Regeneration of Capto Core 700 resin through high throughput and laboratory scale studies and impact on production of a SARS‐CoV‐2 vaccine candidate

Abstract During the development of a SARS‐CoV‐2 vaccine candidate, at the height of the COVID‐19 pandemic, raw materials shortages, including chromatography resins, necessitated the determination of a cleaning in place (CIP) strategy for a multimodal core‐shell resin both rapidly and efficiently. Here, the deployment of high throughput (HT) techniques to screen CIP conditions for cleaning Capto Core 700 resin exposed to clarified cell culture harvest (CCCH) of a SARS‐CoV‐2 vaccine candidate produced in Vero adherent cell culture are described. The best performing conditions, comprised of 30% n‐propanol and ≥0.75 N NaOH, were deployed in cycling experiments, completed with miniature chromatography columns, to demonstrate their effectiveness. The success of the CIP strategy was ultimately verified at the laboratory scale. Here, its impact was assessed across the entire purification process which also included an ultrafiltration/diafiltration step. It is shown that the implementation of the CIP strategy enabled the re‐use of the Capto Core 700 resin for up to 10 cycles without any negative impact on the purified product. Hence, the strategic combination of HT and laboratory‐scale experiments can lead rapidly to robust CIP procedures, even for a challenging to clean resin, and thus help to overcome supply shortages.


INTRODUCTION
Cost of goods for bioprocess is typically dominated by downstream processing [1] and hence the implementation of cleaning in place (CIP) techniques for chromatography resins is important for mitigating the costs associated with the use of this unit operation. [2] The development of CIP strategies for biopharmaceuticals, such as monoclonal antibodies, has been reported for affinity, ion exchange, and hydrophobic interaction-ion exchange (e.g., [3][4][5] ) resin modalities. This typically includes the deployment of multiple CIP agents, such as high conductivity and caustic solutions, aiming to remove tightly bound residuals from the resin that would otherwise lead to its fouling and to a potentially significant reduction of the resin's performance in purifying a target product over multiple cycles. The number of cycles can vary from small to large in batch and continuous processes [6] and column re-use in such processes requires the completion of studies validating the lifespan of chromatography media. [7] Apart from cost savings, the re-use of chromatography resins can be a necessity in situations wherein supply limitations are in place.
This was the case during the development of a SARS-CoV-2 vaccine candidate at the height of the pandemic. The processing of live virus vaccines (LVVs) often requires the purification of large targets (>100 nm in diameter) that are more complex than many recombinant subunit protein therapeutics counterparts. [8][9][10][11] LVV size, along with the accrued avidity of interactions with functionalized stationary phases, often leads to low binding capacities and recoveries when purified via bind and elute chromatography. Hence, chromatography steps run in flowthrough mode may be preferred. Capto Core resin technology (Cytiva, Uppsala, Sweden) provides a unique mode of separation for the purification of LVVs. Here, the inactive outer shell acts as a sieve allowing solutes below a molecular weight cut-off to diffuse into and bind to the functionalized inner bead, which displays a triple mode of action due to the octylamine ligand. Consequently, LVVs will flow through, and impurities will be removed from the product pool by adsorbing to the resin. This mode of separation has made the Capto Core resins highly desirable for LVV processing, including SARS-CoV-2 vaccines. [12] This, and its application for the purification of additional vaccine products [13,14] led to uncommonly severe supply shortages during the COVID-19 pandemic.
The CIP of the Capto Core 700 resin is expected to be challenging since the nature of its functionalized inner core can lead to irreversible binding of solutes that may not be easily interrupted. [15] This challenge is further compounded in that the resin is often applied in flowthrough mode, typically as the first step in a purification process following primary recovery. Hence, a higher content of diverse solutes bind to the resin and require removal prior to resin re-use. Here, a methodology for screening cleaning agents and testing cleaning strategies for Capto Core 700 resin, exposed to clarified cell culture harvest (CCCH) expressing a replication-competent chimeric SARS-CoV-2 LVV candidate, is presented. Experiments were completed at microscale and laboratory scales to ultimately determine the feasibility of CIP and re-using a Capto Core 700 column during the production of batches of the LVV candidate. Miniature column chromatography, with the use of RoboColumns (e.g., [16] ), was employed to design a multi-step CIP strategy involving cleaning agents disrupting the binding of solutes to the resin. Leading candidate CIP agents were then deployed in resin re-use experiments, also performed using RoboColumns. Here, the performance of the CIP strategy was assessed across ten cycles by tracking multiple outputs, such as chromatographic traces, product and impurity flowthrough yields, and by determining directly the presence of bound solutes post-CIP in resin extracts using a procedure that combined and expanded on earlier approaches. [17,18] The results from the microscale experiments were verified at the lab scale where resin re-use experiments were also performed by scaling up the CIP strategy. While the same rigorous analysis was applied to the scale-up experiments, for confirming the absence of a negative impact of the CIP strategy on the chromatography step, here a holistic approach was adopted by characterizing the impact of the resin's re-use on the entire purification process generating the final purified product. It is shown that the microscale experiments are scalable and the formulated CIP strategy can be adopted to re-use the Capto Core 700 resin for the production of a SARS-CoV-2 vaccine candidate without any adverse impact on the delivered purified product. The combination of microscale and lab-scale experiments can, therefore, determine the feasibility of CIP for a challenging to clean resin, exposed directly to CCCH for the production of an LVV. This leads to the highly desirable mitigation of costs and supply limitations.

