Tangential flow filtration of nucleic acids and LNPs in mRNA manufacturing workflow

The plasmid DNA (~6 kb) that encodes for an mRNA GFP construct (~1000 nt) was linearized using EcoRI restriction enzyme digestion at 37°C for 60 min. A TFF step (UF/DF0) was subsequently performed to remove the enzyme and to buffer exchange the DNA into HyClone™ HyPure RNAse-free water. An initial feasibility process development (PD) scale run of the linearized DNA TFF was performed at small-scale using a 100 kDa Xampler™ hollow fiber cartridge (UFP-100-C-3MA,140 cm²) with 30 cm length (data not shown). A total of 120 mg DNA was used at 8.5 g DNA/m²  or 24 L/m² loading. Operation at 3667 s^-1 shear rate and 0.7 bar TMP to achieve ~ 4× volumetric concentration factor (VCF) and five diafiltration volumes (DV) was completed in slightly over 2 h (128 min) at an average permeate flux of 18 LMH. Recovery of 75% was obtained and the protein content was reduced by 10-fold while maintaining integrity of DNA (Table 3). This small-scale run provided feasibility data that the linearized DNA was retained with a 100 kDa hollow fiber, and quality was maintained with acceptable recovery and impurity removal after UF/DF.

Table 3. Quality and yield data for UF/DF0 of linearized DNA 

For the production scale run, TFF was performed on an ÄKTA readyflux™ system modified with a 1/4” Pump Modification Kit and using a SU 1/4” flow kit. Sanitization was performed by recirculating 0.1 N NaOH for 30 min followed by neutralization and equilibration with water for WFI. To be amenable to manufacturing scale, 100 kDa Xampler™ hollow fiber cartridge (UFP-100-C-4X2MA) of 0.14 m² EFA with 60 cm length was used to process 0.93 g of plasmid DNA after restriction enzyme digestion. Production scale TFF was performed at a higher shear rate of 5700 s^-1, lower TMP of 0.55 bar (8 psi / 0.055 MPa), and lower loading of 6.6 g/m² or 6.6 L/m² compared to the small-scale run, with the aim to complete the process in less than 2 h as well as to obtain improved removal of enzyme. Diafiltration of five DVs was performed first followed by 1.6× VCF of linear DNA in HyPure water. Feed concentration was already at ~ 1 g/L as required for target concentration and hence, only a final concentration step of 1.6× was performed for recovery buffer flush. Compared to the 10-fold reduction in the small-scale run, an approximately 20-fold reduction in protein content was observed in the large-scale run (Table 3) with levels at 0.3% in the purified linear DNA, which is well below the 1% to 10% as per pDNA USP guidelines (2). The run was completed in < 2 h (102 min) with an average permeate flux of 26 LMH and a recovery of ~ 90% with integrity maintained after UF/DF0 (Table 3).

Fig 2. Permeate flux and TMP over time for linearized pDNA product exchanged into HyPure water.

Figure 2 shows the permeate flux throughout the UF/DF0 operation under a constant feed pressure of 1.58 bar (22.9 psi / 0.158 MPa). After initial spike to 40 LMH the process permeate flux equilibrated to approximately 17 LMH, which increased over the course of the diafiltration to 36 LMH therefore averaging to 26.1 LMH overall. During concentration the flux decreased slightly to 33 LMH. Conductivity of the permeate decreased throughout the diafiltration from 11.1 to 0.84 mS/cm, indicating an almost complete removal of solutes and sufficient exchange into water.

Fig 3. Permeate flux and 280 nm absorbance over diafiltration volumes for linearized pDNA product exchanged into HyPure water.

Figure 3 shows the relationship between permeate flux and absorbance vs DVs. The average permeate flux during the diafiltration step was 27 LMH and it stabilized as it approached four DV. Permeate absorbance at 280 nm decreased from 974 mAU to 896 mAU by four DV where it also stabilized, indicating the limit of enzyme removal under the conditions employed here. Since 3.2 µg residual protein content per mg of DNA still exists in the purified DNA (Table 3), process parameters could be optimized further for complete removal of enzyme/protein, if desired. RNase activity was not detected in purified DNA. After the UF/DF0 step, ~ 1 L of the linearized DNA was filtered with Supor™ EAV Mini Kleenpak capsule for bioburden reduction, with an EFA of 260 cm² (KA02EAVP2S). Gel electrophoresis data of the purified linearized DNA showed an expected single band at 6000 bp, demonstrating DNA integrity was not affected during TFF (Fig 4). Aliquots of the filtered material were stored at 4°C and used the next day for IVT reactions.

