Titer estimation for quality control (TEQC) method: A practical approach for optimal production of protein complexes using the baculovirus expression vector system

The baculovirus expression vector system (BEVS) is becoming the method of choice for expression of many eukaryotic proteins and protein complexes for biochemical, structural and pharmaceutical studies. Significant technological advancement has made generation of recombinant baculoviruses easy, efficient and user-friendly. However, there is a tremendous variability in the amount of proteins made using the BEVS, including different batches of virus made to express the same proteins. Yet, what influences the overall production of proteins or protein complexes remains largely unclear. Many downstream applications, particularly protein structure determination, require purification of large quantities of proteins in a repetitive manner, calling for a reliable experimental set-up to obtain proteins or protein complexes of interest consistently. During our investigation of optimizing the expression of the Mediator Head module, we discovered that the ‘initial infectivity’ was an excellent indicator of overall production of protein complexes. Further, we show that this initial infectivity can be mathematically described as a function of multiplicity of infection (MOI), correlating recombinant protein yield and virus titer. All these findings led us to develop the Titer Estimation for Quality Control (TEQC) method, which enables researchers to estimate initial infectivity, titer/MOI values in a simple and affordable way, and to use these values to quantitatively optimize protein expressions utilizing BEVS in a highly reproducible fashion.

Key technological advancements of the BEVS lie in the ease and efficient generation of recombinant baculoviruses harboring single genes [1,5], or in case of the MultiBac system, harboring multiple genes, enabling expression of numerous proteins simultaneously [22][23][24].
Well-established standard procedures for BEVS allow a streamlined workflow, accounting for its ease of use and high adaptability to routine laboratory operations for many users. Despite of all these technological advances to simplify the generation of recombinant baculoviruses, when it comes to protein expression, the key indicator for optimal expression of proteins remains elusive.
For example, there is no clear answer to the question of whether or not a user would obtain twice as much protein if twice as much baculovirus is added to the insect cells. The situation is a marked difference from the bacterial expression system in which key factors for optimization of protein expression (e.g. IPTG concentration) have been well established [25].
A lack of clarity regarding optimization of protein expression in BEVS may be attributed to conflicting results from previous studies [26][27][28][29][30][31][32]. In these studies, MOI (Multiplicity Of Infection: how many infectious units per cell) was considered to be a key indicator for expression of a model protein, β-galactosidase. While some of these studies showed that varying MOI values had no significant effects [26][27][28], others showed high MOI [29,30] or even low MOI [28,29,31,32] conditions resulted in optimal protein expression. The results of these studies did not provide a clear answer about the correlation between MOI and protein production. The problem is a lack of understanding the physiological outcome of MOI values: for example, MOI=2 means having twice as much virus per cell, but how it affects the cells and how it represents protein expression levels are uncertain. Even if MOI is a good indicator, classical titer/MOI measurement methods, such as the plaque assay [33], and the immunohistochemical staining method (Clontech Inc), are too cumbersome and time-consuming to be practically used for high throughput expression studies. Other titer/MOI measurement methods such as qPCR [34], flow cytometric assays (FCA) [35], measurement of cell diameter change of post infected cells [36], and a cell line containing eGFP for titer measurement [37] are less tedious. However, they require specialized equipment, and setting them up is not trivial.
Therefore, these methods are not practical for many users, whose goal is to produce proteins of their interest and not to measure MOI per se. Furthermore, the condition at which the titer is measured differs vastly from that at which protein expression is carried out, prompting a question as to how relevant such a measurement is for protein production. Taken all together, a simple and affordable experimental method to quantify and optimize expressions of proteins or protein complexes using BEVS needs to be established for many end users.
Mediator Head module is an essential sub-complex of the Mediator complex, which plays a key role in transcription regulation in eukaryotes [38]. We used BEVS to generate the Mediator Head module for structural and functional studies [16,39]. We often encountered the problem of inconsistency in expression level of the Head module, which prompted us to investigate a way to optimize expression of the Mediator Head module. Our research led us to discover the 'initial infectivity' as an excellent indicator of overall production of protein complexes, and we devised a simple and affordable method to estimate this initial infectivity without the need for any specialized equipment or setting. Our finding led to the development of the 'Titer Estimation for Quality Control' (TEQC) method, which enables users to quantify and optimize expression of protein complexes in insect cells, ensuring quality control of protein expression in the BEVS.