Chimeric VSV∆G-SARS-CoV-2 virus production
Replication-competent, chimeric VSV∆G-SARS-CoV-2 LVV candidate was generated by replacing the live vesicular stomatitis virus (VSV) glycoprotein (G) gene with a coding sequence for the SARS-CoV-2 spike glycoprotein (S). VSV∆G-SARS-CoV-2 was produced in Vero cells. containing the expressed VSV∆G-SARS-CoV-2 vaccine candidate, was either processed further immediately upon its generation or aliquoted and stored at -70 • C until further use.

2.2
High throughput chromatography

Robotic station
High throughput (HT) chromatography studies using PreDictor Robo-Columns (Cytiva, Uppsala Sweden), packed with either 200 or 600 μL of Capto Core 700 resin (Cytiva), were carried out based on the method described in Ref. [16] Here, a Tecan Freedom EVO 150 robotic station was employed, which was controlled by Freedom EVOware v2.8 Measurements at 900 and 990 nm were also made for path length correction purposes. [19] 2.2.3 Resin CIP and high throughput scale resin re-use experiments The robotic system was also deployed to run multi-cycle HT scale purifications of the VSV∆G-SARS-CoV-2 vaccine candidate with resin CIP between each cycle. A total of 10 cycles were performed ( Figure 1A), since they represented an estimate of the minimum number of batches of the VSV∆G-SARS-CoV-2 vaccine candidate produced annually without parallel processing, with eight 600 μL Capto Core 700 RoboColumns (RC1-RC8). Fractions were collected in Axygen 2.2 mL 96-well deep square well plates. Here, at the beginning of each cycle, the columns were flushed with system liquid for 5 CVs to remove the storage solution. This was followed by their equilibration with 10 CVs of 10 mM Tris, pH 7.5, and 150 mM NaCl with a residence time of 2 min.
The columns were then loaded with 70 CVs of CCCH with a residence time of 6 min and fractions were collected every 1.75 mL. Following this, the columns were washed with the equilibration buffer for 1 CV with a 6 min residence time while collecting the effluent in a 0.6 mL fraction. The columns were then cleaned in place in three sequential steps (i.e., CIP1 -CIP3), each applied with a residence time of 6 min while collecting 0.6 mL fractions. CIP1, CIP2, and CIP3 were applied for 3, 5, and 3 CVs, respectively.
At the end of each of cycles 1, 4, 7, and 10, two columns were removed at a time from further cycles ( Figure 1A); one was used for resin extraction, shortly after the completion of the experiment, and the second was sealed and stored at 4 • C. Hence, from RoboColumns 1-8, only RCs 7 and 8 were used across all 10 cycles. For cycles 1, 4, 7, and 10, those columns that were to be removed from further cycling experiments (e.g., RCs 1 and 2 at end of cycle 1), were also flushed with 5 CVs of equilibration buffer, following CIP3, with a residence time of 6 min and while collecting 1.5 mL fractions. Alternatively, those RoboColumns that were also tested in subsequent cycles were stored in 1 N NaOH (i.e., CIP3) until their use. At the end of each cycle, the fractions collected from each RoboColumn during their loading and F I G U R E 1 Design of studies for implementing resin re-use experiments while purifying the VSV∆G-SARS-CoV-2 vaccine candidate from clarified cell culture harvest (CCCH) and cleaning in place (CIP) the used Capto Core 700 columns at (A) High throughput scale using eight RoboColumns (RCs); and (B) Lab-scale using a 20 mL pre-packed column. In (B) the chromatography flowthrough product pool (ChromP) was further processed via ultrafiltration/diafiltration (UF/DF) at cycles 1, 4, 7, and 10 and the column was used to measure the dynamic binding capacity (DBC) of bovine serum albumin (BSA) at the end of cycles 5 and 10 CIP1-CIP3 applications were combined into half-or full-area UV transparent plates, or Matrix 2D barcoded tubes, to create separate pools by mixing equal size aliquots. The pools, fractions, and resin extracts were stored at 4 • C prior to their analysis or at -70 • C for long-term storage.
Chromatograms were generated by aliquoting up to 200 μL of fraction volumes into half-or full-area UV transparent plates and measuring their absorbance at 280, 900, and 990 nm.

