Experimental Validation with Amino Acid Supplementation
To validate the OPLS Model for each Criterion, experimental conditions
were designed similarly to the high nutrient process control with the
addition of amino acid cocktail feeds. CHO cell cultures fed with amino
acid cocktails based on Criterion 1 showed a nominal increase in total
cell growth at day 9 for both models (Growth Model vs Control: no
significant change; Production Model vs Control: no significant change)
(Fig. 3a). Day 14 CVC was also analyzed to assess if the early additions
of amino acid up to day 9 would have an impact on the overall cell
growth of the culture. Similar to day 9, there was only a slight
increase of less than 10% in totals cells from the growth model (p =
0.1712) suggesting that increased supplementation of amino acids that
were already being consuming in excess based on the theoretical demands
did not further drive cell growth. In some instances, overfeeding
nutrients can have deleterious effects on the cell culture (Salazar,
Keusgen, & von Hagen, 2016). For instance, the growth model showed a
significant drop in titer near 20% (p < 0.05) and a slight
drop in titer for the production model (p = 0.159) (Fig. 3b). For
comparison, day 9 titer was also analyzed to determine if the growth
model for Criterion 1 had any early impact on productivity.
Unfortunately, no deviation from the control was observed (Growth Model
vs Control: no significant change; Production Model vs Control: no
significant change).
In contrast, Criterion 2 was designed to test those amino acids that
were being consumed below their theoretical demand. However, since
Criterion 2 was built on a basis of high nutrient feeds in the training
dataset, we postulated that information gained could also support
various phases of media development such as moderate nutrient or low
nutrient feeds. Therefore, three feed media conditions for the growth
and production models were included with biological replicates for the
validation run. Similarly to Criterion 1, amino acid cocktail feeds were
prepared and supplemented on model-informed days. For statistical
comparisons of the model conditions to the control, CVC and relative
titer values were only considered from day 8 to day 14 because a
majority of the cultures showed similar CVC and mAb titer during the
first half of the culture. Although day 9 CVC did not show a significant
change for high nutrient conditions, day 14 CVC showed a modest increase
of about 10% for the growth model and about 20% for production model
(Fig. 4). A more significant increase was observed within the moderate
nutrient and low nutrient conditions when supplemented with the same
amino acid cocktail feeds. Increase in CVC was noticed as early as day 7
for both models in the moderate nutrient condition reaching about a 35%
increase in total cells by day 14 for the growth model (p <
0.001) and about 30% increase for the production model (p <
0.05). The low nutrient condition also showed an increase in CVC as
early as day 7 but relative cell growth propagated to near 55% increase
in total cells for both models by day 14 (Growth Model vs Control: p
< 0.01; Production Model vs Control: p < 0.05).
Interestingly, the relative cell growth rates were only slightly
impacted by the amino acid supplementations (Fig. S3). The growth rate
for high nutrient conditions showed a minor increase at day 3 during the
log-phase of growth but also maintained the slight increased up to day 7
(p < 0.05) which could have been the driver to the 20%
increase in CVC. Similarly, the growth rates for moderate and low
nutrient conditions also slightly increased, but unlike the high
nutrient condition, the growth rate remained elevated for a longer
duration up to day 9. Moderate nutrient conditions had a 5% increase in
growth rate at day 3 in the growth model whereas the low nutrient
condition had a 5% increase around day 5 reflecting a delayed log-phase
but overall greater increase in CVC relative to the control.
The increase in cell growth for both models was further reflected in mAb
productivity. Although the high nutrient growth model did not result in
any significant increase in titer, the production model yielded about
25% increased productivity (p = 0.163) (Fig. 4). The moderate and low
nutrient productivity in contrast reflected a significant increase in
cell growth by producing about 80% more relative titer by day 9 and
60% more relative titer by day 14 when compared to control cultures (p
< 0.001). Although low nutrient conditions had a higher boost
in cell growth, it also resulted in a similar 60% increase in
productivity by day 14 suggesting an upper threshold for mAb production
by this cell line with those particular feeds. The large increases in
mAb titer for moderate and low nutrient conditions were further
reflected by the significant increases in relative specific productivity
(qP) (Fig. S2). For example, the moderate nutrient condition showed a
20% increase in qP for the growth model (p < 0.05) and showed
a near 30% increase for the production model during the stationary
phase (p < 0.01). In contrast, the low nutrient condition
showed about a 40% increase in qP starting from day 7 (p <
0.05) suggesting the increased mAb production was not just an artefact
of increased cell growth.