Fig 4. 1% agarose gel electrophoresis of the linearized DNA after UF/DF0 and sterile filtration. Sample bands (~6000 bp) indicate the triplicate runs at 50 ng and 100 ng.

Using the average permeate flux of the production-scale UF/DF0 that generated purified linearized DNA, calculated parameters and loading/EFA for the UF/DF0 process of hypothetical scale-down pilot and scale-up batches are provided in Table 4. SU hollow fibers and ÄKTA™ TFF systems can be scaled for the required EFA for different feed flow rate (liter per min, LPM) with SU flow kits (1/8’’ to 3/8’’) for different scales. However, depending on the process, we would like to note that plasmid DNA linearization and TFF purification can also be a part of the pDNA manufacturing workflow targeted towards mRNA manufacturing, and in such cases scaling factors and the corresponding process volumes (or mass) may vary.

Table 4. Scalable parameters for UF/DF0 of linearized DNA

Linearized DNA after UF/DF0 and bioburden filtration was used to synthesize mRNA (GFP construct of ~ 1000 nt, uncapped, encoded polyA tail) by performing in vitro transcription (IVT) reactions using our ReadyToProcess WAVE™ 25 system. T-series cassettes with 100 kDa Delta regenerated cellulose membrane were used for the UF/DF1 and UF/DF2 steps of the mRNA manufacturing workflow. Centramate™ LV and Centrasette™ cassette holders were used for small-scale feasibility and large-scale manufacturing runs, respectively. The entire flow path of the TFF system was sanitized by recirculating 0.25 M NaOH for 30 to 60 min followed by neutralization with WFI and equilibration in appropriate buffer of the process step. Freeze-thawed and pre-filtered mRNA was used as feed for both UF/DF1 and UF/DF2 steps. This was to mitigate filter clogging due to possible precipitation resulting from crude IVT mixtures and the freeze-thaw cycle. This was a pre-emptive measure as no visible precipitation was observed in the work described here.

In a typical mRNA manufacturing workflow, UF/DF1 is generally employed to remove excess nucleotides, enzymes, and by-products from the post-IVT crude mRNA mixture to the extent possible, and condition the product into a suitable buffer for the subsequent process step(s).

Two small-scale experiments were carried out to de-risk operation at production scale (data not shown). Optimal TFF parameters of 5 LMM feed flux (liter per min per m² EFA) and 15 psi TMP (1.03 bar / 0.1 MPa) were chosen based on previously published data (3), but we further tested the limits of loading ≥ 100 g/m² on performance of T-series 100 kDa cassettes with Delta membrane for mRNA purification here. Small-scale IVT reactions were run (0.1 to 0.5 L) in  1 L Cellbag™ bioreactor containers using the ReadyToProcess WAVE™ 25 system, and crude IVT mRNA mixtures were pooled after freeze thaw and pre-filtered before TFF. The TFF flow path containing the Centramate™ cassette of 0.0093 m² EFA was sanitized followed by flushing with WFI and equilibration in HyClone™ 10 mM Tris and 1 mM EDTA, pH 7.5 (TE buffer). A diafiltration of seven DV in the first run was performed with a loading of 14 L/m² (140 g/m²) that had a run time of 63 min with an average permeate flux of 93 LMH. In a subsequent run, two Centramate™ cassettes were stacked (0.0186 m² EFA) and a 3-fold concentration followed by a five DV diafiltration was performed by loading 86 L/m² (102 g/m²), which was completed in 123 min with an average permeate flux of 98 LMH. mRNA purification (1 and 2 kb RNA constructs) by UF/DF with 100 kDa hollow fiber filters was found to have an average permeate flux of 35 LMH at a membrane load capacity of 10 g/m² with flux decays of up to 70% at 45 g/m² (4).  The data presented here suggest that crude IVT mRNA purification could be completed using Delta 100 kDa membrane cassettes with an average flux of 95 LMH and > 90% recovery within ≤ 2 h even at a membrane load capacity ≥ 100 g/m². These two small-scale runs provided feasibility data to perform UF/DF1 for a manufacturing-scale process.