Lab-scale resin re-use experiments
Lab-scale resin re-use studies ( Figure 1B Aliquots of these were transferred to Matrix 2D barcoded tubes and stored at -70 • C until analytical testing. In addition to using the column to purify the VSV∆G-SARS-CoV-2 vaccine candidate in ten cycles, the cleaned column was also used to measure the dynamic binding capacity of bovine serum albumin (BSA) at the end of cycles 5 and 10 ( Figure 1B).

BSA dynamic binding capacity measurements
To determine the dynamic binding capacity of BSA at 10% breakthrough (DBC 10% ) the cleaned in place 20 mL column ( Figure 1B)

Resin extraction
Capto Core 700 resin in 200 and 600 μL RoboColumns, which had been previously exposed to CCCH and subsequently cleaned in place, was extracted by removing the resin from the column housing and treating it with a combination of a reducing agent and a detergent. For this purpose, the top cover of the columns was removed, and the housing, containing the resin, was placed upside down inside a pre-weighted Corning Falcon 15 mL conical centrifuge tube (Corning Life Sciences).
For 600 μL columns, the housing was first placed upside down in a weighted 1.5 mL Thermo Scientific Nalgene cryogenic tube (Thermo Fisher Scientific Inc.) before they were both transferred to the 15 mL tube. The 15 mL conical tubes were then centrifuged at 500 g for 5 min using a Sorvall Legend XTR centrifuge (Thermo Fisher Scientific Inc.

Product assays
VSV∆G-SARS-CoV-2 LVV product was tracked by analyzing samples for their content in nucleoprotein (N), specific for VSV, [20] and spike

Data analysis
The data generated from the analytical methods were employed to make qualitative (i.e., SDS-PAGE and TEM images) and quantitative (i.e., chromatograms, product yields, and impurity contents) assessments. For the latter, data analysis is based on inspection of trends, by plotting and processing chromatograms, and one-way analysis of variance, with pairwise comparisons using Tukey's method to control Type I errors, [21] using the resin re-use cycle number as the independent variable. In the absence of an estimate of pure error from technical and analytical replicates, the non-parametric Spearman's rank correlation coefficient, [22] ρ, was employed to determine the presence of a relationship between assay results and resin re-use cycle number. The aforementioned processing of chromatograms involves the aggregation of measurements in total signals for each phase of a chromatography run. For HT scale chromatograms, the measured absorbances of the collected fractions were corrected via subtraction against a corresponding blank (i.e., mobile phase) and then normalized over their pathlength (negative values were replaced by zero). These processed absorbances were then summed over the fractions collected during each phase to yield the total signals (Au/cm) corresponding to the loading, washing, CIP, and post-CIP flushing of a column (i.e., FT, Wash, CIP1, CIP2, CIP3, and PCIP, respectively). For lab scale chromatograms, the total signals (AuÍmL) were estimated by integrating the measured absorbances from each of the aforementioned six phases.
For impurity data, log reduction values were estimated using the base 10 logarithm of the ratio between the starting and final impurity levels.
This accounted for volumetric concentration factors where applicable.

