ORIGINAL RESEARCH published: 19 June 2020 doi: 10.3389/fbioe.2020.00617 An Alternative Perfusion Approach for the Intensification of Virus-Like Particle Production in HEK293 Cultures Jesús Lavado-García*, Laura Cervera and Francesc Gòdia Grup d’Enginyeria Cellular i Bioprocés, Escola d’Enginyeria, Universitat Autònoma de Barcelona, Barcelona, Spain Virus-like particles (VLPs) have gained interest over the last years as recombinant vaccine formats, as they generate a strong immune response and present storage and distribution advantages compared to conventional vaccines. Therefore, VLPs are being regarded as potential vaccine candidates for several diseases. One requirement for their further clinical testing is the development of scalable processes and production platforms for cell-based viral particles. In this work, the extended gene expression Edited by: Maizirwan Mel, (EGE) method, which consists in consecutive media replacements combined with cell International Islamic University retransfections, was successfully optimized and transferred to a bioreactor operating in Malaysia, Malaysia perfusion. A process optimization using design of experiments (DoE) was carried out Reviewed by: to obtain optimal values for the time of retransfection, the cell specific perfusion rate Mohd Azmir Arifin, Universiti Malaysia Pahang, Malaysia (CSPR) and transfected DNA concentration, improving 86.7% the previously reported Audrey Yi-Hui Teh, EGE protocol in HEK293. Moreover, it was successfully implemented at 1.5L bioreactor St George’s, University of London, United Kingdom using an ATF as cell retention system achieving concentrations of 6.8·1010 VLP/mL. VLP *Correspondence: interaction with the ATF hollow fibers was studied via confocal microscopy, field emission Jesús Lavado-García scanning electron microscopy, and nanoparticle tracking analysis to design a bioprocess jesus.lavado@uab.cat capable of separating unassembled Gag monomers and concentrate VLPs in one step. Specialty section: Keywords: bioreactor, perfusion, ATF, design of experiments, VLP, HFM This article was submitted to Bioprocess Engineering, a section of the journal INTRODUCTION Frontiers in Bioengineering and Biotechnology Since the human immunodeficiency virus 1 (HIV-1) was first reported in 1959, 74.9 million Received: 16 April 2020 people have been infected (Whiteside and Wilson, 2018). Due to the high impact and Accepted: 20 May 2020 importance of this disease, many efforts have been devoted to understand and characterize Published: 19 June 2020 the mechanisms taking place in HIV infection and replication (von Schwedler et al., 2003; Citation: Pincetic and Leis, 2009; Meng and Lever, 2013; Patters and Kumar, 2018). For replication Lavado-García J, Cervera L and and virion production, HIV-1 Gag polyprotein play a key role (Bell and Lever, 2013). Upon Gòdia F (2020) An Alternative expression of this polyprotein, it self-assembles forming the virion protein core which can Perfusion Approach for the Intensification of Virus-Like Particle be later packed with the genetic material and bud off the host cell (von Schwedler et al., Production in HEK293 Cultures. 2003; Jolly and Sattentau, 2007; Meng and Lever, 2013). This characteristic can be exploited Front. Bioeng. Biotechnol. 8:617. to obtain protein core structures comprising Gag monomers and lacking genetic material. The doi: 10.3389/fbioe.2020.00617 use of Gag to generate virus-like particles (VLPs) has become a promising option to develop Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 1 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production alternative platforms for recombinant vaccines being able to product is being constantly harvested and operation times target several diseases (Lua et al., 2014; Cervera et al., 2019). extended. The fact of achieving higher cell densities in shorter Having multimeric structures such as Gag-VLPs helps induce times was one of the issues encountered in the first attempt a stronger immune response (Meador, 2017) and since they of transferring the EGE protocol to a bioreactor operated in are produced by a budding process, the lipid envelope can be continuous mode without further optimization (Fuenmayor further used for particle pseudotyping (Kueng et al., 2007, 2011; et al., 2019). Indeed, cells need to be maintained around 2·106 Sharma et al., 2017). The addition of conjugated antigens can cells/mL for an efficient transfection. Higher cell densities are be used to target different diseases, such as influenza (Venereo- more difficult to transfect due to the widely described cell density Sanchez et al., 2017). The use of mammalian cell cultures to effect (CDE) (Bernal et al., 2009; Le Ru et al., 2010; Petiot express HIV-Gag in order to produce VLPs offers the advantage et al., 2015). Transfection efficiency and productivity decrease of providing the necessary cellular machinery to produce when performed at higher cell densities. The energetic demand enveloped particles via membrane budding and performing is solved with the constant addition of fresh media but the the correct post-translational modification (Bandaranayake and CDE comprises molecular causes that cannot be solved only Almo, 2014). One methodology to express heterologous proteins by perfusion (Petiot et al., 2015). In this work, we propose like Gag, is transient gene expression (TGE) (Gutiérrez-Granados overcoming this problem by changing the approach of perfusion et al., 2018a). It consists in the introduction of a plasmid using cell retention devices, like alternating tangential flow carrying the gene encoding the protein of interest that will (ATF), not to achieve high cell densities, but to ensure the remain episomal in the nucleus, while it is being expressed. constant media replacement in order to maintain the cell culture This expression method allows production of relatively high in a re-transfectable state for a longer period of time in the titers of protein in a short period of time, reduces time costs bioreactor and therefore, prolong production. This perfusion related to the generation of a stable cell line and offers high approach is not intended to operate at steady state and cell versatility when developing a production platform regardless bleeding is not necessary, as continuous perfusion is used to of the final product, as only the plasmid of interest needs to constantly renew media while cells are retransfected to keep be changed for a new product. Nevertheless, this method is a high proportion of cells producing VLPs, up to a certain limited in its timespan, since the plasmid is not integrated in limited process span. To achieve a scalable optimization of the the cell genome and upon consecutive rounds of cell division EGE protocol, key parameters important to improve production it is eventually lost. To overcome this issue and envisioning a yields were considered. The time of retransfection, the cell future step toward industrialization, a protocol named Extended specific perfusion rate (CSPR), and the DNA concentration Gene Expression (EGE) was designed allowing to advance added in each transfection were the three variables optimized from a typical harvest at 72 h post transfection (hpt) to a using a Design of Experiments (DoE) method. VLP production continuous harvest up to 240 hpt (Cervera et al., 2015). The was assessed and a new approach to develop the EGE EGE protocol increased protein production up to 12-fold and protocol at bioreactor scale using ATF is suggested. This consisted of a combination of media replacements every 48 h intensified and optimized procedure can reduce costs and and two retransfections at 48 and 96 hpt. The fact of having increase VLP production for a future potential implementation a constant rate of media replacement makes the EGE protocol at industrial scale, offering a production platform for novel suitable for continuous biomanufacturing using cell retention VLP-based vaccines. devices. Continuous production allows the constant renewal of nutrients and depletion of cell waste, toxic by-products or metabolites that might interfere with the culture homeostasis EXPERIMENTAL SECTION (Butler, 2005). As a result, higher volumetric productivity can be reached compared to batch and fed-batch culture. Indeed, HEK293 Mammalian Cell Line, the reduction of bioreactor size and facilities infrastructure, with Culture Conditions a reduction in production costs, is facilitating the adoption of The cell line used in this work is