For the production scale run, 1 L and 2 L IVT reactions were performed in 2 L and 10 L Cellbag™ bioreactor containers. Crude IVT mRNA mixtures were pooled after a controlled freeze-thaw cycle by RoSS.pFTU large-scale system with RoSS freezer shells (Single Use Support) using 5L Allegro™ biocontainers. The pooled crude IVT mRNA mixture was pre-filtered using a 0.45 µm Fluorodyne^TM II DBL filter for use as the feed material for UF/DF1 operation on the ÄKTA readyflux™ system equipped with a SU 3/8” flow kit. Sanitization was performed followed by equilibration with TE buffer. 12.7 L of crude IVT mRNA mixture was loaded onto a 0.5 m² T-series Centrasette™ with 100 kDa Delta membrane at a loading of 25 L/m² or 38 g/m² and purified by buffer exchange into TE buffer for six DV followed by ~ 3× VCF from 1.4 g/L to 4 g/L at 5 LMM feed flux and 1.03 bar (15 psi / 0.103 MPa) TMP. The run was completed in 65 min with an average permeate flux of 110 LMH.

Figure 5 shows the permeate flux throughout the UF/DF operation. Initial flux was 60 LMH and this increased over the course of the diafiltration until stabilizing at 123 LMH which averaged out to 112 LMH. During concentration the flux decreased to 110 LMH for an average of 114 LMH. Permeate conductivity stabilized around 50 min, indicating a successfully completed DF.

Fig 5.  Permeate flux, conductivity and TMP over time for mRNA crude IVT product that was diafiltered into 10 mM Tris, 1 mM EDTA, pH 7.5 buffer and concentrated to ~4 g/L.

Fig 6. Permeate flux and 280 nm absorbance over six diafiltration volumes for post-IVT mRNA exchanged into 10mM Tris, 1mM EDTA, pH 7.5 buffer.

Figure 6 shows the relationship of the permeate flux as a function of the six DVs. The average permeate flux during the diafiltration step was 118.5 LMH and it took 3.5 DV to stabilize. Absorbance signal at 280 nm in the permeate began above the limit of detection and decreased over 3.5-fold throughout the six DV indicating a significant removal of impurities. However, the permeate absorbance did not stabilize indicating additional DVs could be beneficial to further purify the mRNA, if desired. Product recovery of 98% was achieved after UF/DF1 with a total run time of 65 min. Capillary electrophoresis data showed no detectable impact to mRNA integrity during TFF. Residual protein was reduced by 2-fold and residual DNA by 7.4-fold after UF/DF1 with the 100 kDa Delta membrane (Table 5).

Table 5. Analytical data for mRNA purification by UF/DF1 and UF/DF2.

^*BLQ- lower limit of quantitation for NanoOrange, 300 ng/mL as per the manufacturer’s instructions.

In an RNA manufacturing workflow, a TFF step (UF/DF2) may be performed following chromatography purification to concentrate and buffer exchange into a storage buffer (1 mM sodium citrate, pH 6.5 was used in this work). We have not performed chromatography in the demonstrated use case of an mRNA workflow process here, but mRNA was diluted to ~ 0.5 g/L in TE buffer prior to performing UF/DF2, to simulate typical mRNA elution concentrations after chromatography. We chose optimal TFF parameters of 5 LMM feed flux and 1.03 bar (15 psi / 0.103 MPa) TMP based on previous data (3) and further optimized the process to determine the performance of Delta 100 kDa membranes for UF/DF2. We performed a small-scale feasibility run (data not shown) for the UF/DF2 step using a Centramate™ cassette with EFA of 93 cm² on an ÄKTA flux™ s, using our flow kit, at a loading of 20 L/m² or 9 g/m². Diluted 0.5 g/L mRNA in TE buffer was freeze-thawed and UF/DF2 operation was performed to concentrate to 1.9 g/L (~ 4× VCF) followed by five DVs of buffer exchange into 1 mM sodium citrate, pH 6.5, which was completed in 11 min. mRNA integrity was 100% before and after TFF, and the product recovery was 97% (Table 5).