High throughput screening of CIP agents
Capto Core 700 is a multimodal resin with a highly hydrophobic ligand.
Recommendations for CIP this resin are provided in the manufacturer's instructions. However, these require flammable solvents, such as isopropanol and n-propanol. The use of such solvents can become limiting at the pilot plant and commercial scales due to OSHA regulations requiring handling to take place within an explosion-proof facility. [23,24] To overcome this, the screening of the cleaning agents at the HT scale sought to identify alternatives by testing 36 conditions, each designed to deploy up to three cleaning agents per condition in three sequential steps (CIP1, CIP2, and CIP3) (Table S1, Supporting Information). For each test, columns were used to purify CCCH and the chromatographic traces were converted into total signals ( (Table S1, Supporting Information), led to the lowest signal amongst CIP1-CIP3 except for conditions 3, 4, 9, and 10 ( Figure 2A). This suggested that the application of CIP3 could be beneficial in cleaning the resin by removing any additional residuals present after the application of CIP1 and CIP2. Such behavior was not supported, however, by the analysis of the resin extracts from conditions 1-6 ( Figure 2B) and 9-12 ( Figure 2C) demonstrating that only conditions 11 and 12 led to extracts free of residuals ( Figure 2C). For these two conditions, CIP1 included a mixture of 30% n-propanol/1 N NaOH and was found to be the only cleaning agent leading to resin extracts free of residuals when deployed during CIP1 in either a two-step cleaning condition (conditions 15, 16, and 22 in Table S1, Supporting Information, with resin extracts shown in Figure 2C and D, respectively) or when deployed alone (condition 29 in Table S1, Supporting Information, with resin extract shown in Figure 2E).
Conditions including the 30% n-propanol/1 N NaOH mixture in CIP1 led to both high CIP1 and overall signals (Figure 2A). This agreement between the overall absorbance signal and the absence of residuals in the resin extracts could support the employment of chromatograms as a screening tool to identify conditions with a high likelihood of leading to residual-free resin extracts. However, conditions including a 30% isopropanol/1 N NaOH mixture in CIP1 (5,6,19,20,24, and 31 in Table S1, Supporting Information) also led to high overall signals (Figure 2A) while the corresponding resin extracts were neither free of residuals ( Figure 2B, D and E) nor considerably cleaner from extracts obtained from alternative conditions which had a considerably lower overall absorbance signal (e.g., conditions 1, 2, 7 and 8 vs. 5 and 6 in Table S1, Supporting Information, and Figure 2A and B).
Hence, the use of chromatograms to assess cleaning conditions could potentially lead to erroneous conclusions. Here, it is important to highlight that while both high content n-propanol and isopropanol mixtures with 1 N NaOH are recommended in the manufacturer's instructions as cleaning solutions for Capto Core 700, only the former was found to be effective while the latter was similar in performance to, for example, condition 36 including a 1 N NaOH cleaning agent in CIP1 (condition 24 vs. 36 in Table S1, Supporting Information and Figure 2H). Hence, the use of the isopropanol and NaOH solution mixture represented a sub-optimal cleaning agent. The presence of NaOH was determined to be a necessary component of the 30% n-propanol/1 N NaOH mixture since in its absence (condition 28 in Table S1, Supporting Information) the returned resin extract was not clean (condition 29 vs. 28 in Figure 2E). The screening experiments identified therefore a single cleaning agent capable of cleaning Capto Core 700 resin used to purify CCCH. The use of n-propanol and NaOH was characterized further to determine a cleaning strategy for the resin.

Cleaning in place strategy
While screening experiments did not lead to a flammable solvent-free cleaning condition, a potential clean-in-place strategy for Capto Core 700 resin, capable of meeting regulatory requirements, was identified: (1) After its loading and washing, the column would be flushed with 3 CVs, the presence of a hold appeared to lead to marginally cleaner lanes ( Figure 3B) whereas the application of CIP2 for 15 CVs combined with resin storage in CIP3 for 15 days did not lead to any additional removal of foulants from the resin ( Figure 3B). These suggested that given an effective cleaning condition, the use of a hold would not add to its efficiency.
While these results demonstrated that a 30% n-propanol/0.75 N NaOH solution could be employed in cleaning a Capto Core 700 column used for one purification of the VSV∆G-SARS-CoV-2 vaccine candidate, a decision was made to adopt a more aggressive CIP2 solution (i.e., 30% n-propanol/1 N NaOH) and a contact time under flow of 30 min for multi-cycle resin re-use experiments. This aimed to increase the likelihood of a successful CIP strategy in multiple resin re-use cycles with a more conservative approach.

