Optimizing cell harvesting in a closed and automated manner is important when it comes to reducing processing time for expanded cell therapy products, and can help quickly deliver treatments to patients. Cytiva offers the CultureWash C-Pro application software for the Sepax™ C-Pro system to perform automated processing, maximize cell recovery, and minimize operator intervention while avoiding open, manual techniques. Using a model human B-lymphoblastoid cell line 721.221-mIL21 (source: Rutgers University – hereinafter “221s”), we tested a predictive mathematical formula for calculating minimum intermediate volume that is applicable up to 4 billion total cells for a final product dose of 20 mL, and suggest an optimal parameter set for the CultureWash application when above this cell number. We achieved recovery of > 88% for harvesting up to 4 billion cells using a standard parameter set and were able to increase recovery from ~ 68% to > 86% for larger cell number pellet size with our optimal parameter set.

Introduction: CultureWash C-pro harvest formula

In our prior study (Optimizing recovery on the Sepax™ cell processing system: Rinse volume and manual extraction guidelines for 20 mL final product volume), we used a model T-cell line (Jurkat cells) to investigate cell recovery during cell harvest using the CultureWash C-Pro application for the Sepax™ C-Pro system. We found that cell recovery for a given total cell number is a function of the final (FV), intermediate (IV), and rinse (RV) volumes, with the major driver of cell loss coming from the dead volume (DV) left in the kit. In this study, we propose a predictive formula (Equation 1) to determine the minimum IV required for different numbers of cells to harvest. The formula hypothesizes that the cell diameter plays the most important role in the overall pellet size, which, in turn, determines the minimum intermediate volume needed to accommodate that pellet. We set out to test 3 conditions (total cell number:intermediate volume) and kept the FV at 20 mL for each condition:

  1. 2 billion:5 mL
  2. 4 billion:10 mL
  3. 6 billion:15 mL

Equation 1: Hypothesized formula to predict minimum IV for Sepax™ C-Pro kits.


IV = minimum intermediate volume [mL]

C = Total cell number being processed [# cells]

D = average cell diameter [m] (e.g., 16 µm = 1.6e-5 m)

1E6 = volumetric unit conversion constant [cm3 per m3]

1.2 = 20% extra volume buffer factor

As an example, with the average cell diameter of 221s being 16 µm, our formula predicts that roughly 2.57 mL of IV is required for every 1 billion cells including 20% extra volume as a buffer region. Ideally this would be a linear relationship since the diameter of the cells did not change between runs, thus every extra billion cells should take up the same volume in the chamber. However, we did not observe this in practice, as we explain later.

We use the CT-60.1 single-use kit that is compatible with the CultureWash C-Pro application. While we would typically recommend a CT-90.1 single-use kit for volumes of 20 mL or less to allow for manual extraction (ME), we wanted to keep ME off as this option does not allow for changes to IV to with a fixed FV — i.e., we couldn’t alter IV while keeping the FV fixed at 20 mL.

Though we set out with the goal to show the linearity of the formula for a given cell diameter, irrespective of total pellet size, we observed higher-than-expected cell loss as the pellet size grew beyond 4 billion cells. Thus, as shown in the Results and discussion section below, the linearity of the formula for a fixed cell diameter was only demonstrated up to a cell pellet size of 4 billion in the CultureWash C-Pro application for a final volume of 20 mL, with further guidance on alteration to the parameter set for applications requiring greater cell number in a fixed FV.

In summary, the study aimed to:

  1. Evaluate the hypothesis that IV is a function of cell diameter, and in turn, cell pellet volume.
  2. Evaluate the impact of IV on total cell recovery.
  3. Verify the validity of the formula as cell pellet size increased.