a serum-free suspension- continuous processes in the biopharmaceutical industry over adapted HEK 293 cell line (HEK293SF-3F6) kindly provided by the last few years (Gutiérrez-Granados et al., 2018b). One of Dr. Amine Kamen from McGill University (Montreal, Canada). the main advantages of continuous bioprocessing is that this Cells were cultured in disposable polycarbonate 125 mL flasks mode of operation allows the support of high cell density with vent cap (Corning R ) at 37◦ C, 5% of CO2 and 85% cultures, dramatically increasing volumetric productivity as the RH at 130 rpm in a LT-X Kuhner shaker (LT-X Kuhner, Birsfelden, Switzerland). Cell culture media was HyCellTM Abbreviations: ATF, Alternating tangential flow; CCD, Central composite design; TransFx-H media from HyCloneTM (GE Healthcare, Chicago, CDE, Cell density effect; CSPR, Cell specific perfusion rate; DoE, Design of IL, USA) supplemented with 4 mM GlutaMAXTM (Gibco, Life experiments; EGE, Extended gene expression; eGFP, enhanced green fluorescence Technologies, ThermoFisher, San Jose, CA, USA) and 0.1% protein; Gag::eGFP, translational fusion of HIV-Gag protein and eGFP; HFM, PluronicTM F-68 Non-ionic Surfactant (Gibco, Life Technologies, Hollow fiber module; Hpt, hours post transfection; PBS, Phosphate-buffered saline; ThermoFisher, San Jose, CA, USA). PEI, polyethyleneimine; PES, Polyethersulfone; RFU, Relative fluorescence units; RSM, Response surface methodology; FESEM, Field emission scanning electron Cell concentration and viability were determined using the microscopy; SF, Shake Flasks; SPR, Sequential Perfusion Recovery; STR, Stirred NucleoCounter R NC-3000 automatic cell counter (Chemometec, tank bioreactor; TGE, Transient gene expression; VLPs, Virus-like particles. Allerod, Denmark) according to manufacturer’s instructions. Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 2 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production For the pseudoperfusion experiments, the total culture for 15 min at room temperature and then added to the volume was 20 mL and media replacement (MR) was carried out cell culture. centrifuging the culture at 3000xg for 5 min every 12 h ensuring In the different series of perfusion experiments performed, that the proportional volume of media was replaced depending the first and second retransfections (t2 and t3 ) were carried on the condition. To maintain a MR rate of 2 reactor volume out following the previous protocol and incorporating some per day (RV/day), 20 mL were replaced every 12 h. For a rate of variations depending on each of the optimization experiments. 1 RV/day, 10 mL were replaced every 12 h and for a rate of 0.5 The specific conditions used in each case are detailed in the RV/day, 5 mL were replaced every 12 h. Results and Discussion section. Bioreactor Culture Conditions and Flow Cytometry Setup Description Samples were taken every 24 h after transfection and cells were A BioStat B Plus bioreactor (Sartorius Stedim Biotech, fixed using formaldehyde 2% for 10 min, centrifuged and then Goettingen, Germany) equipped with a 3-blade segment resuspended in PBS for FACS analysis. The percentage of GFP dual impeller with UP-DP configuration (Buffo et al., 2016) positive cells was assessed using a BD FACS Canto flow cytometer was used for HEK293 cell cultivation. The agitation was set (BD Biosciences, San Jose, CA, USA). Laser 488 was used for a 200 rpm. Temperature was set a 37◦ C. pH was set at 7.1 GFP measurement. The results were analyzed with FACS DIVA and controlled with CO2 and NaHCO3 (7.5%w/v). Dissolved software (BD Biosciences, San Jose, CA, USA). oxygen was controlled at 40% of air saturation by supplementing air by sparger at a constant flow of 0.1L/min and additional HIV-1 Gag VLP Quantification pure oxygen when needed. HEK293 growing exponentially in by Fluorimetry disposable polycarbonate 1 L shake flasks (Corning R ) were used The concentration of HIV-1 Gag VLPs was assessed by to seed the bioreactor at 0.5·106 cells/mL in 1.5 L. Samples were fluorimetry using a developed and validated quantification assay taken every 24 h for cell counting and viability determination. (Gutiérrez-Granados et al., 2013). VLP containing supernatants After transfection, assessment of the percentage of GFP positive were recovered by cell culture centrifugation at 1000×g for cells and VLP quantification was also performed every 24 h. 5 min. Relative fluorescence unit values (RFU) were calculated Perfusion was achieved using an alternating tangential flow by subtracting fluorescence unit (FU) values of non-transfected (ATF) cell retention device (Repligen, Waltham, MA, USA) with negative control samples. 0.5 and 0.2 µm pore size and 0.13 m2 of filtration area hollow fiber modules (Repligen, Waltham, MA, USA) and a ATF flow HIV-1 Gag VLP Quantification by rate of 0.6 L/min. When performing transfection, perfusion was stopped to incubate the cells with the DNA/PEI solution and Nanoparticle Tracking Analysis Nanoparticle Tracking Analysis (NTA) was also used to quantify reestablished 2 h after transfection (Cervera et al., 2017). To carry fluorescent particles. NTA measurements were performed with out media replacement, the filtration rate was set at 0.26 mL/min a NanoSight R LM20 device (NanoSight Ltd., Amesbury, UK) at the beginning of the process using a MasterFlex L/S peristaltic equipped with a blue laser module (488 nm) to quantify HIV- pump (MasterFlex Group, Gelsenkirchen, Germany) and was 1 Gag::eGFP VLPs and neutral density filter for total particle modified every day depending on the viable cell density to by light scattering. Data was analyzed with NanoSight R NTA maintain a cell specific perfusion rate (CSPR) of 30 pL/cell/day. 3.1 software. Briefly, samples were injected, and independent The bioreactor and the ATF system were placed over a scale analyses were carried out. Three video recordings of 60 s duration (MSE36200S-000-D0 Cubis, Sartorius, Goettingen, Germany) were taken for each sample. Capture settings were recorded to control the mass displacement created due to the filtration of spent media. The scale was connected to a Scilog peristaltic with an sCMOS camera (camera level of 8 for Gag::eGFP VLP pump (Scilog tandem 1081, Parker, Oxnard, CA, USA) which samples, and 11 for controls, viscosity: 0.9 cP) and analyzed with controlled the addition of fresh media upon detection of mass a detection threshold of 4. displacement to maintain constant volume in the bioreactor. Field Emission Scanning Electron Transient Transfection Microscopy (FESEM) Visualization The initial transfection (t1 ) was carried out at a cell density of Morphometry of the hollow fiber inner layer and fluorescence 2·106 cells/mL using a final DNA concentration of 1 µg/mL. detection at nanoscale were determined by Field Emission PEI/DNA complexes were formed by adding PEI to plasmid Scanning Electron Microscopy (FESEM). The analyzed samples DNA diluted in fresh culture media (10% of the total culture were 0.5 µm pore size hollow fiber samples exposed to non- volume to be transfected). Transfection reagent PEIpro R transfected HEK293 as a control, 0.5 and 0.2 µm pore size (Polyplus-transfection, Illkirch-Graffenstaden, France) was used. hollow fibers exposed to VLP-producing HEK293 cell cultures. The plasmid used contained the gene coding for HIV-Gag Hollow fibers were longitudinally cut in pieces of ∼2–3 mm2 protein fused to eGFP (Gag::eGFP). Briefly, the corresponding and deposited in carboncoated gold grids (200 mesh) during DNA was diluted with supplemented HyCellTM media and 1 min, air dried and observed in a FESEM Zeiss Merlin (Zeiss, vortexed for 10 s. Then PEI was added in 1:2 (w/w) DNA:PEI Jena, Germany) operating at 1.5 kV and 3.4 mm of working ratio and vortexed three times, then the mixture was incubated distance. Samples were then randomly checked with an in-lens Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 3 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production secondary electron detector for morphology and with a Back- TABLE 1 | Box-Behnken design, results and ANOVA analyses for optimization of scattered Electron (BSE) detector for fluorescence detection. the extended gene expression (EGE) protocol for VLP production. Representative images were obtained at a wide range of high −1 0 1 magnifications (from 200,000x to 500,000x). Retransfection time (hpt) 24 48 72 CSPR (pL/cell/day) 30 51.5 100 DNA (µg/mL)a 0.5 1.25 2 Confocal Microscopy Visualization Experimental