High throughput resin re-use experiments
The formulated CIP strategy was deployed in HT resin re-use experiments where 8 RoboColumns (i.e., RC1-RC8) were employed to purify the VSV∆G-SARS-CoV-2 vaccine candidate across 10 cycles ( Figure 1A). Chromatographic traces ( Figure 4A and Figure   contributes the most to the cleaning of the resin and the application of CIP3 functions predominately as a storage solution (Figure 2A). Trends between the total signals and the cycle number were observed for CIP1 ( Figure 4C) and CIP2 ( Figure 4D). For the former, strong negative correlations were in place for RC7 and RC8 whereas, for the latter, RC5-RC8 displayed strong positive correlations (Table S2,  Chromatography flowthrough product pool yields, based on anti-S quantitative western blotting, ( Figure 5A) were estimated to be on average 73.80% ± 6.02% across cycles 1, 4, 7, and 10 for RC7 and RC8, and the obtained data returned an insignificant effect of cycle number on yields, as determined by one-way ANOVA (p-value = 0.4561, Table   S3, Supporting Information). Here, it is noted that low amounts of the VSV∆G-SARS-CoV-2 vaccine candidate bound to the resin despite its diameter (≈70 nm) being greater than the diameter of the Capto Core 700 resin's pores. [25] Likewise, an insignificant impact of the cycle number on the product yields was also determined based on the infectivity data (p-value = 0.0632, Table S3, Supporting Information) ( Figure 5A) which led to an average yield of 91.63% ± 43.41% across the four considered cycles. Here, the high variability of the average yield was due to the individual yields from cycle 1 which were considerably lower than those from cycles 4, 7, and 10 (i.e., 32.94% ± 0.56% vs. > 85% in Figure 5A). This was attributed to a lack of cryoprotectant addition to the cycle 1 flowthrough product pool aliquots before they were frozen for their analysis with the infectivity assay. This can lead to a loss of infectivity in the stored samples due to their freezing and thawing prior to their testing. Excluding the cycle 1 chromatography product yield data led to an even lower F statistic and hence did not change the derived conclusions (Table S3, Supporting Information).
Higher-resolution chromatography flowthrough product pool yield results were obtained based on the anti-N quantitative western blotting data since this assay was deployed for each RoboColumn across the 10 cycles ( Figure S3 and Table S4, Supporting Information). These data indicated a significant difference in the product yields for RC8 (pvalue = 0.0066, Table S4, Supporting Information); a difference was observed between cycle 3 and cycles 4 and 10 (i.e., 61.32% ± 8.45% vs. 90.74% ± 10.51% and 96.67% ± 16.10%, respectively) ( Table S5, Supporting Information). Conversely, for RC7, which also went through all 10 cycles of resin re-use ( Figure 1A), no significant differences were detected (Table S4, Supporting Information). This, along with the fact that for RC8 only 2 out of 45 pairwise comparisons (Table S5, Supporting Information) were shown to be significantly different from each other, and none of the pairwise comparisons between cycle 1 and later cycles were statistically significant, led to the conclusion that these results corroborated the infectivity and anti-S quantitative western blotting yield data. Hence, these data support that Capto Core 700 chromatography flowthrough product pool yields are not dependent on the number of resin re-uses.
The absence of persistent foulants, accumulating from one re-use cycle to the next, was also indicated by the analysis of the product pools ( Figure 5B and C) and resin extracts ( Figure 5D) for impurity presence. The chromatography flowthrough product pools across all cycles displayed an identical band pattern and purity based on SDS-PAGE ( Figure 5B) and the ELISA HCP assay, for cycles 1, 4, 7, and 10, showed that the chromatography step reduced considerably and consistently the HCP content in CCCH to < LOQ in the product pools ( Figure 5C). The deployment of the higher throughput HCP quantitative western blotting assay agreed with these results since the product Similar to the HT scale results ( Figure 4A), the recorded chromatographic traces from the lab-scale runs ( Figure 6A) showed a near-perfect overlap across the cycles. This was especially true during the loading ( Figure 6B) and washing ( Figure 6B) of the column wherein the integrated chromatograms from these two phases led to signals that were independent of the column re-use cycle. For the former, the flowthrough peak showed an increase from cycles 1-5 to cycles 6-10, which was < 1% and hence insignificant. A step increase between these two sets of cycles was also observed in the recorded signal for CIP3 whereas within each set the CIP3 signal remained virtually constant ( Figure 6B). For cycles 6-10, the increase in the CIP3 signal is observed in the recorded chromatograms ( Figure 6A); it increased ≈1 CV after the completion of CIP2 and remained constant until the completion of CIP3. Hence, such an increase represents a baseline shift from CIP2 to CIP3 instead of indicating the elution of solutes. This behavior was attributed to an error in the preparation of the CIP3 solution as no other sources of error could be identified.
Conversely, for CIP1 and CIP2 ( Figure 6B) a strong positive correlation was observed between the integrated signals and the number of re-use cycles (ρ = 0.95, p-value < 0.0001 and ρ = 0.99, pvalue < 0.0001, respectively). Contrary to the HT scale data, here the CIP2 step did not lead to the highest observed signal amongst CIP1-CIP3 ( Figure 6B). This was attributed to the fact that at the HT scale the recorded signals were blank corrected whereas this was not the case for the lab-scale data. Nevertheless, the existence of strong positive correlations for CIP1 and CIP2 for the lab-scale data could indicate an incomplete cleaning of the resin, also considered for the HT scale resin re-use study ( Figure 4D). This was sought to be verified by the analysis of chromatography flow through product pools, resin extracts, and intermittent BSA DBC measurements ( Figure 1B).
Product yields, based on anti-S quantitative western blotting and infectivity data ( Figure 7A), were determined to be on average 69.60% ± 4.83% and 102.40% ± 19.91%, respectively, which were close to those obtained from the HT resin re-use experiments (i.e., 73.80% ± 6.02% and 91.63% ± 43.41%, respectively). The higher variability observed for the infectivity data-based yields is attributed to the nature of the assay itself and not to the variability of the chromatography step. The product yields per cycle from both assays were also found to be independent of the cycle number in both cases (ρ = -0.12, p-value = 0.7588 and ρ = 0.13, p-value = 0.7329, respectively). At the same time, the ELISA HCP and hcDNA results from the chromatography product for each cycle were < LOQ, a result also corroborated for the former by the SDS-PAGE analysis of the flowthrough product pools ( Figure 7B).
While these results suggested the success of the CIP strategy, additional rigor was applied in evaluating its effectiveness. For this purpose, BSA DBC 10% measurements were made between cycles 5 and 6 (DBC 10% = 10.55 g L -1 ) and after cycle 10 (DBC 10% = 10.43 g L -1 ) ( Figure 1B). The estimated DBC 10% values were close to each other and to the reference, DBC 10% (11.34 ± 0.10 g L -1 ), determined using three fresh 20 mL columns. In comparison, when a fourth fresh 20 mL column was used to purify 245 CVs of CCCH and to then determine BSA's DBC 10% , without being cleaned in place, the returned DBC 10% was estimated to be 4.66 g L -1 . This significant reduction demonstrates the extent of the impact of an ineffective CIP of a CCCH exposed Capto Core 700 column. Hence, while the BSA DBC 10% values obtained during the re-use experiments were marginally lower than the reference resin was confirmed to be comprised of a fibrous mesh with large pores ( Figure S4A, Supporting Information), as observed previously. [25] The images of the three resin samples from the re-used 20 mL Capto Core 700 column (Figures S4B-D, Supporting Information) were identical to each other and to the image from the fresh resin sample ( Figure   S4A). Hence, the structure of the resin itself was not affected by the re-use of the column. Moreover, previously published TEM images of used but not cleaned Capto Core 700 resin typically depict foulants as large dark globules. [25] Based on this, the absence of such globules in the TEM images in Figures S4B-D Figure 7E). A third control experiment at the RoboColumn scale was conducted to simulate the BSA DBC measurements but here the column was cleaned before the resin was extracted (i.e., BSA extract in Figure 7E). These analyses confirmed the significant pres-  Table   S6, Supporting Information); a higher log reduction was achieved in cycle 7 compared to cycles 1, 4, and 10 ( Figure 8B and Table S7, Supporting Information). The analysis of variance results, however, was driven by a considerably low pure error since the returned HCP concentrations from the four cycles had a variability of ≈1.5%-3%.
Hence, the increase in log reduction for cycle 7, compared to cycles 1, 4, and 10, corresponded to a difference of < ≈4.5% and combined with the fact that for cycle 7 the concentration factor overshot the target by ≈60% led to attribute this difference to acceptable process variability instead of the re-use of the column. This was also indicated by observing the similarity of the hcDNA log reduction trends ( Figure 8B) to those from the HCP data since cycle 7 was also shown to lead to the highest log reduction compared to the rest of the four cycles. Here, the achieved hcDNA log reduction across the four cycles (1.84 ± 0.07) was lower than the one for HCP.
The SDS-PAGE analysis of the UF/DF intermediates and final product also indicated similarity of product yields for UFP for cycles 1, 4, 7, and 10 ( Figure 8A). This was supported by anti-S quantitative western blotting and infectivity data returning average UFP step yields from the chromatography flowthrough product pool, across the four cycles, of 74.16% ± 4.13% and 47.70% ± 6.75%, respectively ( Figure 8B). (B) In left-hand side y-axis, log reduction of host cell protein (HCP) (○) and DNA (hcDNA) (•) impurities at UFP from CCCH obtained by ELISA and qPCR analysis, respectively. On the right-hand side y-axis, step yield at UFP from ChromP based on anti-Spike (S) protein quantitative western blotting (□) and infectivity data (■). In (B), the HCP log reduction data error bars correspond to ± 1 standard deviation (sd) from analytical replicates strategy, enabled the robust production of the VSV∆G-SARS-CoV-2 vaccine candidate for 10 batches.