Materials and Methods

Cell culture

Previously expanded 221s (Cell source: Rutgers University , expanded internally) were thawed from a 1 mL vial of 1e8 cells frozen in CS-10 (Crystor) . 9 mL of warmed RPMI (Hyclone™ serum) with 10% FBS (Hyclone™ serum) was added to the thawed vial to neutralize the freezing medium and then centrifuged at 300 × g for 10 minutes. Supernatant was discarded and cells were resuspended in complete medium (RPMI (1L), 10% FBS, 1% Pen/Strep (100 ×, Hyclone™ serum), 1% SG-200 (200 mM Hyclone™ serum), 1.25% v/v Glucose solution (200 g/L stock, Gibco) at 5e5 cell/mL in T-175 flask. Flasks were subcultured daily, splitting into a new T-175 flask diluting with fresh media to 5e5 cells/mL. All cell counts were done daily in triplicate on the NC-200 (Chemometec), along with biochemistry data being acquired on the Bioflex2 (Nova Biomedical) to verify sufficient metabolite levels and monitor growing conditions. Once the culture reached 5e8 cells total, the cells were seeded into a 10 L Xuri™ system Cellbag™ bioreactor at 5e5 cells/mL in 1L of complete media (parameters: 6 RPM rock rate, 6° angle, 0.1 mL/min gas flow rate with compressed air, 5% CO¬2, 37°C). Media was added daily to reset the culture to 5e5 cells/mL until 5 L total volume was reached. The day following maxing out the working volume at 5 L, a 0.5 × perfusion rate was used. The day following the perfusion start, the perfusion rate was set to 1 ×. This method for maxing out bag volume and increasing media perfusion in a step-wise fashion was followed whenever cells were taken out, resetting the total number of cells in the bag. The only exception was for the final two data points (IV 15 mL, 6B, slow fill/extraction) where we manually perfused (harvesting all the volume, centrifuging cells, and resuspending in 5 L of fresh media, 1 × mock perfusion) due to the filter fouling from extended culturing. This feeding and perfusion schedule was necessitated by the high number of cells required for the runs, while utilizing the same 10 L Xuri™ system Cellbag™ vessel for the completion of the study. While we could just remove cells and batch feed after removal for the lower cell number conditions, to reach densities required for the more extreme cases at 5 L working culture volume, we had to perfuse overnight to provide proper growth conditions.

Sepax™ C-Pro system runs

Cell counts were taken prior to the Sepax™ C-Pro system run to calculate the volume necessary to harvest from the culture. The proper volume was harvested via the Xuri™ harvest line using the CBCU unit peristaltic pump into an appropriately sized Labtainer™ bioprocess container (Thermo Scientific). Cells were then bulk centrifuged to remove spent media then resuspended in 117 to 206 mL for initial bag volume. The bulk centrifugation step was required as the densities in the Xuri™ bag were not great enough to get to the processing cell number required below the chamber volume (220 mL). Keeping the initial volume below chamber volume allowed all runs to have an equal number of concentration steps (1), irrespective of cell number, thereby normalizing for any variability that might have arisen from not doing so. The wash buffer for the run was then made (PlasmaLyte A (Baxter), 10% Human Serum Albumin (25% stock, Gemini cat. 800-120), 1 mM EDTA (0.5 M EDTA stock, Invitrogen). The Sepax™ C-Pro system was then programmed with the proper pre-determined settings (see table below), and we followed all on-screen prompts. An exception to the prompts was during the line stripping into the final bag, both the filter line and line that goes from the stopcock to the chamber were also stripped.

Table 1. Sepax™ C-Pro parameters: parameters that varied across conditions in bold

CultureWash C-Pro v432 parameters Units Values
Total cells cells 2e9-6e9
Initial volume mL 117 to 206
Detect initial volume No (0) Yes (1) 0
Dilution ratio 0
Dilution speed mL/min NA
60s-mixing after dilution No (0) Yes (1) NA
Thaw bag validation No (0) Yes (1) 0
Hang bag validation No (0) Yes (1) NA
Input bag rinsing No (0) Yes (1) 1
Pause input bag rinse No (0) Yes (1) NA
Optical cell detection No (0) Yes (1) 0
Product filling speed mL/min 120 (30*)
Waste extraction speed mL/min 120 (30*)
Intermediate volume mL 5-15
Intermediate volume mL 5-15
Pre-wash cycles cycles 1
Pre-wash g-force × g 400
Pre-wash sedimentation time seconds 300
Switch washing solution No (0) Yes (1) 0
Post-wash cycles cycles 0
Post-wash g-force × g NA
Post-wash sedimentation time seconds NA
Switch resuspension solution No (0) Yes (1) 0
Exchange waste bags No (0) Yes (1) 1
Final volume mL 20
Manual extraction No (0) Yes (1) 0
Syringe for manual extraction No (0) Yes (1) NA

*Value differs in the 4th run set only, demarcated as “-slow” in Figure 1 and Figure 3.