Retransfection CSPR DNA/mL Psp The visualization of the hollow fiber modules was performed Run time using a FluoView R FV1000 confocal microscope (Olympus, Tokyo, Japan) at sampling speed of 2 µm/pixel, excitation at 1 1 0 −1 1.83E+09 488 nm and detection at 500–600 nm. Step size was 2 µm/slice 2 1 0 −1 2.20E+09 3 −1 0 −1 2.68E+09 using XYZ scan mode. Transversal and longitudinal cuts of the 4 −1 0 −1 3.29E+09 hollow fiber were made from samples of 0.5 µm pore size exposed 5 1 0 1 1.85E+09 to non-transfected HEK293 as a control and 0.5 and 0.2 µm 6 1 0 1 1.92E+09 pore size hollow fibers exposed to VLP-producing HEK293 cell 7 0 −1 −1 5.06E+09 cultures. For the longitudinal cuts, objective lens UPLAPO 20x, 8 0 −1 −1 4.33E+09 9 0 1 −1 2.32E+09 NA: 0.75 and optical zoom x3 was used. For the transversal 10 0 1 −1 2.36E+09 cuts, objective lens UPLAPO 10x, NA: 0.40 and optical zoom 1x 11 0 1 1 2.10E+09 was used. 12 0 1 1 2.47E+09 13 −1 1 0 4.64E+09 14 −1 1 0 2.91E+09 Optimization of Retransfection Parameters 15 1 −1 0 5.21E+09 16 1 0 5.56E+09 Using Design of Experiments 17 0 −1 0 0 2.07E+09 Retransfection parameters were optimized in order to maximize 18 0 0 0 1.92E+09 VLP specific productivity. The three variables chosen for 19 0 0 0 1.72E+09 optimization were the time of retransfection, the cell specific 20 0 0 0 1.89E+09 21 0 −1 1 5.19E+09 perfusion rate (CSPR) and the DNA concentration added in 22 0 −1 1 4.10E+09 each retransfection. The adjusted VLP specific productivity (Psp : 23 1 1 0 2.25E+09 VLP·106 cells−1 ·days−1 ) for the DoE response was calculated 24 1 1 0 1.94E+09 taking into account the volume of media replacement and the cell 25 −1 0 1 2.40E+09 culture viability in each day along the culture using the equation 26 −1 0 1 4.07E+09 27 −1 −1 0 5.65E+09 showed below (Equation 1): 28 −1 −1 0 5.19E+09 29 0 0 0 2.12E+09 Pd CVLP 30 0 0 0 2.38E+09 i=1 ( Cx KMRi vi ) Psp = (1) Model Multiple R2 p valueb Lack of fitc d  MR < 1 1 − MR 0.9126 1.06E-08 0.9316 KMR = RV MR MR RV (2) RV > 1 RV Parameters Co-efficient t Pb Constant 2.02E+09 10.225 <0.0001 where CVLP is the VLP concentration (VLP/mL), Cx is the [Time] −5.04E+08 −4.1761 0.0005 concentration of viable transfected cells (106 cells/mL), vi is [CSPR] −1.21E+09 −9.9874 <0.0001 the cell viability (%), d is the number of culture days after the [DNA] 1.88E+06 0.0155 0.988 first transfection and KMR is the media replacement coefficient, [Time]·[CSPR] −4.11E+08 −2.4077 0.026 [Time]·[DNA] −9.50E+07 −0.5562 0.584 proportional to the total volume replaced each day. KMR is [CSPR] · [DNA] −1.25E+06 −0.0073 0.994 calculated as shown in Equation 2, where MR is the replaced [Time]2 5.95E+08 3.3492 0.003 media and RV is the reactor volume. Thus, estimating the final [CSPR]2 1.56E+09 8.7562 <0.0001 volume at the time extracellular fluorescence was measured. A [DNA]2 −8.21E+07 −0.4617 0.649 Box-Behnken design was selected to define the optimal value for Optimal values Time of CSPR DNA Psp each variable. The three variables were screened at three levels: Retransfection a low level coded as −1, an intermediate level coded as 0, and −1 −1 0.5977154 5.49E+09 a high level coded as +1 as indicated in Table 1. Box-Behnken At 24hpt 30 1.7 µg/ml pL/cell/day experimental results were fitted to a second order polynomial equation described below (Equation 3) by non-linear regression a DNA/PEI ratio was always maintained at [1:2]. analysis: where Y is the response in VLP·Mcell−1 ·days−1 , β0 is bp values under 0.05 are considered statistically significant with 95% confidence, and p values under 0.1 are considered statistically significant with 90%. the offset term, βi is the linear coefficient, βii is the quadratic c p values associated to lack of fit test above 0.05 mean that the hypothesis arguing that coefficient, βij is the interaction coefficient, and Xi and Xj are the model is suitable cannot be rejected. the independent variables. This equation was used to predict the hpt: hours post transfection. Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 4 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production optimal values of the independent variables it can be observed that it tends to converge to a range going from 30 to 60 pL/cell/day range, suggesting that maintaining the media replacement rate within this interval is enough to maintain βii Xi2 + X X X Y = β0 + βi Xi + βij Xi Xj (3) cell growth. using R software (RStudio, Inc., Boston, MA, USA). The quality Optimization of Retransfection by Design of the fit of the model equation is expressed by the coefficient of Experiment (DoE) Method R2 and p-values obtained by regression analysis. Additionally, In order to optimize the EGE protocol, the continuous variables a lack of fit test was performed to compare the experimental time of retransfection, CSPR and the DNA concentration used error to the prediction error. The overall significance of the to retransfect, were selected. These variables were studied by model was determined by analysis of variance (ANOVA) F test, means of a design of experiments (DoE). A response surface whereas the significance of each coefficient was determined by methodology (RSM), specifically a Box-Behnken design, was the corresponding t test. selected as it allows to reduce number of experimental runs and yet obtain statistically relevant information (Montgomery, 2019). RESULTS AND DISCUSSION The parameters of the first transfection point had been already optimized in previous work and fixed in 2·106 cells/mL and Study of the Cell Growth Upon Different 1 µg/mL of DNA (Cervera et al., 2013; Fuenmayor et al., 2018). Media Replacement Rates The variables of [DNA:PEI] ratio, nature of the complexation Monitoring cell growth under three different media replacement agent and time for the maturation of the DNA:PEI complexes rates showed that maximum cell concentration, reaching 22·106 were also individually determined in previous works (González- cells/mL, was achieved in the shake flask replacing media at Domínguez et al., 2019a,b). The range of DNA was also 1RV/day and no difference was observed at 2RV/day (Figure 1A). selected based on this previous experience and most importantly, This suggests that 1RV/day is enough to meet the nutrient restrained by the PEI cytotoxicity. In addition to this, multiple demand of the culture and further cell growth could be variable optimization was selected for the presented study since hindered by other factors such as oxygen diffusion in the VLP production wanted to be tested combined with CSPR and flask. Cell growth monitoring was ceased when viability reached time of retransfection to take into account possible synergic 50% or below and no more data from that condition was effects and simulate the closest condition to the bioreactor. acquired. The observed fluctuations are due to the process The ratio [DNA:PEI] was fixed to [1:2] and maintained for all of centrifugation and resuspension, which added an intrinsic the experiments following the standard procedure. Regarding error to the monitoring process. However, this fluctuation does the time of retransfection, the working range was set from not affect the following CSPR calculation. When the CSPR is 24 to 72 hpt, based on the EGE protocol where the first calculated for every time point in each condition (Figure 1B) retransfection was performed at 48 hpt. As for the CSPR, the FIGURE 1 | Pseudoperfusion growth curves (A) HEK293 cell culture growth curves in shake flasks where media replacement was performed at a rate of 0.5, 1, or 2 reactor volume per day (RV/day) and a batch condition where no media replacement was carried out. Blue and red dotted lines represent 80 and 50% of cell viability, respectively. Experiment performed n = 3. (B) Representation of the cell specific perfusion rate (CSPR) from cell cultures with different rate of media replacement. Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 5 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production working range was set based on the previous study of the cell = 4) was performed using the set of parameters determined by growth upon different media replacement rates. The upper limit the optimal solution predicted from the model. A Psp of (5.7 was set at 100 and the lower limit at 30 pL/cell/day. Since the ± 0.8)·109 VLP·Mcell−1 ·days−1 was obtained, confirming the DNA concentration used for retransfection in the original EGE model adequacy. protocol was 0.5 µg/mL, the lower limit was set at this value. The upper limit for the DNA was set in 2 µg/mL due to the Optimization of the Second Retransfection cytotoxic effects of PEI (Cervera et al., 2013; Gutiérrez-Granados et al., 2016). Working ranges for each variable are presented Parameters Shake Flasks Study Following the EGE protocol and after optimizing the first in Table 1. A three-factor three-level Box-Behnken design was retransfection, a similar approach was performed to optimize constructed using the selected ranges for each variable as the the second retransfection. This time, based on the results from design space boundaries. A 15-experiment matrix was defined in the first optimization, a CSPR of 30 pL/cell/day was maintained which the central point was triplicated, and each experimental for all the experiments. In order to estimate the optimal values run was duplicated to assess the pure experimental error. This for the retransfection time and DNA concentration, a central led to a 30-experiment matrix (Table 1). Each condition was composite design (CCD) was used. The same working range retransfected at its corresponding time and harvested 72 h as in the Box-Behnken design was maintained for the DNA after. The response variable considered in this analysis was the and for the time of retransfection, the working range was set adjusted VLP specific productivity (Psp ) calculated as described from 24 to 72 h post retransfection. A 10-experiment matrix in Equation 1. This equation is proposed to optimize specific was defined, and each experimental run was duplicated to assess productivity taking into account two aspects of VLP production. the experimental error. In this case, the temporal proximity of A media replacement coefficient (KMR ) was added, reflecting the three transfections was so close that cell toxicity caused by how many times a reactor volume of media was replaced every PEI made unfeasible the completion of the experiment for the 12 h, making it directly proportional to the amount of replaced tested conditions. Therefore, no statistically significant model media to favor the conditions where the VLP concentration was obtained (data not shown). As this approach did not succeed in the supernatant was diluted due to a higher replacement to optimize the second retransfection, a further analysis of the rate. A term reflecting cell viability (v) was also added to culture behavior once transferred to the bioreactor was carried penalize conditions with low viability which are incrementing out as discussed further. extracellular relative fluorescent units (RFUs) due to the release of unassembled free monomer caused by cell death. This is one of the reasons why specific productivity was chosen over Intensification of the Optimized Protocol in total volumetric productivity as the DoE response to be based Bioreactor Using ATF on. Otherwise conditions reporting high fluorescence intensity, The ATF configuration previously described in the Methods but low cell densities and viabilities had been favored. Data section and illustrated in Figure 3A was implemented to apply was fitted to a second-order model by non-linear regression the optimal solution obtained using the Box-Behnken design analysis. The statistical significance of the non-linear regression to the stirred tank bioreactor (STR) culture operating in was confirmed by ANOVA analysis, also showed in Table 1. Data perfusion mode. Transfecting at 2·106 cells/mL at 0 h and from the model was used to construct response surface plots then retransfecting at 24 hpt reduced cell growth, making the where the interaction between factors can be observed (Figure 2). viable cell density (VCD) to remain stable at 3–4·106 cells/mL Noticeably, lower CSPRs led to higher adjusted specific up to 144 hpt (Figure 3B). Up to this point viability steadily productivities (Psp ) (Figures 2D–F). This condition allows to dropped from 98 to 50%, following the same trend in both minimize media consumption while increasing productivity. shake flasks (SF) and in the STR. From 144 hpt viability started Higher CSPRs led to the same cell densities at the moment to restore and cell growth recovered exponential rate until 264 of transfection, but the transfection percentage dropped faster hpt when the culture was stopped. In this phase of the culture, in these conditions as the subpopulation of non-transfected the STR reached 22–23·106 cells/mL compared to the 10·106 cells grew faster than the transfected one, losing the plasmid cells/mL observed in SF. Although the same CSPR was used more rapidly (Figures 2J–L). This can be explained as overall for both systems, in the STR, perfusion was automatically and cell homeostasis is disrupted upon transfection, reducing cell continuously operated while in SF a pseudoperfusion was carried growth and downregulating the cell cycle (Lavado-García et al., out, manually replacing the corresponding amount of media 2020). Regardless of the time of retransfection, decreasing CSPR every day. Operating in continuous mode proved to be more led to an increase in the specific productivity. It should also efficient as nutrients were continuously replenished and waste be considered that, due to cytotoxicity of the complexation removed. The percentage of transfection and the fluorescence agent (PEI), an increase in the DNA concentration compromises intensity were monitored daily. There was a peak in transfection viability. The model was used to predict an optimal combination at 96–120 hpt reaching 90% of transfected cells (Figure 3D) of the three factors to maximize Psp , reaching a value of 5.5·109 and a peak of production at 168 hpt reaching 115–120 RFUs VLP·Mcell−1 ·days−1 . The optimal set of parameters for the in the bioreactor (Figure 3E). RFU were converted to VLP first retransfection point was to perform it at 24 hpt using concentration using a quantification method by fluorimetry 1.7 µg/mL of DNA and maintaining a CSPR of 30 pL/cell/day. previously described Gutiérrez-Granados et al. (2013). Taking In order to validate the model, a confirmation experiment (n into account the VLP concentration in the STR, the harvest Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 6 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production FIGURE 2 | Response surface graphs based on Box-Behnken experimental results. Maximum VLP specific production in cell culture supernatants as a function of (A–C) CSPR vs. Time; (D–F) time vs. DNA; and (G–I) CSPR vs. DNA. The graphs were constructed by depicting two variables at a time and maintaining the third one at a fixed level. +1, 0 and −1 correspond to 0.5, 1.25, and 2 mg/mL for the DNA concentration; 30, 51.5 and 100 pL/cell/day for the CSPR and 24, 48, and 72 hpt for the time of retransfection. (J–L) represent the changes of the percentage of transfection measured in percentage of GFP positive cells along the studied time course for conditions reaching the sixth day of the process (J) and conditions reaching the fifth (K). (L) Shows the merged plot. Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 7 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production FIGURE 3 | (A) Schematic representation of the implemented ATF perfusion system. (B) Viable cell density and cell viability graphs along the studied time course. The blue and red dotted line indicate 80 and 50% of cell viability, respectively. Cells were transfected at time point 0 and 24 h post-transfection (hpt) noted by red arrows. (C) VLP specific productivity rate along the studied time course. (D) Flow cytometry analysis of the transfection percentage (showed in GFP positive percentage of cells) along the time course. (E) Measurements of fluorescence in relative fluorescent units (RFUs) from the reactor, the harvest and the shake flasks. (F) Total VLP concentration produced per day. Two different pore size for the hollow fiber module were used in the ATF system: 0.5 and 0.2 µm denoted as 0.5 and 0.2, respectively. (G) Total cumulative VLP concentration. Experiments performed n = 3. Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 8 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production and the corresponding CSPR, the total VLP production per replacement in order to prolong production a defined period day was calculated. As it can be observed in Figure 3F, the of time. system ceased producing VLPs at 168 hpt, indicating the 13 days process can be reduced to 9 days subsequently reducing time, VLP Retention by the ATF Hollow Fiber resources and costs. Cumulative VLP production (Figure 3G) Module achieved was 6.8·1010 VLPs/mL, improving in 86.7% or 7.54 For the ATF configuration, hollow fiber modules (HFM) fold the previous reported results (Cervera et al., 2015) using with pore sizes of 0.5 and 0.2 µm were used in order to the original EGE protocol in shake flasks. Comparing the compare their VLP recovery efficiency. Regardless of the HFM, EGE protocol carried out in bioreactor prior to the developed transfection efficiency and total VLP production remained optimization (Fuenmayor et al., 2019), the presented work constant (Figures 3F–G). However, despite having a diameter achieved a reactor and media volumetric productivity of 7.1·1012 of 140 nm, VLPs were clearly retained inside the bioreactor. and 8.7·1011 VLP·L−1 ·day−1 , respectively, improving 26.8% or Therefore, a detailed study of VLP retention by the HFM was 1.36 fold and 67.8% or 3.1 fold, respectively, the reported results. carried out. In order to characterize the interaction between VLPs This improvement led to the achievement of an average VLP and the HFM, confocal and field emission scanning electron specific productivity of 3,000 VLP·cell−1 ·day−1 (Figure 3C). The microscopy (FESEM) studies were performed. In these analyses same VLP concentration, specific productivity and trend in for characterization of VLP retention in the ATF system, a transfection efficiency were achieved in SF and STR, proving that sample from fibers exposed to conditioned media free of VLPs the optimized EGE protocol was successfully transferable to the from non-transfected HEK293 culture was used as a control. 1.5L bioreactor. Longitudinal cuts of the lumen of the hollow fibers from the three conditions were visualized under FESEM to evaluate the Optimization of the Second Retransfection morphology and presence of nanoparticles deposited on the fiber Parameters Bioreactor Study (Figure 5). Detection of secondary electrons (SE) is routinely After the difficulties of the first attempt of optimizing a used to evaluate size, size distribution and topography while second retransfection at small scale using pseudoperfusion, a detection of backscattered electrons (BSE) reveals differences in different approach was taken. Taking into account the trends chemical composition and allows the detection of nanoparticles in cell viability, a second retransfection following the standard (Kowoll et al., 2017; Österreicher et al., 2018) and substrates with protocol with 1 µg/mL of DNA, was carried out at 168 hpt. different composition deposited on the fiber surface. Contrast of At this point, viability surpassed 60% and started recovering. BSE for imaging has been widely used for metal nanoparticles Cells need to be growing to be efficiently transfected as they but it can also be used to detect fluorophores or molecules with need to be dividing for exogenous DNA to enter the nucleus electron excitation states whose electronic rearrangement after (Cervera et al., 2017). Therefore, at 168 hpt their recovering being exited under the electron beam, behave like a heavier atom. state would allow a new retransfection. In addition to this, BSE originate from the elastic scattering of primary electrons at 168 hpt VLP production ceased. At this moment the VCD upon interaction with the nucleus of the atoms in the sample. was 7–8·106 cells/mL. Figure 4 illustrates the different studied Therefore, atoms with heavier nuclei cause stronger scattering, process variables such as specific VLP production, percentage of providing a more intense signal for the BSE detector. Fluorescent transfection and total VLP production. Interestingly, this second molecules produce a different BSE signal compared to the non- retransfection had no effect. VCD and cell viability maintained fluorescent background (Vancová and Nebesárová, 2015; Seras- the trend previously observed and no additional changes in Franzoso et al., 2016; Garming et al., 2017; Fokkema et al., 2018). transfection and production were observed. As a result, a second The SEM analyses of the different fiber samples revealed that retransfection was discarded. This can be explained by the cell there is a deposit of particles in the fibers exposed to VLPs. density effect (CDE), which shows the difficulty to transfect cell The morphology of polyethersulfone (PES) fibers showed no cultures at high cell densities (Petiot et al., 2015). Indeed, these difference but the BSE analyses showed clusters of particles results confirmed the CDE, which was originally reported to be of different chemical composition and a higher BSE intensity related to energy depletion (Bereiter-Hahn et al., 1998; Bernal compared to the inert PES. In order to confirm the retention et al., 2009; Petiot et al., 2015). However, perfusion cultures are and deposit of VLPs in both 0.5 and 0.2 HFM, transversal and fed with a continuous addition of fresh media and removal of longitudinal cuts were also visualized using confocal microscopy metabolic waste, preventing energy depletion. Encountering the (Figure 6). The transversal cuts revealed that fluorescent deposits CDE in continuous cultures has been already reported (Genzel appeared in the luminal space of the fibers in the conditions et al., 2014; Fuenmayor et al., 2019) and different molecular exposed to VLPs. Interestingly, the longitudinal cuts analysis and metabolic factors have been described to influence the showed fluorescent deposits following the same clustering CDE (Lavado-García et al., 2020). Therefore, using perfusion to pattern previously detected using FESEM. Coherently, in both achieve very high cell densities, as implemented when working transversal and longitudinal cuts, the detected fluorescence with stable gene expression (SGE) cell-based platforms, is no intensity increase in the 0.2 µm HFM due to smaller pore and longer useful for bioprocesses using TGE. Instead, the ATF higher retention. Therefore, it is proved that VLPs are being perfusion system can be used to maintain a transfectable cell retained in the HFM forming clusters of particle aggregates density and perform perfusion to maintain the constant media of ∼2.5 and 5 µm of diameter in the 0.5 and 0.2 µm HFM, Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 9 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production FIGURE 4 | Second retransfection at 168 hpt in bioreactor. (A) Viable cell density and cell viability graphs along the studied time course. The blue and red dotted line indicate 80 and 50% of cell viability, respectively. Cells were transfected at time point 0, 24, and 168 h post-transfection (hpt) noted by red arrows. (B) VLP specific productivity rate along the studied time course. (C) Flow cytometry analysis of the transfection percentage (showed in GFP positive percentage of cells) along the time course. (D) Measurements of fluorescence in relative fluorescent units (RFUs) from the reactor, the harvest and the shake flasks. (E) Total VLP concentration produced per day (F) Total cumulative VLP concentration. respectively, with consistent results via confocal microscopy and when using ATF as the retention device for production of viral SEM. Nevertheless, the amount of deposited VLP was negligible particles. In the case of viruses like Influenza A which are about as VLP production in 0.2 and 0.5 conditions were equivalent 80–100 nm (Genzel et al., 2014; Coronel et al., 2019), modified (Figure 3G). The use of HFM has been previously reported to vaccinia Ankara (MVA) virus (Vázquez-Ramírez et al., 2018) lead to particle retention. Especially when manufacturing viral or even the production of yellow fever and Zika virus, both particles, as hollow fibers are used to concentrate VLPs (Negrete about 50 nm, recoveries < 1% of the virus titer in the harvest et al., 2014). Significant retention of viruses has been reported fraction have been reported (Nikolay et al., 2018). Therefore, VLP Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 10 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production FIGURE 5 | Visualization of a ATF hollow fiber inner side from a longitudinal cut using field emission scanning electron microscopy (FESEM). Rows represent the different studied conditions: control fibers used in non-producing-VLPs HEK293 cultures and fibers used in VLP-producing HEK293 cultures of 0.5 and 0.2 µm of pore diameter. Images were taken at 3.4 mm of working distance and 129,000x. The first column shows the morphology of the fibers using scanning of secondary electron (SE) detection. The second column shows back-scattered electron (BSE) detection for fluorescence detection from the same area. The third column shows the merge of the SEM image and the electron back scattering acquisition. retention observed in this work agrees with the already described different fractions and to determine the percentage of retention results of viral particle retention. (Figure 7A). The size distribution curve of the fluorescent particles in the bioreactor, regardless of the HFM used, described Quantification of VLP Retention particles of 140 nm and was maintained along the whole NTA analyses of samples from the bioreactor and from the bioprocess, confirming the stability of the particles. More harvest fraction were performed at different days to evaluate VLPs were retained in the 0.2 compared to the 0.5 µm HFM VLP quality and percentage of fluorescent particles in the along the process. The measurement of total particles revealed Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 11 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production FIGURE 6 | Visualization of ATF hollow fibers using confocal microscopy. Columns represent the different studied conditions: control fibers used in non-producing-VLPs HEK293 cultures and fibers used in VLP-producing HEK293 cultures of 0.5 and 0.2 µm of pore diameter. The first row shows transversal cuts of the fibers under visible light. The second row shows GFP fluorescence of the same transversal cut. The third row shows GFP fluorescence from a longitudinal cut of the luminal space. that, as expected, the 0.2 µm HMF was retaining a greater Comparing the value of VLP concentration obtained by total amount of particles inside the bioreactor compared to the fluorimetry-based and NTA methods of quantification, the 0.5 µm HFM. However, 12–15% of fluorescent particles ratios reflecting the presence of extracellular unassembled free was observed in the bioreactor in both cases. Surprisingly, fluorescent Gag monomers can be assessed. Samples with a high when measuring the fluorescent particles found in the harvest concentration of free monomer will provide a high fluorescence fractions, the concentration found was significantly lower than intensity signal while presenting a low particle concentration by expected. The size distribution curve showed an heterogeneous NTA. Inside de bioreactor, the observed ratio VLP:monomer is population of particles, representing only the 0.08–0.1% of around 1:4 while in the harvest fraction can reach up to 1:1762. the total particles found in the harvest, in both HFM used. These ratios are showed in Figure 7B. The free monomer ratio Calculations for both HFM revealed a 99% of VLP retention observed in the bioreactor did not depend on the HFM, as inside the bioreactor. the difference found comparing 0.5 and 0.2 µm HFM was not Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 12 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production FIGURE 7 | (A) Nano particle tracking analysis (NTA) from cultures where hollow fiber modules (HFM) with pore size of 0.5 and 0.2 µm were used. The left column shows the particle size distribution of the detected fluorescent particles (VLPs) along different days of the culture, in the reactor and harvest. The right column shows the particle size distribution of all particles detected (VLPs and extracellular vesicles) along different days in the reactor and harvest. Dotted blue lines indicate particle size of 140 nm. D labels indicate the time (day of the bioprocess) (B) Fold change ratio between the estimated particle concentration using fluorimetry and the detected actual particle concentration using NTA. This fold change suggests the presence of unassembled free Gag monomers. Significance calculated using two-way ANOVA, DF = 6 and n=3 in each condition. significant. On the other hand, the ratio found in the harvest Selection of Conditions for a VLP was around 103 and significantly changed depending on the Production Platform HFM used. The bigger pore of the 0.5 µm HFM facilitated Considering the results obtained in the previous sections, the filtration of free monomer together with VLPs compared the conditions for the implementation of the intensified and to the 0.2 µm HFM, reducing the ratio VLP/free monomer to optimized EGE method at bioreactor scale using an ATF system 1:816 while in the 0.2 µm HFM the ratio significantly increased for perfusion could be defined. The use of 0.2 µm HFM to 1:1762 due to the separation of VLP and free monomer provided the most advantages for the development of a VLP- via its filtration through the fiber. This can explain the low based production platform. The retention of the product within percentage of fluorescent particles found in the harvest and the reactor allowed the one-step production and concentration can be used for the direct separation of unassembled free Gag of VLPs, achieving 12–15% VLPs, compared to the total of monomer, providing substantial bioprocessing advantages as extracellular vesicles, increasing from the previously reported discussed further. 1% (Fuenmayor et al., 2019). Thus, eliminating the problem of Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 13 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production having the total amount of VLPs diluted in the harvest fraction, TABLE 2 | Optimized and intensified process performance. increasing media volumetric productivity 67.8% or 3.1 fold. The Accumulated VLPs 8.7E+13 problem of having the product retained might concern VLP Max. VLP concentration in a day (VLP/mL) 1.5E+10 stability but particles were proved stable at 37◦ C throughout VLP concentration in reactor (VLP/mL) 6.8E+10 the process, overcoming the stability issue. This might be due Mean VLP specific productiviy (VLP·cell−1 ·day−1 ) 3.0E+03 to the nature of enveloped particles. Biological membrane- Volumetric productivity / Reactor (VLP·L−1 ·day−1 ) 7.1E+12 bound extracellular vesicles, like exosomes and microvesicles, are Volumetric productivity / Spent medium (VLP·L−1 ·day−1 ) 2.7E+12 stable at physiological conditions such as temperature and pH Total volume of spent media (L) 3.8 (Madison and Okeoma, 2015; Ellwanger et al., 2017), equivalent to the bioreactor conditions. In addition to this, the 0.2 µm HFM allows the separation of unassembled free Gag monomer in the same step. Hence, improving the consecutive downstream processes not only by increasing the initial VLP concentration an approach based on the design of experiments method was regarding the total number of particles, such as exosomes applied to optimize the retransfection time, the CSPR and the or microvesicles, but also regarding the quality, separating DNA concentration. The optimal solution for the variables VLPs from free extracellular monomers. Therefore, overcoming of retransfection at 24 hpt, a CSPR of 30pL/cell/day and a challenges typically associated with VLP purification. However, DNA concentration of 1.7 µm g/mL were used to perform the the CDE prevented production from being extended in time, process in a 1.5L stirred tank bioreactor. The same volumetric not being able to prolong gene expression further than 168 productivity, specific productivity and VLP concentration were hpt, when cells ceased producing. However, product retention achieved in the parallel shake flask culture and the bioreactor. in the bioreactor is forcing to stop the process at a harvest The process was successfully operated at a 1.5L, reaching time point and to be restarted, as in batch culture mode. In a concentration of 6.8·1010 VLP/mL and a total cumulative order to operate continuously and not being restricted to the particle amount of 8.7·1013 VLPs in the bioreactor, improving batch process limitations, a continuous approach is proposed. the original EGE protocol 86.7% or 7.54 fold and the first Future work will be performed to demonstrate the validity of bioreactor approach using acoustic filter 67.8% or 3.1 fold. such an approach. When reaching the VLP harvest point at A summary of the process variables is presented in Table 2. 