3.4
Assessment of the high throughput approach to establishing a cleaning-in-place strategy  Table S1, Supporting Information). This demonstrated a lack of a synergistic action between cleaning agents. For example, the use of 0.5 M acetic acid alone led to a high amount of residuals bound to the column postcleaning (condition 32 in Figure 2F). Conversely, condition 1, which employed both 0.5 M acetic acid and 1 N NaOH led to a cleaner resin extract ( Figure 2B), which was, however, nearly identical to cleaning condition 36 which employed 1 N NaOH alone ( Figure 2H). The lack of such synergistic action for the evaluated cleaning agents, the inability of high concentration acids and chaotropes (condition 35, 34 in Table   S1 and Figures 2G, 2F, respectively) to produce clean resin extracts, and time constraints led to the decision to focus on evaluating the deployment of the 30% n-propanol/1 N NaOH solution in a cleaning strategy instead of extending screens of cleaning agents.
The screening experiments also demonstrated the limitations of evaluating cleaning approaches solely based on chromatographic absorbance traces; conditions leading to high absorbance signals were not always those returning the cleanest resin extracts (e.g., conditions 19 and 20 in Figure 2A and D). Such false positives were avoided here by also assessing the performance of cleaning agents through resin extraction and subsequent SDS-PAGE analysis. This leads to a rigorous assessment since the absence of residuals in resin extracts consists of direct observation, but SDS-PAGE is characterized by low throughput. However, during the screening experiments, the generated CIP results indicated rapidly the requirement of a solvent/caustic mixture.
Hence, while the efficiency of these experiments could be improved, by deploying higher throughput and quantitative analytics, no bottleneck was experienced due to the small number of samples for evaluation. Conversely, when quantitative results were desired, along with higher throughput, the HCP quantitative western blotting assay was deployed. This was implemented when screening the concentrations of the NaOH and n-propanol components in the cleaning agent and its contact time, with or without a hold, with the resin (Figure 3). The same applied when testing resin extracts from the resin re-use experiments to determine quantitatively a significant presence of residuals as a function of cycle number. While this approach resembled the one proposed by, [17] the assay used here is characterized by high sensitivity and the adopted approach allows for multiple analytics to be run, with increased levels of replication, due to extracting higher resin volumes.
The success of the screening experiments in identifying a cleaning agent was further supported by the HT scale resin re-use experiments where even after 10 purification cycles the resin remained clean and continued to deliver a chromatography product unaffected by its reuse. While a shorter and more efficient version of the CIP strategy could be applied (Figure 3), a more conservative approach was adopted here employing the 30% n-propanol/1 N NaOH mixture in CIP2 with a 30 min contact time under flow. This was based on the observation that this condition was successful in cleaning columns packed with Capto Core 700 resin after they were deployed to purify the VSV∆G-SARS-CoV-2 vaccine candidate and additional LVVs produced in adherent Vero cells, when none of the other tested conditions, including the isopropanol/NaOH mixture, were successful. Furthermore, such a conservative approach assisted the success of the strategy in the resin re-use experiments and enabled the generation of results supporting the production of 10 batches of the VSV∆G-SARS-CoV-2 vaccine candidate.
During the HT scale resin re-use experiments, the cleaning of the columns was assessed by product and impurity flowthrough yields and analysis of resin extracts ( Figure 5). Typically, resin re-use experiments also employ the measurement of dynamic binding capacities, at various points between cycles, to determine the presence of persistent column fouling. This was not implemented here since RoboColumns can return variable DBCs compared to their lab-scale counterparts. [26] Moreover, the main focus of the HT investigation was to achieve a high number of re-uses, which can be challenging with RoboColumns as they are a disposable technology; the seal at the upper frit of these columns can lose its integrity based on the type of used sample, the number of re-uses, and also due to batch-to-batch variability of the columns themselves. Typically, five cycles can be completed without observing any beading of liquid at the top of these columns. Including intermittent DBC measurements in the resin re-use experiments would therefore reduce significantly the number of cycles achieved with no beading observation. Here, re-using the columns ten times, while applying caustics, led to beading towards the end of the cycling experiments. This and the lack of the DBC estimates at the HT scale were two driving forces for performing lab-scale resin re-use experiments in addition to the generation of material for UF/DF studies post-Capto Core 700 chromatography and the desired verification of the HT scale results. Comparing the results from lab-scale runs to those of the RoboColumns would assist to determine whether beading could lead to spurious conclusions or whether DBC measurements were required to correctly assess the performance of the devised CIP strategy. Furthermore, this comparison was critical in determining the impact of scaling RoboColumn-based separations on a constant residence time basis and their packing quality on the generated data.
Based on the observed agreement between scales, these limitations of the RoboColumn technology do not prevent them from posing as an excellent scale-down model for designing and testing cleaning-inplace approaches for challenging to clean chromatography resins such as Capto Core 700.