Maintenance of insect cells
Hi5 and Sf9 cells were obtained from Expression Systems (Davis, CA). The cells were maintained in ESF921 media (Expression systems) in shaker flasks at 27°C. Both cells were split to a density of 0.5 × 10 6 cells/ml every 3 days.

Construction of baculovirus transfer vectors for expression of multiprotein complexes
All vectors used in this study are summarized in Supplemental Table S1. Construction of the baculovirus transfer vector for expression of modified Mediator Head module used for our structure determination was described in [16]. Construction of the baculovirus transfer vector bearing the genes encoding the yeast Mediator Middle module is as follows: Open reading frames (ORFs) of genes encoding Med7, Med21, Med9, Med4, Med10, Med19 and Med1, were amplified by PCR from the yeast (S. cerevisiae) genomic DNA. Med31 with a C-terminal HA and 10xhistidine tag (Med31-HA-10His) was amplified by PCR from pBacPAK9-Med31-HA-10His vector, which was generated by sub-cloning MED31 gene into BamHI and XhoI sites of pBacPAK9 followed by addition of HA and 10xHis tag sequences by two rounds of QuikChange site-directed mutagenesis (Agilent Technologies).

Virus production and storage
Productions of recombinant baculoviruses in Sf9 cells were performed as described [23].
Liquid viruses were stored at 4°C. Frozen viruses were generated as described (Supplemental protocol) [44]. Frozen virus stocks were stored under liquid nitrogen.

Estimation of Multiplicity of Infection (eMOI)
The initial infectivity is defined as the infectivity 24 hours after infection (I24). An estimation of the initial infectivity is denoted as eI24. The multiplicity of infection (MOI), which is a ratio of infectious units per cell, can be estimated using the following two equations: Where a growth constant, α, is the growth rate of uninfected insect cells in a 24 hours period, C 0 is an initial cell number, and C 1 is the cell number 24 hours after addition of virus. The detailed procedure for estimation of infectivity and MOI is described in the supplementary protocol in Supplemental Information.

Titer measurement by Expression Systems, Inc
Our liquid viruses were shipped to Expression Systems (Davis, CA) for titer measurement.
Titer measurements were performed via flow cytometric analysis (FCA) for expression of the baculoviral coat protein gp64 by Expression Systems, Inc as described [35]. The numbers of gp64 positive cells identified by FCA were used to calculate the infectivity of the baculovirus expressing the Mediator Head module. hours, and were harvested by centrifugation, frozen in liquid nitrogen, and stored as a pellet at -80°C until future use.

Purification of protein complexes
The Mediator Head module was purified as described previously [45]. The yeast Mediator Middle module, core TFIIH, CycC-CDK8, and human Taf8 Hepes-KOH pH7.6, 100 mM potassium acetate, 10% Glycerol, 300 mM Imidazole, and 5 mM βmercaptoethanol). The recombinant yeast TFIIF complex was purified via an IgG affinity column. Lysate from 200 ml culture prepared as described above, was loaded onto 0.5 ml of IgG resin (GE Healthcare). After wash with the high salt and low salt buffer, the recombinant yeast TFIIF was recovered by TEV protease digestion for 1 hour at room temperature. Concentrations of purified protein complexes were measured by Bradford assay.

Assay for solubility of the Head module subunits by quantitative western blotting
Cells infected with the various eMOIs were expressed in 50 ml culture of Hi5 cells (1.0 × 10 6 cells/ml), following the same experimental procedure described above. Cells were harvested in 1.5 ml tubes, cell pellets were frozen in liquid nitrogen, and stored at -80 °C until use. Cell Bands were quantified using ImageJ software package [47]. 9 pmol (2 µg) of the purified Mediator Head module was used as standard.