Based on these small-scale data, at five LMM feed flux and 1.03 bar (15 psi / 0.103 MPa) TMP and a time average permeate flux of 153 LMH, processing the scale-up batch of 13.7 L for 4× VCF and five DV using a 0.5 m² EFA at a loading of 27 L/m² would be expected to complete in ~ 22 min run time. Shorter processing time is desired for UF/DF2 step to complete the pre-filtration and post-UF/DF2 bioburden control filtration unit operations within the 8 h manufacturing shift. For the production-scale run, after a controlled freeze-thaw cycle, 7.4 g (13.7 L) of the diluted ~ 0.5 g/L feed material in TE buffer was pre-filtered for bioburden control and/or precipitate removal using Supor™ EKV membrane in a Kleenpak™ capsule filter with a 375 cm² EFA in 52 min. Similar to UF/DF1, UF/DF2 was performed on the ÄKTA readyflux™ system equipped with a SU 3/8” flow kit, and sanitization was performed followed by equilibration in 1 mM sodium citrate, pH 6.5. A 0.5 m² T-series cassette with Delta 100 kDa membrane was loaded at 27 L/m² or 15 g/m² to concentrate from ~ 0.5 g/L to ~ 2 g/L (~ 4× VCF) followed by buffer exchange (4.5 DV) from TE buffer into sodium citrate at 5 LMM feed flux and ~ 0.7 bar TMP and the run was completed in ~ 20 min.

Figure 7 shows permeate flux, retentate conductivity and TMP across 4.5 DVs. There was no effect on integrity during TFF (100% main peak area before and after TFF) and 96% of product recovery (Table 5). Interestingly, the amount of residual protein in mRNA was found to be below the quantitation limit of the assay (NanoOrange) before UF/DF2, possibly due to dilution, but was expected to be ~ 3.4 µg protein/mg RNA post-UF/DF1 as shown in Table 5. However, after UF/DF2, residual protein content was undetectable within the assay limits but was calculated to be < 0.2 µg protein/mg mRNA, which meets the product quality attribute target (< 0.3 to 0.5 µg/mg RNA) (1). Residual DNA content remained the same before and after UF/DF2 (Table 5) at a level of 8 ng/mg of RNA, which is much below the required CQA target (< 330 ng/mg RNA) (1). Other purification steps may be employed prior to UF/DF2 to reduce the levels of residual protein and DNA even further, if required. The double-stranded RNA (dsRNA) by-product levels were < 1% under the IVT conditions employed here but additional purification steps may be utilized prior to UF/DF2 to bring the dsRNA levels to < 0.1% (1 ng/µg mRNA) (1), if needed. After the UF/DF2, 7.1 g (3.7 L) of the mRNA was filtered using a Supor™ Prime sterilizing grade filter in a Kleenpak™ Spectrum capsule with a 240 cm² EFA in less than 25 min for bioburden control and frozen at -80°C. Pre-filtration, UF/DF2, and bioburden filtration steps to make the bulk mRNA DS were completed in < 2 h of run time (not including the set up time for the unit operations).

Fig 7. Permeate flux and conductivity over 4.5 diafiltration volumes for mRNA exchanged from 10 mM Tris, 1 mM EDTA, pH 7.5 into 1 mM sodium citrate, pH 6.5 for UF/DF2.