3.5
Evaluation of the Capto Core 700 resin CIP strategy for the production of the VSV∆G-SARS-CoV-2 vaccine candidate with multiple resin re-use cycles Scaling up the CIP strategy, derived from the HT scale experiments, to lab-scale using a 20 mL pre-packed column, led to the same conclusions as those from the HT scale resin re-use experiments; the application of the CIP strategy delivered a chromatography product that remained unchanged across the 10 re-use cycles (Figure 7) and the two measured intermittent BSA DBC 10% values were similar to the ones obtained from columns that were not employed to purify the VSV∆G-SARS-CoV-2 vaccine candidate. The resin structure was also found to be unaffected by its exposure to the cleaning agents across the 10 cycles ( Figure S4, Supporting Information). The effectiveness of the CIP strategy was also demonstrated by assessing the impact of using a CCCH exposed, but not cleaned column, on BSA DBC 10% measurements where it was shown that in absence of column cleaning the measured dynamic binding capacity undergoes a significant reduction (≈60%). Based on these results, the conclusion can be made that the lab-scale data corroborate the findings from the HT scale investigations.
Focusing on the chromatography step alone assesses the impact of the CIP strategy on an intermediate process product instead of determining its impact on the final VSV∆G-SARS-CoV-2 vaccine candidate product. The lack of breakthrough of HCPs (based on ELISA analysis) and hcDNA, with an increasing number of column re-uses, (all results were below LOQ) contributed to this conclusion ( Figure 7). However, the purification process includes a UF/DF step wherein the chromatography product is concentrated 50-fold and this leads to impurity contents greater than LOQ at the final product during typical processing without any resin re-use. This concentration step could therefore act cumulatively, across the resin re-use cycles, and compound an elevated level of impurities in the final product. This could occur in a limiting situation wherein there is a weak breakthrough of impurities, which while it increases with the number of resin re-use cycles is not significant enough to be observed in the chromatography product without the additional concentration offered in the UF/DF step.
Hence, processing the chromatography product through the UF/DF step, which includes a concentration step adds rigor to the evaluation of the CIP strategy. This showed that re-using the resin for 10 cycles had no impact on the final product in terms of both purity and yield ( Figure 8) even in the presence of a 50-fold concentration across the UF/DF step. This finding also indicated that more than 10 batches of the VSV∆G-SARS-CoV-2 vaccine candidate could potentially be produced while CIP the Capto Core 700 column with the identified CIP strategy. While the data supporting more than 10 re-uses would need to be generated, this prospect exemplifies the importance of identifying optimally performing cleaning conditions for chromatography columns since sub-optimal conditions would not allow for the maximization of column re-use cycles. The latter would be incremental in maintaining a low cost of goods with an increase in the annual production of the VSV∆G-SARS-CoV-2 vaccine candidate.
The demonstrated scalability of the HT scale resin re-use results is significant with regards to the applicability of microscale HT chromatography techniques for the development of CIP strategies. As demonstrated here, such CIP development efforts are complex and require considerable resources in terms of both time and quantities of chromatography resin and feed material. These can make their completion at a laboratory scale prohibitive, especially at the early stages of development where feed material availability can be limited. The deployment of HT chromatography techniques can mitigate these challenges since they require small amounts of materials, and their parallelization and automation enable the rapid and systematic screening of large experimental spaces to identify optimal resin cleaning conditions. While the above considers predominantly the screening of cleaning agents, demonstrating here that HT scale chromatography columns can be employed during resin re-use experiments, and can return scalable results, support their deployment for such studies; it can lead to considerable savings in terms of used feed materials (e.g., ≈2 L at HT scale vs. ≈13 L at laboratory scale based on the approach described in Figure 1) and by offering an alternative scale for performing such long campaigns of experiments it alleviates experimental bottlenecks for laboratory-scale development teams. These render the adoption of HT chromatography technologies an attractive alternative to laboratory-scale feasibility studies.