To further understand the differences between the resulting cell growth
and mAb productivity phenotypes observed, glucose consumption rates were
analyzed to assess how the model-driven conditions and supplementation
of presumably limiting amino acids could alter cellular metabolism.
Posed differently, was differential glucose consumption artefact of the
increased percentage of cells within each model condition or a result of
altered metabolic activity at the cellular specific level? Total glucose
consumption within high nutrient feeds did not show any change in
overall consumption whereas under moderate and low nutrient conditions,
there was about a 2 g/L decrease in peak glucose consumption (Fig. 5).
In contrast, the specific glucose consumption rates highlighted a more
significant difference between control and model conditions within
moderate and low nutrient feeds. Growth and production model conditions
resulted in almost a 1000 pmol/cell*day decrease in specific glucose
consumption in moderate feed and about a 2000 pmol/cell*day decrease in
low nutrient feeds (p < 0.01). Accordingly, the marked
increase in cell growth and mAb production for moderate and low nutrient
conditions when compared to the high nutrient condition was further
reflected by the larger magnitude of specific glucose consumption. Taken
together, the decreased specific glucose consumption rates in
model-driven conditions suggests that cells within these conditions
presumably consumed other nutrients such desired amino acids for
bioproduction and energy generation that ultimately resulted the cells
to be in a higher metabolic state to produce more mAb.
Since the amino acid cocktails were supplemented in addition to process
feed media and prepared at 100x concentrations, modulations in feed
volume, pH, and osmolality were warranted. Specific control conditions
for each model at all three nutrient feed levels for Criterion 2 were
included within the experiment to understand if the increase in CVC or
productivity was in fact due to the amino acids and no other cell
culture variations. The solution/pH control was designed for the high
nutrient feed only and was supplemented with base at the same volume and
pH as the amino acid cocktail with time points representative of either
of the growth or production models. Similarly, osmolality control
cultures were fed with a NaCl solution with the same osmolality and
volume as the highest osmolality among the amino acid feed cocktails.
High nutrient feed cultures supplemented with the solution/pH control
feed showed about a 20% decrease in day 14 CVC and day 14 titer (Fig.
S3). Due to the high pH levels, supplementation of the solution/pH
control could have increased cell death and a decline in productivity.
Comparably, cultures supplemented with the osmolality control feed for
high nutrient conditions also showed about a 20% drop in total cell
growth and productivity (p < 0.35). However, moderate and low
nutrient conditions supplemented with the osmolality feed control did
not show any significant differences from the control cultures
suggesting that the accumulation of additional osmolality by the
cultures in high nutrient conditions was an artefact of the already
nutrient rich environment and not of the control feed. Therefore, the
effects of osmolality and the pH change from the amino acid cocktail
feeds did not contribute to the overall increase in cell growth or
productivity, but it was in fact a result of the key amino acids
supplementation.
The growth and production models for Criterion 3 represented those
nutrients that were consumed less than their theoretical demand but were
not deemed important by the model VIP statistic. Presumably, these
nutrients were considered not to significantly impact cell growth or
productivity within all three nutrient conditions. Although day 9 CVC
was chosen as the response variable, relative cell growth rates from
Criterion 2 showed that the peak cell growth occurred prior to day 7.
This is further reflected in the experimental results for Criterion 3
growth model where there was insignificant change to total cells
measured in response to day 8 amino acid supplementation for all three
nutrient feed levels (Fig. 6a). Similarly, growth model conditions for
all three nutrient levels resulted in about a 20% decrease in relative
titer when compared to the control (Fig. 6b). Overall, the production
model resulted in about a 5% increase in total cells for high and
moderate nutrient conditions, and about a 15% decrease in total cells
for the low nutrient conditions when compared to the control. Relative
titer on the other hand showed a 10% increase for the high nutrient
condition (p < 0.3) and about a 40% increase for both
moderate and low nutrient conditions (p < 0.05). Although 40%
increase was presumed to be significant, it was not as significant as
the 90% increase in the analogous cultures within Criterion 2 (p
< 0.001). We presume that the large increase in titer from
moderate and low nutrient feed conditions in Criterion 3 was due to the
low nutrient levels in the culture and that any additional nutrients
could provide a positive effect.