Results and Discussion
The initial infectivity strongly correlates with overall protein complex production In our previous work, we used the MultiBac baculovirus expression system as our method of choice for high-yield, recombinant protein production [16]: As illustrated in Figure 1A, the multi-gene construct harboring all 7 genes encoding the Mediator Head subunits was generated, and integrated into baculovirus genome (DH10MultiBac) followed by virus production, expression and purification of the complex [16] ( Figure 1A). We often encountered variability in production of the Mediator Head module when switching from one batch of baculovirus to the other, ranging from 3 mg per 1 L culture to 6 mg per 1 L culture (see below). As described in the introduction, given a lack of clarity regarding optimization of protein expression in BEVS, we showing 100 % infectivity after 24h, remained unchanged in their cell density until the end of the incubation. ( Figure 1C). In contrast, infection with batch MB33 virus was incomplete after 24 hours, and uninfected cells continued to divide until 48 hours ( Figure 1C), resulting in a noticeable difference in amount of the cells being harvested (illustrated in Figure 1B below). In short, the discrepancy in the overall expression levels of the Mediator Head module correlates with a difference in initial infectivity of the two batches of virus. Therefore, we hypothesize that the initial infectivity of the baculovirus, which we define here as infectivity 24 hours after virus addition: I 24 , could be an excellent indicator of an overall production of the Mediator Head module.
To test our hypothesis, we pursued to develop a new and simple approach to determine initial infectivitiy (I 24 ) in order to examine the correlation between initial infectivities and expression levels of the Mediator Head module in-depth. Technically, flow cytometric assays [35], or the measurement of cell diameter change of post infected cells using Vi-CELL XR Cell Counter (Beckman Coulter) [36], could be used to measure an initial infectivity (I 24 ). However, as mentioned above, our goal was to devise an easy and affordable experimental method without the use of specialized equipment and setting so that many end users can utilize it. We reasoned that an initial infectivity (I 24 ) could be "estimated" by measuring cell densities: C 0 (number of cells per ml) being the initial cell density and C 1 being the cell density at 24 hours after addition of virus, based on the assumption that (i) infected cells lead to immediate growth arrest [48], resulting in no cell division, and that (ii) uninfected cells divide with a growth constant, α in 24 hours: α indicates how much fold change of uninfected cell growth for a 24 hours period, and it is defined as a ratio of cell densities at the time t and at the time, (t+24 hours): α = cell density at (t+24)/cell density at t. For example, α=2 means that cell number doubles in every 24 hours. We like to emphasize at this point, that we were not detecting the number of infected cells directly.
Rather we attempted to estimate how many cells are infected based on the consequence of virus infection (infected cells led to cell arrest). Thus, we designate this value as estimate of I 24 : eI 24 .
Based on the assumption, the cell density at 24 hours after addition of virus, C 1 can be described as a sum of (i) the number of infected cells (C 0 x eI 24 ) and (ii) the number of the uninfected cells that were divided [(C 0 -C 0 x eI 24 ) x α]: C 1 = C 0 x eI 24 + (C 0 -C 0 x eI 24 ) x α. Therefore, eI 24 can be described as follows: We evaluated if our new method provides estimates of infectivity, that are in line with those measured by commercially available FCA [35]. In this assay, Sf9 cells were infected by serially  Figure 2B). In this experiment, infected cells were incubated for 96 hours (our default setting). As shown in Figure 2B, there is a clear correlation between estimated initial infectivity (eI 24 ), and protein amount of the Mediator Head module: the Mediator Head module yield increases nearly proportionately to an increase in the value of eI 24 until it reaches close to 100%. Furthermore, we examined the correlation between eI 24 and the protein complex yield when infected cells were incubated for a shorter time (48 and 72 hours) ( Figure 2C). The length of incubation period influenced total yield of the complex. However, it did not influence the overall pattern of correlation between eI 24 and yield ( Figure 2C). Taken together, these data supports our hypothesis that our estimated initial infectivity, eI 24 , could be a good indicator of the overall Mediator Head module expression in insect cells. The significance of this finding is that despite the complexity of expressing seven proteins simultaneously, the overall protein yield could be influenced by a 'single' factor, an initial infectivity, which can be easily estimated by counting cell density using a hemocytometer or cell counter. We assume that I 24 estimate (eI 24 ) = 1-e -MOI estimate -Equation 3 (Eq. 3).