CONCLUSIONS
High throughput chromatography techniques, based on RoboColumns, were employed to devise a cleaning strategy for Capto Core 700 to alleviate supply limitations met during the development of the VSV∆G-SARS-CoV-2 vaccine candidate. Screens of cleaning agents revealed the need for both caustic (NaOH) and solvent (n-propanol) at high concentrations (≥0.75 N and 30%, respectively) to yield resin extracts free of residuals. Conversely, acids, bases, and chaotropes, either deployed alone or in a combination, were not capable of cleaning the resin nor was the manufacturer recommended isopropanol and NaOH CIP solution mixture. The use of n-propanol in the successful cleaning agent would require the CIP of a pilot and commercial-scale columns to occur in an explosion-proof facility. Consequently, a CIP strategy was devised employing the flushing of the column with NaOH before and after its cleaning with the solvent-containing solution in an inactivation and storage step, respectively. This strategy was tested at a high throughput scale and was demonstrated to be effective for up to 10 resin re-use cycles. These results were also verified at the lab scale, using a 20 mL pre-packed column. Here, the application of the CIP strategy was found to be effective for the Capto Core 700-based purification step and did not impact the subsequent UF/DF purification step. The final purified VSV∆G-SARS-CoV-2 vaccine candidate product, delivered while cleaning and re-using the column, was unaffected for up to 10 column re-use cycles. The results generated in this study serve to support the application of high throughput chromatography techniques for screening, implementing, and evaluating cleaning strategies for chromatography resins. This led to demonstrating the potential elimination of a high-risk factor, such as supply shortage, for the development of the VSV∆G-SARS-CoV-2 vaccine candidate. The capability to implement these microscale studies rapidly and efficiently, along with their scalability, is therefore a valuable tool in enhancing the responsiveness of purification development teams in the face of unprecedented challenges.