168 hpt, a bleeding step might be added to harvest most of Hollow fiber modules of 0.5 and 0.2 µm of pore size were tested the culture and maintain the corresponding amount of cells to when implementing the ATF perfusion system. A 99% of VLP seed the bioreactor again at 0.5·106 cells/mL with fresh media retention in the bioreactor was observed in both alternatives, and to recover the initial state, restarting the optimized EGE confirmed by FESEM, confocal microscopy and NTA analyses. transfection process. Bearing this in mind, 168 hpt proved to be the optimal VLP This novel perfusion approach is different to the conventional harvesting time from the bioreactor. Interestingly, the 0.2 µm use of perfusion for achieving high cell densities and maintaining HFM provided the advantage of separating unassembled Gag steady state through purge or bleeding. Here, continuous monomers from correctly assembled VLPs and concentrating perfusion is used to keep a constant media replacement while them in the bioreactor, facilitating the subsequent downstream bleeding would be used to recover the product every 168 processing. The use of the ATF system also allowed the hpt and continue the process. Although the recovered culture intensification of the process, increasing production while should undergo a clarification process to separate the cell reducing spent media. Although there is no previous work in mass from the VLPs in the supernatant, this method would VLP production by TGE using an ATF system for perfusion, the allow the acquirement of higher volumetric productivities while proposed platform reduce spent media from 9-10 to 3.8L and maintaining the constant media replacement. Thus, overcoming therefore resources and costs compared to the only additional the batch and fed-batch limitations, the implementation reported study regarding TGE at bioreactor scale (Hong et al., of Sequential Perfusion recovery (SPR) is the proposed 2019). Due to product retention, a proposed procedure of mode of operation. Conventional perfusion processes using controlled cell bleeding at 168 hpt to harvest and restart the ATF has been reported to successfully run for 50–80 days culture at an initial cell density of 0.5·106 is proposed under (Clincke et al., 2013; Zhang et al., 2015; Schwarz et al., the name of Sequential Perfusion Recovery (SPR). The cycle 2020). Therefore, the ATF membrane integrity should not be could be repeated as many times as needed to achieve the compromised for several 168-h cycles, depending on the desired desired titer. This strategy is an intensified and optimized VLP VLP titer. production platform that could encourage the potential further development VLP-based therapies, potentially applicable to several diseases. CONCLUSIONS The implementation of the extended gene expression protocol DATA AVAILABILITY STATEMENT has been achieved at bioreactor scale in perfusion mode circumventing limitations previously observed, such as a The original contributions presented in the unfavorable effect on the purity of the product. In this work, study are included in the article/supplementary Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 14 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production materials, further inquiries can be directed to the ACKNOWLEDGMENTS corresponding authors. The authors thank Dr. Amine Kamen (McGill University, AUTHOR CONTRIBUTIONS Montreal, Canada) for providing the HEK 293 SF-3SF6. The following reagent was obtained through the National Institutes JL-G designed and performed the experiments, analyzed and of Health AIDS Reagent Program, Division AIDS, National interpreted the data and wrote the manuscript. LC and Institute of Allergy and Infectious Diseases, National Institutes FG supervised the experiments and edited the manuscript. of Health: pGag::eGFP (Cat#11468) from Dr. Merilyn Resh. All authors contributed to the article and approved the The support of Dr. Jose Amable Bernabé from the Institute submitted version. of Material Science of Barcelona (ICMAB-CSIC, Barcelona, Spain) with particle tracking techniques and the help of FUNDING Alejandro Sánchez Chardi and Helena Montón Domingo from the Microscopy Service from Universitat Autònoma de The project that gave rise to these results received the support Barcelona (Barcelona, Spain) in the performance of scanning of a fellowship from la Caixa Foundation (ID 100010434). The electron microscopy and confocal microscopy, respectively, fellowship code is LCF/BQ/ES17/11600003. The research group whose analysis are greatly appreciated. Thanks to Juan Manuel is recognized as 2017 SGR 898 by Generalitat de Catalunya. Lora-Alonso for his help in editing the final figures. REFERENCES Coronel, J., Behrendta, I., Bürgin, T., Anderlei, T., Sandig, V., Reichlad, U., et al. (2019). Influenza A virus production in a single-use orbital shaken Bandaranayake, A. D., and Almo, S. C. (2014). Recent advances in mammalian bioreactor with ATF or TFF perfusion systems. Vaccine 37, 7011–7018. protein production. FEBS. Lett. 588, 253–260. doi: 10.1016/j.febslet.2013.11.035 doi: 10.1016/j.vaccine.2019.06.005 Bell, N. M., and Lever, A. M. (2013). HIV Gag polyprotein: processing Ellwanger, J. H., Veit, T. D., and Chies, J. A. B. (2017). Exosomes in HIV and early viral particle assembly. Trends Microbiol. 21, 136–144. infection: a review and critical look. Infect. Genet. Evolution 53, 146–154. doi: 10.1016/j.tim.2012.11.0060 doi: 10.1016/j.meegid.2017.05.021 Bereiter-Hahn, J., Munnich, A., and Woiteneck, P. (1998). Dependence of energy Fokkema, J., Fermie, J., Liv, N., van den Heuvel, D. J., Konings, T. O. M., Blab, metabolism on the density of cells in culture. Cell Struct. Funct. 23, 85–93. G. A., et al. (2018). Fluorescently labelled silica coated gold nanoparticles as doi: 10.1247/csf.23.85 fiducial markers for correlative light and electron microscopy. Sci. Rep. 8:13625. Bernal, V., Carinhas, N., Yokomizo, A. Y., Carrondo, M. J., and Alves, P. doi: 10.1038/s41598-018-31836-1 M. (2009). Cell density effect in the baculovirus-insect cells system: a Fuenmayor, J., Cervera, L., Gòdia, F., and Kamen, A. (2019). Extended gene quantitative analysis of energetic metabolism. Biotechnol. Bioeng. 104, 162–180. expression for Gag VLP production achieved at bioreactor scale. J. Chem. doi: 10.1002/bit.22364 Technol. Biotechnol. 94, 302–308. doi: 10.1002/jctb.5777 Buffo, M. M., Corrêaa, L. J., Esperançaa, M. N., Cruza, A. J. G., Farinasa, C. S., Fuenmayor, J., Cervera, L., Gutierrez-Granados, S., and Godia, F. (2018). Transient Badinoa, A. C., et al. (2016). Influence of dual-impeller type and configuration gene expression optimization and expression vector comparison to improve on oxygen transfer, power consumption, and shear rate in a stirred tank HIV-1 VLP production in HEK293 cell lines. Appl. Microbiol. Biotechnol. 102, bioreactor. Biochem. Eng. J. 114, 130–139. doi: 10.1016/j.bej.2016.07.003 165–174. doi: 10.1007/s00253-017-8605-x Butler, M. (2005). Animal cell cultures: recent achievements and perspectives in Garming, M. W. H., Weppelman, G. C., de Boer, P., Martínez, F. P., Schirhagl, the production of biopharmaceuticals. Appl. Microbiol. Biotechnol. 68, 283–291. R., Hoogenboom, J. P., et al. (2017). Nanoparticle discrimination based doi: 10.1007/s00253-005-1980-8 on wavelength and lifetime-multiplexed cathodoluminescence microscopy. Cervera, L., Gòdia, F., Tarrés-Freixas, F., Aguilar-Gurrieri, C., Carrillo, J., Nanoscale 9, 12727–12734. doi: 10.1039/C7NR00927E Blanco, J., et al. (2019). Production of HIV 1-based virus-like particles Genzel, Y., Vogel, T., Buck, J., Behrendt, I., Vazquez Ramirez, D., Schiedner, G., for vaccination: achievements and limits. Appl. Microbiol. Biotechnol. 103, et al. (2014). High cell density cultivations by alternating tangential flow (ATF) 7367–7384. doi: 10.1007/s00253-019-10038-3 perfusion for influenza A virus production using suspension cells. Vaccine 32, Cervera, L., Gonzalez-Dominguez, I., Segura, M. M., and Godia, F. (2017). 2770–2781. doi: 10.1016/j.vaccine.2014.02.016 Intracellular characterization of Gag VLP production by transient transfection González-Domínguez, I., Cervera, L., Gòdia, F., and Roldán, M. (2019a). of HEK 293 cells. Biotechnol. Bioeng. 114, 2507–2517. doi: 10.1002/bit. Quantitative colocalization analysis of DNA delivery by PEI-mediated cationic 26367 polymers in mammalian cells. J. Microsc. 273, 53–64. doi: 10.1111/jmi.12760 Cervera, L., Gutiérrez-Granados, S., Berrow, N. S., Segura, M. M., and Gòdia, F. González-Domínguez, I., Grimaldi, N., Cervera, L., Ventosa, N., and Gòdia, (2015). Extended gene expression by medium exchange and repeated transient F. (2019b). Impact of physicochemical properties of DNA/PEI complexes transfection for recombinant protein production enhancement. Biotechnol. on transient transfection of mammalian cells. N. Biotechnol. 49, 88–97. Bioeng. 