Biomedical Engineering Reference
In-Depth Information
120.00
100.00
80.00
60.00
mAb1 % yield
Main charge isoform
% Monomeric mAb1
CHOP
DNA
40.00
20.00
0.00
4
4.25
4.75
5
5.25
5.5
6
6.5
6.8
Harvest pH
Figure
7.2.
Effect of harvest pH on product attributes and the level of CHO host-cell impurities
after centrifugation. The
-axis represents the levels of product and impurities in the centrate,
normalized to the pH 6.8 harvest condition. Adjustment of the cell culture harvest medium to
acidic pH causes cells to flocculate, resulting in more efficient centrifugation. In addition, the
pH-inducedflocculation results in reduced levels of CHOhost-cell protein (CHOP) andDNA in the
centrate. Harvest at reduced pH did not alter the product (mAb1) recovery, charge profile, or
level of aggregation.
y
the product concentration in the centrate or product quality attributes such as aggregation
and charge variants.
These findings indicate that harvest pH affects the purity of the load onto the rPA
capture column and the harvest-capture design space was explored together in a series of
DOEs. Other centrifugation parameters (e.g., g-force and flow rate) did not significantly
alter the properties of the filtered centrate (rPA column load) across relatively wide
ranges of operation. These parameters were not a significant source of process variation
and offered little or no opportunity for process control. Thus, these “nonkey” parameters
were not included in the subsequent design space studies. As mentioned earlier, product
recovery and product attributes are relatively insensitive to the harvest pH. In addition,
DNA is easily removed by the rPA and subsequent chromatography steps. Thus, the
primary response for the DOE studies was determined to be the clearance of CHOP. The
sequence of these experiments is summarized in Table 7.2. Briefly, of the 19 input
parameters identified for the operation of the rPA capture column, 7 were identified as
nonkey parameters based on prior process knowledge and were not carried forward into
the resolution IV screening design. Of the 12 parameters tested in the screening DOE, 4
were shown to have significant effects on the process or the product and carried forward
into the subsequent modeling design. The experimental design matrix for the modeling
DOE is shown in Table 7.3. For this exercisewith four input variables, a D-optimal design
comprising 32 chromatographic runs carried out in two blocks was selected. This
orthogonal response surface design (power
>
80 and 95% confidence interval) was