Recovery of various processes were calculated as output cell number expressed as a percentage of the total cells that went into the run based on average cell counts in triplicate from the NC-200 (Chemometec). Cells quantified in various parts of the process/kit were diluted in media to keep within the range of 5e5 cells/mL to 2e6 cells/mL to get the most accurate counts within the linear range of the NC-200. Some endpoint solutions fell below that optimal range, but were still within the specification of the counter (5e4-5e6 cells/mL) and could still be utilized. If the solution fell below the lower limit (5e4 cells/mL), the solution was centrifuged at 300×g for 5 minutes and resuspended in ~1 mL media so that it fell within the counter range. To get the volume left in the kit, we utilized the manual purge tool to move the piston to displace the remaining volume into a pre-weighed conical tube.


We used a BioProfile™ Flex 2 Automated Cell Culture Analyzer (Nova Biomedical) to monitor the Xuri™ culture’s chemistry and gas via offline sampling. All cell counts were performed in triplicate using a Via1- Cassette™ device on a NucleoCounter™ NC-200 Automated Cell Counter (Chemometec). Samples were diluted in media as appropriate prior to cell count. We estimated volumes in bags and tubes by subtracting the empty vessel weight from the full vessel weight.


We determined statistical significance in GraphPad Prism 8 using an ordinary one-way ANOVA and Tukey’s multiple comparison test with a single pooled variance. Significance is denoted in each graph as * for p < 0.05, ** for p < 0.01, *** for p < 0.0001.

Results and discussion

Cell recovery and viability

Average recovery for the 2 and 4 billion cell conditions was greater than 88% (Fig 1) (2B:IV5 = 90.8% +/- 0.5%, 4B:IV10 = 89.2% +/- 0.6%). The 6 billion cell condition with an IV of 15 mL had a significantly lower recovery, while using the same CultureWash C-Pro parameter set. We were able to improve the 6 billion cell condition recovery to 86.7% +/- 6.2% by decreasing the filling/extracting speed of the piston chamber, see discussion below. The drop in cell viability during a run was less than 1% for each run, that is, either a drop from 98% in the initial bag to 97% in the final bag, or 97% to 96%, depending on the day.

Fig 1. Recovery in the final bag for each condition. Each bar represents the average of technical triplicate +/- SD, with each replicate shown in black. Differences are only significant when compared to 6-Billion-IV15, where higher cell loss was experienced.

We expected that increasing IV in accordance with the total cell number would maintain high recovery, but we found that recovery was weakly and negatively correlated with total cell number (Fig 2). As the cell number increases, the integrity of the cell pellet starts to diminish, as suspension cells do not stick together well. Once the pellet grows large enough, cells on the top layer are not held strongly enough by the centrifugation force alone and are more susceptible to shearing off during supernatant extraction (one of the main drivers of cell loss). Another interesting correlation was around cell diameter. As the actual cell diameter fluctuated slightly over the culture as density and media conditions change, there was also a weak negative correlation between cell diameter and total recovery. This makes intuitive sense, as smaller cells will form a smaller pellet that will fit better within the IV and be less susceptible to sheering off into the wash.

In our predictive formula, we assumed that a cell occupies the volume of a rigid sphere of constant diameter as part of our theoretical equation and we do not include the impact of packing efficiency. In practice, however, this is not the case, as the mechanobiology of the cell membrane is much more complex. Factors such as cell cycle, growth media, buffers, activation status, and cell-cell interactions can provide biophysical stimuli that affect cell deformation, diameter, and therefore packing efficiency. The addition of a 20% buffer in the equation partially accounts for this deformation, however, additional refinement and testing may be required to account for cell-type- and process-specific factors.

Fig 2. Pearson Correlation table between IV, recovery, diameter, and cell number.

Total cell loss

Breaking down compartmental loss shone some light on where cell loss was occurring (Fig 3A). In all runs, less than 0.7% of cells were left in the initial bags, measured by washing the initial bag with 50 mL of media. Cells lost in the kit (calculated from cells extracted by manually purging the Sepax™ C-Pro separation chamber) also were minimal; < 1% in all runs except for one run of the 4 billion cell condition. In the two lower cell number conditions, the loss was roughly 5%, or lower (Fig 3B). In the initial 6 billion cell condition we experienced high cell loss, upwards of 37%. Mitigations will be discussed in the next section, in which we explain how we brought the average cell loss in the final condition down to an average of 3.4%. The remaining cell loss was calculated by subtraction to complete the mass balance (this likely represents a combination of cells remaining in the lines/chamber and systematic error in volume estimation and cell counts) and is denoted as “Mass balance – Loss to kit” in Figure 3A.