112, 934–946. doi: 10.1002/bit.25503 doi: 10.1016/j.nbt.2018.09.005 Cervera, L., Gutiérrez-Granados, S., Martínez, M., Blanco, J., Gòdia, F., Mercedes Gutiérrez-Granados, S., Cervera, L., Gòdia, F., Carrillo, J., and Segura, Segura, M., et al. (2013). Generation of HIV-1 Gag VLPs by transient M. M. (2013). Development and validation of a quantitation assay for transfection of HEK 293 suspension cell cultures using an optimized fluorescently tagged HIV-1 virus-like particles. J. Virol Methods. 193, 85–95. animal-derived component free medium. J. Biotechnol. 166, 152–165. doi: 10.1016/j.jviromet.2013.05.010 doi: 10.1016/j.jbiotec.2013.05.001 Gutiérrez-Granados, S., Cervera, L., Kamen, A. A., and Gòdia, F. (2018a). Clincke, M. F., Mölleryd, C., Samani, P. K., Lindskog, E., Fäldt, E., Walsh, K., Advancements in mammalian cell transient gene expression (TGE) technology et al. (2013). Very high density of Chinese hamster ovary cells in perfusion by for accelerated production of biologics. Crit. Rev. Biotechnol. 38, 918–940. alternating tangential flow or tangential flow filtration in WAVE BioreactorTM - doi: 10.1080/07388551.2017.1419459 part II: applications for antibody production and cryopreservation. Biotechnol. Gutiérrez-Granados, S., Cervera, L., Segura, M., L., Wölfel, J., and Gòdia, F. Prog. 29, 768–777. doi: 10.1002/btpr.1703 (2016). Optimized production of HIV-1 virus-like particles by transient Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 15 June 2020 | Volume 8 | Article 617 Lavado-García et al. HEK293 Perfusion for VLP Production transfection in CAP-T cells. Appl. Microbiol. Biotechnol. 100, 3935–3947. Österreicher, J. A., Grabner, F., Schiffl, A., Schwarz, S., and Bourret, doi: 10.1007/s00253-015-7213-x G. R. (2018). Information depth in backscattered electron microscopy Gutiérrez-Granados, S., Gòdia, F., and Cervera, L. (2018b). Continuous of nanoparticles within a solid matrix. Mater. Charact. 138, 145–153. manufacturing of viral particles. Curr. Opin. Chem. Eng. 22, 107–114. doi: 10.1016/j.matchar.2018.01.049 doi: 10.1016/j.coche.2018.09.009 Patters, B. J., and Kumar, S. (2018). The role of exosomal transport Hong, J., Demirji, J., Blackstock, D., Lee, J., Dinh, T., Goh, A., et al. (2019). of viral agents in persistent HIV pathogenesis. Retrovirology 15:79. Development of an alternating tangential flow (ATF) perfusion-based transient doi: 10.1186/s12977-018-0462-x gene expression (TGE) bioprocess for universal influenza vaccine. Biotechnol. Petiot, E., Cuperlovic-Culf, M., Shen, C. F., and Kamen, A. (2015). Influence of Prog. 35:e2831. doi: 10.1002/btpr.2831 HEK293 metabolism on the production of viral vectors and vaccine. Vaccine Jolly, C., and Sattentau, Q. J. (2007). Human immunodeficiency virus 33, 5974–5981. doi: 10.1016/j.vaccine.2015.05.097 type 1 assembly budding and cell-cell spread in T cells take place in Pincetic, A., and Leis, J. (2009). The mechanism of budding of retroviruses from tetraspanin-enriched plasma membrane domains. J. Virol. 81, 7873–7884. cell membranes. Adv. Virol. 2009, 6239691–6239699. doi: 10.1155/2009/623969 doi: 10.1128/JVI.01845-06 Schwarz, H., Zhang, Y., Zhan, C., Malm, M., Field, R., Turner, R. et al. (2020). Kowoll, T., Müller, E., Fritsch-Decker, S., Hettler, S., Störmer, H., Weiss, Small-scale bioreactor supports high density HEK293 cell perfusion culture C., et al. (2017). Contrast of backscattered electron SEM images of for the production of recombinant Erythropoietin. J. Biotechnol. 309, 44–52. nanoparticles on substrates with complex structure. Scanning 2017:4907457. doi: 10.1016/j.jbiotec.2019.12.017 doi: 10.1155/2017/4907457 Seras-Franzoso, J., Sánchez-Chardi, A., Garcia-Fruitós, E., Vázquez, E., and Kueng, H. J., Leb, V. M., Haiderer, D., Raposo, G., Thery, C., Derdak, S. Villaverde, A. (2016). Cellular uptake and intracellular fate of protein V., et al. (2007). General strategy for decoration of enveloped viruses releasing bacterial amyloids in mammalian cells. Soft Matter. 12, 3451–3460. with functionally active lipid-modified cytokines. J. Virol. 81, 8666–8676. doi: 10.1039/C5SM02930A doi: 10.1128/JVI.00682-07 Sharma, J., Uchida, M., Miettinen, H. M., and Douglas, T. (2017). Modular Kueng, H. J., Schmetterer, K. G., and Pickl, W. F. (2011). Lipid rafts, interior loading and exterior decoration of a virus-like particle. Nanoscale 9, pseudotyping, and virus-like particles: Relevance of a novel, configurable, and 10420–10430. doi: 10.1039/C7NR03018E modular antigen-presenting platform. Int. Arch. Allergy Immunol. 154, 89–110. Vancová, M., and Nebesárová, J. (2015). Correlative fluorescence and scanning doi: 10.1159/000320224 electron microscopy of labelled core fucosylated glycans using cryosections Lavado-García, J., Jorge, I., Cervera, L., Vázquez, J., and Gòdia, F. (2020). mounted on carbon-patterned glass slides. PLoS ONE 10:e0145034. Multiplexed quantitative proteomic analysis of HEK293 provides insights of doi: 10.1371/journal.pone.0145034 molecular changes associated to the cell density effect, transient transfection Vázquez-Ramírez, D., Genzel, Y., Jordan, I., Sandig, V., and Reichl, U. (2018). and virus-like particles production. J. Proteome Res. acs.jproteome.9b00601. High-cell-density cultivations to increase MVA virus production. Vaccine 36, doi: 10.1021/acs.jproteome.9b00601 3124–3133. doi: 10.1016/j.vaccine.2017.10.112 Le Ru, A., Jacoba, D., Transfiguraciona, J., Ansorgea, S., Henry, O., Kamen, Venereo-Sanchez, A., Simoneau, M., Lanthier, S., Chahal, P., Bourget, L., Ansorge, A. A., et al. (2010). Scalable production of influenza virus in HEK- S., et al. (2017). Process intensification for high yield production of influenza 293 cells for efficient vaccine manufacturing. Vaccine 28, 3661–3671. H1N1 Gag virus-like particles using an inducible HEK-293 stable cell line. doi: 10.1016/j.vaccine.2010.03.029 Vaccine 35, 4220–4228. doi: 10.1016/j.vaccine.2017.06.024 Lua, L. H., Connors, N. K., Sainsbury, F., Chuan, Y. P., Wibowo, N., Middelberg, von Schwedler, U. K., Stuchell, M., Müller, B., Kaplan, J., Morham, S. G., A. P. J., et al. (2014). Bioengineering virus-like particles as vaccines. Biotechnol. et al. (2003). The protein network of HIV budding. Cell 114, 701–713. Bioeng. 111, 425–440. doi: 10.1002/bit.25159 doi: 10.1016/S0092-8674(03)00714-1 Madison, M. N., and Okeoma, C. M. (2015). Exosomes: Implications in HIV-1 Whiteside, A., and Wilson, D. (2018). Health and AIDS in 2019 and beyond. pathogenesis. Viruses 7, 4093–4118. doi: 10.3390/v7072810 African JAIDS Res. 17, iii–v. doi: 10.2989/16085906.2018.1550202 Meador, L. R., (2017). A heterologous prime-boosting strategy with Zhang, Y., Stobbe, P., Silvander, C. O., and Chotteau, V. (2015). Very high replicating Vaccinia virus vectors and plant-produced HIV-1 Gag/dgp41 cell density perfusion of CHO cells anchored in a non-woven matrix-based virus-like particles. Virology 507, 242–256. doi: 10.1016/j.virol.2017. bioreactor. J. Biotechnol. 213, 28–41. doi: 10.1016/j.jbiotec.2015.07.006 04.008 Meng, B., and Lever, A. M. (2013). Wrapping up the bad news: HIV assembly and Conflict of Interest: The authors declare that the research was conducted in the release. Retrovirology 10:5. doi: 10.1186/1742-4690-10-5 absence of any commercial or financial relationships that could be construed as a Montgomery, D. C. (2019). Design and Analysis of Experiments, 10th Edn. Wiley. potential conflict of interest. Negrete, A., Pai, A., and Shiloach, J. (2014). Use of hollow fiber tangential flow filtration for the recovery and concentration of HIV virus-like Copyright © 2020 Lavado-García, Cervera and Gòdia. This is an open-access article particles produced in insect cells. J. Virol. Methods 195, 240–246. distributed under the terms of the Creative Commons Attribution License (CC BY). doi: 10.1016/j.jviromet.2013.10.017 The use, distribution or reproduction in other forums is permitted, provided the Nikolay, A., Léon, A., Schwamborn, K., Genzel, Y., and Reichl, U. (2018). original author(s) and the copyright owner(s) are credited and that the original Process intensification of EB66 R cell cultivations leads to high-yield yellow publication in this journal is cited, in accordance with accepted academic practice. fever and Zika virus production. Appl. Microbiol. Biotechnol. 102, 8725–8737. No use, distribution or reproduction is permitted which does not comply with these doi: 10.1007/s00253-018-9275-z terms. Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 16 June 2020 | Volume 8 | Article 617