Initial infectivity (I 24 ) of a recombinant baculovirus can
By combining equations 1 and 3, I 24 estimate = eI 24 , and MOI estimate (eMOI) can be described as follows: Since MOI = titer/ cell density x cell culture volume (V c ) x virus volume (V virus ), then titer estimate, eTiter, can be described as follows: Our results so far led us to hypothesize the following as illustrated in Figure 3: first, a single factor, the initial infectivity (I 24 ) is a good indicator of overall protein complex expression level as is MOI despite of complexity of multiple proteins being expressed; second, the relationship between I 24 and MOI is nonlinear ( Figure 2D); third, I 24 could be estimated simply by measuring cell density 24 hours after addition of baculovirus with Eq.
1; fourth, baculovirus "titer" can be estimated using Eq. 5; and finally and most importantly, using eI 24 or eMOI value of a given virus harboring multiple genes, expression of a protein complex can be quantified and thus, optimized ( Figure 3).

The cell density measurement based method provides estimations of titer values that are in line with those measured by other titer measurement method
Our findings so far were very encouraging in terms of developing a simple and affordable method to estimate viral titer, MOI and infectivity. It will allow us quality control of protein complex production using the BEVS in insect cells at very early stages of the experiment.
Moreover, the results from our cell density measurement based method are in good agreement with those from FCA method for assessing the initial infectivity of the baculovirus expressing the Mediator Head module. We further evaluated whether our new method can also provide titer estimates that are in line with those measured by one of the conventional titer measurement assays. To this end, we compared titer values obtained from our formula (Eq. 5) (Figure 3) to those from commercially available flow cytometric assay based titer measurements, which provides titer values, that are in good correlation with those obtained by plaque assay [35]. Titers of a total of twelve baculoviruses were measured by both methods and compared ( Figure 4A Figures S2-3). Comparing both methods for the viruses expressing protein complexes, the obtained titer values are in good agreement for five out of six viruses ( Figure   4B). The other viruses, which do not contain genes for recombinant protein expression, and the one encoding the yeast core TFIIH genes titer values measured using our formula (Eq. 5) are on average 2.1 times higher than those determined by the FCA method. In this regard, there are several other studies published comparing different titer measurement methods. For example, a 2-3 fold difference between the end-point dilution assay, and qPCR based assay was observed [51]; the analysis of end point dilution using GFP and anti-DBP (DNA Binding Protein) based assay showed as much as a 15-fold difference between the titer values [52]. Compared to these previous results, clearly, our measurement results are in a reasonable and acceptable range.
Therefore, we conclude that our formula provides good titer estimations and thus, can be extended in its application to aid in the robust expression of many proteins.
As suggested, titer values can be dependent upon medium and cell types [35].  24 . We also used the corresponding eMOI to display because eMOI is intuitively easy to understand: for instance, eMOI=2 indicates adding twice as much volume of virus as eMOI=1.
There is a good correlation between the eI 24 of the expression baculovirus and the overall protein complex production ( Figure 5). Protein complex yield peaked at or near infection saturation point of eMOI being 3 or greater ( Figure 5) except for the case of the Mediator Middle module: its yield peaked at an earlier infectivity point (39.3%; eMOI=0.5) ( Figure 5A). However, the protein complex yield difference between eMOI=0.5 and eMOI=3-5 is relatively small with 10 mg/L and 9 mg/L, respectively, accounting for only a 10% difference. In other words, the eMOI=3-5 is nearly optimal. Key conclusion is that the condition at which eI 24 =100% or eMOI≥3 provides the optimal, or near optimal expression levels for all cases tested. We call this the "eMOI=3 or greater" rule. It is not entirely surprising to see that optimal protein expression is observed when cells are fully infected within the first 24 hours, considering a delay in infection due to lower virus load consequently postpones reaching the level of maximum amount of recombinant protein producing cells and ultimately yield in the given time-frame. Our "eMOI=3 or greater" captures such condition for an optimal protein production.
We further tested the idea that once the cells are more than 95 % infected (eMOI ≥ 3), any additional excess virus would not necessarily increase overall expression level of protein complexes. To test this idea, the Mediator Head module, hTaf8-Taf10 and yeast core TFIIH were chosen because the Head is our model complex and other two are on different ends of the spectrum in size and expression levels: hTaf8-Taf10 is considered a relatively small complex (81 kDa), and was highly expressed ( Figure 5A) while yeast core TFIIH is a relatively large complex (6 subunits, 320 kDa) and not expressed well ( Figure 5B). For all three complexes (Supplemental Finally, we tested if an increase in yield of protein complex could be attributed to an increase in overall expression of the complex or an increase in solubility of the subunits. We used the Mediator Head module as a test case since it showed most change in its yield between low and high eMOI ( Figure 2D). We compared soluble and insoluble fractions of the Head module with different eMOI ranging from eMOI=0.1 (10% infected initially) to eMOI=4 (~100% infected) by immunoblotting using antibodies against Med17, Med18 and Med11 subunits (Supplemental Figure S6A-C). The ratio between soluble and insoluble fractions of Med17, Med18 and Med11 appears to stay constant, suggesting that an increase of I 24 or eMOI leads to an increase in overall production of the Mediator Head subunits.
The results of a total of six multi-protein complexes are consistent with our hypothesis that eI 24 (or eMOI) can be an excellent indicator of expression levels of multi-protein complexes. By varying eI 24 or eMOI value, the optimum expression points of all complexes can be identified.
But in the end, the condition at which eI 24 =100% or eMOI ≥3 provides the optimal, or near optimal expression levels ("eMOI=3 or greater" rule). In essence, our simple and affordable cell density measurement based method provides eI 24 or eMOI value of a recombinant baculovirus. Researchers (end users) can use these values to quantitatively optimize expression levels of proteins or protein complexes of their interest. We named our newly developed quantitative method "TEQC" method: "Titer Estimation for Quality Control" (TEQC) of protein or protein complex production.

