo, K.R., 2012. MicroRNA profiling of follicular lymphoma identifies microRNAs related to cell proliferation and tumor response. Haematologica 97, 586–594. Wurm, F.M., 2004. Production of recombinant protein therapeutics in cultivated mammalian cells. Nat. Biotechnol 22, 1393–1398. Zuker, M., 2003. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res 31, 3406–3415.