Table 6. Calculated parameters for UF/DF of mRNA purification needed for IVT reactions of different scales

mRNA purification using T-series cassettes with Delta 100 kDa membrane can achieve target concentrations of 2 to 4 g/L as described in the work here. mRNA purification with 100 kDa hollow fibers have been shown to limit target concentrations to 1 to 1.5 g/L (4). Table 6 displays the calculated parameters and conditions for mRNA purification by UF/DF for hypothetical  scale-down, pilot or scale-up batch sizes based on the process data generated for the 3 L IVT production scale here. T-series cassettes with Delta or Omega™ membrane and Cadence™ single-use TFF modules with Omega membrane can be scaled for the required EFA (0.0093 to 2.5 m²) as well as various ÄKTA TFF systems for different feed flow rates (LPM) with SU flow kits (1/8’’ to 1/2’’) for purification of RNA at different scales. In this work, for the production scale run we performed diafiltration first followed by concentration, but the reverse operation of concentration followed by diafiltration for process efficiency can also be performed if appropriate. Table 6 is based on the loading of 38 g/m² of the production scale run of DF followed by UF process but, the PD scale run of UF first and then DF process showed ≤ 2 h process time even at a membrane load capacity ≥ 100 g/m². Hence, it is recommended to test the filter capacity and optimal conditions for different mRNA UF/DF processes.

In a typical mRNA manufacturing workflow, after the encapsulation of purified mRNA into LNPs, the aqueous: ethanol mixture of the mRNA-LNPs is purified by TFF to remove ethanol, exchange into a formulation buffer, and concentrate to the dosing target. A feasibility run for encapsulation of mRNA into LNPs was performed on NanoAssemblr™ Blaze using a NxGen™ 500D cartridge by mixing a proprietary lipid cocktail (in ethanol) and mRNA (diluted in 0.1 M sodium acetate, pH 4.0) with a 2:1 in-line dilution in 1× PBS pH 7.4 (~ 8.3% ethanol in final formulation). Based on the optimal TFF parameters from previous data with DC100T01 at 0.2 g/m² of mRNA in LNPs (3), we chose a feed flux of 6 LMM and a TMP of 1.03 bar (15 psi / 0.103 MPa). UF/DF3 process was further developed at loading of >1 g/m² of mRNA in LNPs and to achieve different VCF and scaling-up of T-series cassettes with Delta 100 kDa membrane.

Small-scale feasibility runs (data not shown) were performed using Centramate™ cassettes with EFA of 93 cm² (DC100T01) and 186 cm² (DC100T02). For the DC100T01 run on an ÄKTA flux™ s using our flow kit, 1.5 g/m² (44 L/m²) of mRNA-LNPs were loaded and a 7× VCF was achieved from 0.03 g/L to a final concentration of 0.24 g/L followed by five DV buffer exchange into a proprietary pH 7.4 buffer. This took 40 min to complete with an average permeate flux of 70 LMH. For the DC100T02 run at a loading of 44 L/m² or 1.6 g/m², an 18× VCF to a final concentration of ~ 0.55 g/L mRNA in the LNPs was achieved followed by five DV. The UF/DF operation was completed in 55 min with an average permeate flux of 75 LMH. Both the runs had more than 90% recovery of mRNA-LNPs, and the hydrodynamic diameter and polydispersity index (PDI) were not changed before (73 to 76 nm and 0.11 to 0.18) and after (69 to 83 nm and 0.10 to 0.24) TFF for both the runs (Table 7). mRNA encapsulation efficiency before and after TFF remained at 99% and mRNA integrity was 100% for both the runs (Table 7). These small-scale runs provided feasibility data that successful LNP encapsulation and UF/DF could be performed while maintaining product stability and quality with the T-series Delta 100 kDa membrane cassettes at loadings > 1 g/m² mRNA dose. Additional optimization could be performed to obtain scale-up parameters at higher loadings while maintaining CQAs.   

Table 7. mRNA-LNP drug product quality and yield after UF/DF3 process

Based on the time average permeate flux of 70 LMH provided by the DC100T01 scale work, a scale-up process of 42 L to achieve 7× VCF and five DV with 1 m² EFA at a feed flux of 6 LMM and 1.03 bar TMP, would be expected to be completed in ~ 57 min. A process time of ≤ 2 h is ideal to complete the entire LNP DP formulation workflow in an 8 to 12 h manufacturing shift, which includes LNP encapsulation, and provide ample time to perform analytics for concentration adjustment followed by sterile filtration.