Fig 3. (A) Breakdown of cell loss for each condition by compartment. (B) Cell loss during washing steps. Differences are highly significant when compared to 6-Billion-IV15, where higher cell loss was experienced during wash steps. Each bar represents the average of technical triplicates, +/- SD.

Decreased cell recovery due to speedy parameter set

During waste extraction of the supernatant (from concentrating the cells and washing) for the 6 billion cell condition, we observed visible cell clumps and cloudiness. While our formula should have included enough headspace in the supernatant above the pellet to avoid losing any cells, turbulence near the pellet during extraction likely contributed to the cells shearing off. One possible driver is the biology of the cell type used in this study; since 221s are derived from a lymphoblastoid lineage and grown in suspension, it is hypothesized they can also easily shear off. In addition, suspension grown cells can rapidly undergo cell diameter changes depending on the suspension medium in which they are grown, affecting the pellet size significantly. An additional factor that can likely contribute to some of the variability is that cells might not distribute evenly upon centrifugation along the periphery of the Sepax™ C-Pro separation chamber, thus extracting the waste too quickly can lead to turbulence that can shear off the cells from the pellet. This was confirmed when we reduced the filling and extraction speed parameter and saw improved recovery comparable to conditions with fewer total cells (Fig 1).

Other Parameters to help increase recovery

Several other conditions were tested prior to adjusting filling speed to increase recovery. As examples, we tested adjusting IV to 20 mL for the 6 billion cells condition (n = 1, data not shown) to see if the formula was incorrect and therefore required a “fudge factor” in the IV (the formula predicts 15 mL as the IV for this condition). Surprisingly, this did not help, as this run had a similar, lower recovery to the 15 mL condition with the fast separation chamber filling/extracting (~ 60%, data not shown). This further illustrates that while IV is important to contain the whole pellet, it cannot recover cells that shear off during the extraction operation, and other factors might need to be considered.

Another parameter we used was increasing the centrifugation g-force to 600 × g and sedimentation time to 10 minutes to get better pelleting (n = 1; data not shown). This led to a marginal increase in recovery to 75% (data not shown) but did not fully rescue the run. It is important to adjust these parameters based on your cell type to avoid causing cell death from the high centrifugation forces.

Note that all these studies were done with the optical sensor disabled, which helps ensure the FV and RV are exactly as set. If cell recovery is more important than a maximum final volume, enabling the sensor could prevent cells from being shunted into the waste bag.

Future work will dive deeper on the impact of centrifugation g-force and other parameters, such as varying cell types and diameters, on overall recovery.


In this study, we tested three conditions varying IV based on total cell number to keep the volume of the cell pellet from being extracted into the waste bags on the CultureWash C-Pro V432; namely, 2, 4, and 6 billion 221 cells, and a final volume of 20 mL. Overall, > 88% recovery was achieved in the linear range of a predictive formula up to 4 billion cells for a given cell diameter of ~16 µm, with similar recovery for large pellets with an adjusted CultureWash C-Pro parameter set.

Cell loss from wash steps in optimal parameter sets was 5% or less, illustrating that the predicted IVs were sufficient to maintain the cell pellet. When we experienced large cell loss during the wash step, we saw large clumps and turbid wash buffer, likely due to the filling/extraction speeds being too rapid and shearing some of the looser cells from the pellet-supernatant interface. Our recommendation for pellets greater than 4 billion cells is to set the filling and extracting speed to 30 mL/minute, particularly if run time is less crucial than cell recovery.

While the formula and parameter guidance tested in this study will aid in preventing cell loss during wash steps, our recommendation is to enable the Optical Cell Detection parameter in the CultureWash C-Pro application to automate the process and further minimize the potential for cell loss.


Cytiva gratefully acknowledges the support from Rutgers University in supplying the 721.221-mIL21 cell line (Yang, Y., et al., Molecular Therapy-Methods and Clinical Development, 2022 [https://pubmed.ncbi.nlm.nih.gov/32695845/]) that was used in this work.