TEQC method can be utilized to achieve reproducible expression of protein complexes to ensure quality control
We tested whether we could use the eMOI value to achieve reproducible expression of protein complexes, resulting in a production of the recombinant protein complexes with maximum yield consistently. The conditions for maximum expression levels of 6 protein complexes are as follows: in the Hi5 cells, the Mediator Head module peaked at eMOI of 3.7; hTaf8-Taf10 at eMOI of 3.5, yeast CDK8-CycC at eMOI = 3.5, and Mediator middle module at eMOI= 0.5; in SF9 cells, yeast core TFIIH at eMOI of 3.0, and yeast TFIIF at eMOI of 3.0 ( Figure 5), respectively. For simplicity, we are referring to eMOI only. All these six protein complexes were expressed using the same optimum eMOI and cell line but were expressed on different days.
These independent expressions were repeated three times for all complexes. The yield of each purified protein complex was measured, the numbers were averaged, and compared to the original expression level that provided the optimum eMOI value in that cell line. For all tested complexes, the yield of purified protein was highly reproducible ( Figure 6A).

Conclusion
This work has revealed the connection among initial infectivity, MOI, and an overall recombinant protein production. We developed a simple and affordable cell density measurement based method to estimate initial infectivity, MOI, and titer, leading toward the development of the TEQC method for optimization of a production of protein complexes using the BEVS. Key finding is that optimal or near optimal expression levels are reached when eMOI ≥ 3, which our "eMOI=3 or greater" rule. Since expression of single proteins could be viewed as "one" subunit multi-protein complex, the TEQC method should be applicable to expression of single proteins as well. Additionally, expression of a protein complex in insect cells was carried out in this work by infecting a single virus harboring multiple genes encoding the subunits of the complex, and not by co-infecting multiple viruses. The strategy for optimization of protein complex expression using the co-infection method will be a future subject of investigation.       Middle were expressed at their optimum infectivity or eMOI, and each protein complex yield was measured; labeled as "initial" (dark blue). Expression was repeated three times using the same eMOI or infectivity value for each complex independently, purified protein complex yields were measured and the numbers were averaged; labeled as repeats. The repeats (green) were compared with the initial protein expression optimization results (dark blue). (B-E) Mediator

Author contributions
Head, Taf8-Taf10, core TFIIH, and TFIIF complexes were expressed using freshly made new high titer viruses as well as the old viruses with declined titers. The eTiter values of both new and old viruses were determined and volume of each old or new virus was adjusted such that the expression of each complex was carried out at the same eI 24 or eMOI. After expression and purification, protein complex yield was measured, and the data were plotted against either eI 24 or eMOI. (B) Mediator Head module was expressed in Hi5 cells (red) (C) Taf8-Taf10 was expressed in Hi5 cells, (D) TFIIH core was expressed in Sf9 cells, (E) TFIIF was expressed in Sf9 cells, and compared with the initial expression results (blue) and the results from old viruses (red).