For the production scale run, purified mRNA after UF/DF2 and bioburden filtration was thawed and diluted to 0.17 g/L in 0.1 M sodium acetate (pH 4.0). Diluted mRNA was encapsulated into LNPs using a proprietary lipid mix by our NanoAssemblr™ commercial formulation system equipped with a SU NxGen™ commercial manufacturing flow kit 48 L/h. A total of 1.9 g of mRNA was encapsulated, resulting in a total output formulation volume of 49 L of mRNA-LNPs in aqueous buffer-ethanol mixture (7.8% ethanol). Of the diluted formulation, 42 L was used as the feed material for UF/DF3, for an input loading of 42 L/m² or 1.6 g/m² of mRNA in the LNP product. Two 0.5 m² T-series cassettes with Delta 100 kDa membrane (DC100T06) were used, for a total filter surface area of 1 m². The UF/DF operation was performed on an ÄKTA readyflux™ TFF system equipped with a SU 0.5” flow kit plus with ReadyMate™ connectors. The filter assembly was sanitized and equilibrated with 1× PBS, pH 7.4. The product was concentrated 5× VCF to a target of ~ 0.2 g/L, then diafiltered into four DVs of proprietary pH 7.4 cryobuffer, which was completed in 46 min. A final concentration was not performed after diafiltration, and product was collected after a recovery buffer chase to obtain ~ 8.3 L of the final product. Collectively, small- (DC100T01 or T02) and large-scale (DC100T06) TFF run data suggest that T-series cassettes with Delta 100 kDa membrane can scale to achieve different VCF (5×, 7× or 18× VCF) and four to five DV at loadings > 1 g/m² of mRNA dose in LNPs to complete the UF/DF process in less than an hour, which is important for maintaining LNP stability/quality in ethanol-containing formulations.

Fig 8. Permeate flux, conductivity, and TMP over time for mRNA-LNP product during concentration to 0.2 g/L and diafiltration into 4× DV of pH 7.4 cryobuffer.

The average permeate flux throughout UF/DF3 operation was 87 LMH (Fig 8). The initial flux was 160 LMH and decreased over the course of concentration through initial fed-batch concentration to 77 LMH, or an average value of 109 LMH during concentration. Figure 9 displays the TFF process during the diafiltration of mRNA-LNPs into cryobuffer. The flux stabilized during diafiltration with an average value of 71 LMH. After the completion of fed-batch concentration and prior to the start of diafiltration, the feed pump was briefly stopped and a momentary reduction in flux and TMP observed at 13 min and 19 min timepoints. During the product recovery stage, a buffer chase with additional cryobuffer was used to increase product recovery. 

Fig 9. Permeate flux, TMP, and conductivity over diafiltration volumes during diafiltration into cryobuffer.

Analytical results indicate that the mRNA-LNP particle size was largely unchanged by TFF processing with a starting size of 70 nm, and a post-TFF size of 68 nm. The PDI increased from 0.05 PDI to 0.14 PDI before and after TFF, a common occurrence with mRNA-LNPs. This may be attributed to the impacts of shear forces experienced by the mRNA-LNPs during processing (5). Encapsulation efficiency was also 99% before and after TFF (Table 7). The differences in size and PDI of the mRNA-LNPs for small- and large-scale runs, as shown in Table 7, could be due to the use of different lipid mix batches, systems, and formulation conditions for LNP encapsulation, but still remain within the CQA specifications (1). The mRNA-LNP drug product was subsequently diluted to a target mRNA dose of 0.1 g/L (Comirnaty dose concentration) for downstream sterile filtration validation studies and frozen at -80°C. Product recovery was > 100% by RiboGreen analysis. Integrity of mRNA in LNPs before and after TFF was ~ 94% and after TFF indicating that the process conditions did not significantly impact the mRNA stability.

Table 8. Calculated conditions and parameters for UF/DF3 of mRNA-LNPs at three different mRNA DS batch sizes.

Table 8 outlines the calculated process parameters and conditions for hypothetical scale-down (0.015 g), pilot scale (0.16 g), and scale-up (8 g) batches, based on the production scale mRNA DS batch size (1.6 g) process data generated here. T-series cassettes with Delta membrane at the Centramate™ (DC100T01 or DC100T12) and Centrasette™ (DC100T06 or T26) scales are fit-for-purpose for these batch sizes of LNP manufacturing. Our diverse range of ÄKTA™ TFF systems (ÄKTA readyflux TFF system 500, flux™ s, flux 6, readyflux™ or readyflux™ XL) with SU flow kits (1/8’’-1’’) are ideal for these different batch sizes that can cover a broad range (0.06 to 30 LPM) of feed flow rates. However, these scaling factors are based on loading of 1.6 g/m² of mRNA dose in LNPs showing feasibility with potential to be further optimized for improved scalability and test the limits of filter capacity. Also, the process volumes in Table 8 are based on the ~ 8% ethanol in the aqueous buffer:ethanol mixture of mRNA-LNP encapsulation volume that may vary depending on the level of dilution performed in this unit operation.

Overall, the small-scale feasibility and large-scale manufacturing runs highlight the scale-up of our 100 kDa TFF filters, whether they are hollow fiber cartridges or flat sheet cassettes, as well as the ÄKTA flux™ s and ÄKTA readyflux™ systems for linear plasmid DNA and mRNA manufacturing process. For process development and large-batch manufacturing scales, hollow fiber cartridges are available in SU format with EFA from 16 cm² up to 28 m² for linear pDNA, whereas the flat sheet cassettes can be stacked to create filter areas in between cassette sizes (0.0093 to 2.5 m²) and > 2.5 m² for mRNA and mRNA-LNPs. Approximately 1 g of linear pDNA was processed in UF/DF0. Total uncapped mRNA processed in UF/DF1 and UF/DF2 was 19 g and 7.4 g respectively, and 1.6 g of uncapped mRNA encapsulated in LNPs were processed in UF/DF3 using the T-series cassettes with Delta 100 kDa membrane. 1.6 g mRNA-LNP drug product at a target Comirnaty vaccine dose mRNA concentration of 0.1 g/L (30 µg/dose) would be equivalent to ~ 50 000 doses of a vaccine. The demonstrated use case here employs the model GFP construct for mRNA (uncapped), and a proprietary lipid composition for mRNA-LNPs, for UF/DF process optimization towards typical vaccine dosage applications. Optimization of the TFF operating conditions at process development scale needs be undertaken for different DNA or RNA construct and size (number of basepairs or nucleotides length) as well as different RNA-LNP lipid compositions before scale-up for drug products at therapeutic dosages.

This application note was authored by Divakara Uppu, Associate Senior Scientist; Roman Kondra, Senior Scientist, MSAT Field Team; Mark Auger Jr., Scientist II; Yaw Sarpong, Senior Scientist, Downstream Process Development; and Kelsey Schwartz, Associate Senior Engineer, Process Development. With thanks for the hands-on support from the teams that were involved in the execution of testing, equipment set up and preparation, and analytical testing that include Sree Gayathri Talluri, Marilyn Patterson, Shawheen Fagan, Steve Turbayne, Matt J. Leprohon, Randolph M. Huelsman, Stephen W. Standring, Brian Bergeron, Mustafees Khan, Stephanie Purkis, Ian Johnson, Logan Molnar, Logan Ingalls, Andrew Tapper, Ashley Braun, Aleksei Angell, and Joanna Yovera.

1. Daniel S, Kis Z, Kontoravdi C, Shah N. Quality by Design for enabling RNA platform production processes. Trends Biotechnol. 2022 ;40(10):1213-1228. doi: 10.1016/j.tibtech.2022.03.012.
2. USP mRNA Draft Guidelines: Analytical Procedures for mRNA Vaccine Quality – 3rd Edition.
3. Uppu D, Auger M Jr, Kondra R, Hejmowski A, Marchand N. T-series TFF cassettes with Delta 100 kDa membranes for RNA and LNP applications. Cytiva. CY42341-120324-AN.
4. Cui T, Fakhfakh K, Turney H, Güler-Gane G, Toloczko A, Hulley M, Turner R. Comprehensive studies on building a scalable downstream process for mRNAs to enable mRNA therapeutics. Biotechnol. Prog. 2023; 39: e3301. doi: 10.1002/btpr.3301.
5. Process development considerations for RNA-LNP therapeutics. Cytiva. December 03, 2024. www.cytivalifesciences.com/en/us/knowledge-center/advance-technology-rna-lnp-process-development