For the mutant leader selections, the fluorescein-binding single chain antibody 4m5.3 (Boder et al. 2000
) was used as a model heterologous protein. The 4m5.3 scFv was cloned under the direction of a library of S. cerevisiae
alpha mating factor 1 prepro signal sequence peptides preceding a Kex2p recognition site for leader peptide cleavage. The library was created through error prone PCR mutagenesis of the native prepro sequence followed by transformation into yeast via
homologous recombination with the 4m5.3 ORF to yield a library of approximately 108
variants. The selection for improved secretion was carried out using the Cell Surface Secretion Assay (CeSSA) (Manz et al. 1995
; Rakestraw 2006
). The CeSSA technique uses cell surface-bound antigen to capture the secreted scFv against the antigen (in this case using a fluorescein antigen covalently bound to the library of cells secreting the anti-fluorescein scFv 4m5.3). This assay has been shown to be an effective way to select for the improved secretion of the scFv. After the last round of sorting, plasmid DNA was isolated and retransformed into fresh cells that were then tested for 4m5.3 scFv secretion. All of the isolated and characterized mutant leaders imparted improvement in 4m5.3 secretion over the wild-type leader (WTαpp) with the most productive mutant leaders yielding up to a sixteen-fold improvement (the eight best clones are shown in ). The isolated leaders were sequenced and analyzed for similarities (sequences of the eight best leader sequences are shown in ). The sequence data show definite trends in leader mutations. The 22nd
valine residue in the WTαpp is universally changed to an alanine residue. Moreover, the hydrophobic LLFI motif spanning the 63rd
to the 66th
residue is mutated to residues with more polar side chains. The impact of these two types of mutations will be addressed later.
One of the greatest difficulties in generating hyperproductive strains for secretion is the inability to develop strains with enhanced secretion independent of the protein being expressed. To test whether the αpp8 and αppS4 leaders isolated from the selection could be used to improve the secretion of proteins other than 4m5.3 scFv, the two leaders were used to direct the secretion of the lysozyme binding scFv D1.3 and a disulfide-stabilized version of the carcinoembryonic antigen (CEA) binding scFv sm3E (Graff 2004
). The data show that the αpp8 and αppS4 leaders enhance the secretion of both types of single-chains (). Additionally, the two leaders were used to enhance the secretion of horseradish peroxidase as well as a poorly expressing mutant of interleukin-2, IL-2 (Q126R). These results demonstrate that leader-imparted improvements in heterologous protein secretion can be generalized to other scFvs as well as structurally unrelated proteins.
Figure 2 Secretion of the scFvs 4m5.3, D1.3, and sm3E as well as the IL-2 mutant 2-4(Q126R) and HRP was assayed under the direction of the WT, αpp8, and αppS4 leaders. Student′s t-tests comparing average secretion titers between the WTαpp (more ...)
Immunoglobulin fragments such as scFvs and Fabs have been successfully expressed in yeast (Gasser 2007
). However, with previously reported yields as low as 50 μg/L, S. cerevisiae
has not been shown to secrete the full-length IgG in titers sufficient for use as an expression host (Horwitz et al. 1988
). In hopes to remedy this shortcoming, the αpp8 and αppS4 leaders were used to direct the secretion of the full-length heavy chain and the N-terminally FLAG-epitope-tagged light chain of a mouse/human chimeric 4m5.3 IgG1
. These chains were expressed from two different auxotrophically marked low copy plasmids. Protein gels of protein A and FLAG-purified supernatant show a band consistent in size with a full-length human IgG1
() that is reactive with both anti-human and anti-FLAG antibodies when probed by Western blot (data not shown). Additionally, smaller molecular weight bands consisting of partially assembled IgG are purified. When reduced with DTT, the IgG bands collapse into a 30 kDa light chain band (containing the FLAG epitope tag), two Fc-containing bands around 50 kDa, as well as a good deal of higher molecular weight protein. Treating the IgG with the N-glycosidase EndoH results in the two major forms of the heavy chain exhibiting a slightly increased mobility. The new mobility of these two bands is consistent with that of the DTT-reduced, non-N-glycosylated 4m5.3 N297Q mutant, which removes the lone N-linked glycosylation site in hIgG1
, demonstrating that the fully assembled IgG contains N-linked Fc glycosylation. The Fc site is the only N-linked glycosylation site in the final, fully processed form of the molecule. Higher molecular weight, EndoH sensitive forms of the N297Q mutant (as well as some of the higher molecular weight WT IgG) are likely non-Kex2 cleaved protein still retaining the N-glycosylated pro region. Further treatment of the IgG with mannosidase causes the upper heavy chain band to collapse into the lower band, suggesting that a subpopulation of the secreted heavy chain has some O-linked glycosylation in addition to the typical N-linked Fc glycosylation.
Figure 3 (a) SDS-PAGE of Protein A and FLAG purified 4m5.3 IgG1 generated with the αpp8 leader. Abbreviations: std (control human IgG1 standard), WT (wild-type 4m5.3 IgG with app8 leader), N297Q (4m5.3 N297Q non N-glycosylated mutant with app8 leader). (more ...)
4m5.3 IgG secretion was further enhanced through manipulation of the host strain. IgG secretion directed by the αpp8 and αppS4 leaders was compared to the wild-type leader with and without the overexpression of protein disulfide isomerase (PDI), a chaperone previously shown to be beneficial to scFv expression () (Robinson et al. 1994
; Shusta et al. 1998
). Both improved leaders, in combination with PDI, enhanced IgG secretion 40-fold over the wild-type leader alone with about 25-fold of that improvement due to the mutant leaders. In previous studies, it has been shown that heterologous protein expression can put significant negative selection pressure on low-copy plasmid retention resulting in as much as 50% of cells to be plasmid negative (see (Boder et al. 2000
) and (Rakestraw and Wittrup 2006
) for comparison of a yeast surface displayed protein utilizing genes expressed from low-copy plasmids versus integrations). To compensate for expression instability caused by the dual plasmid system, chromosomal integrations of the two genes were made resulting in an additional four-fold improvement in secretion. The total yield of 4m5.3 IgG secreted from the enhanced expression strain in five mL culture was determined to be approximately 9 mg/L by a fluorescein quench assay and quantitative Western blotting. The total yield of a one-liter shake flask culture after protein A and FLAG purification was determined to be approximately 1.5 mg. Combined, these strategies impart a 180-fold improvement in IgG secretion over previously reported yields making S. cerevisiae
a productive host for the expression of full-length IgG.
To test the functionality and specificity of the yeast-produced IgG, purified 4m5.3 IgG was used to bind fluorescein-labeled cells. Specificity was measured by labeling the cells with biotin instead of fluorescein. Additionally, the 4m5.3 IgG was preincubated with 10μM fluorescein before being added to the cells (). A human, type-matched, polyclonal IgG was also used as a negative control in the labeling experiments. To test the flexibility of the yeast host for the generation of IgGs with different variable domains, the variable heavy and light chains of the anti-EGFR 225 IgG and the anti-CEA scFv sm3.E were cloned into the yeast IgG vector to express as hIgG1 chimeras. Antibody titers for 225 and sm3E in 1 L cultures were similar to or better than the yields derived for 4m5.3 expression. The yeast-produced 225 IgG binds to the EGFR-expressing A431NS cell line. Competition experiments show that the IgG does not show strong binding when cells are pre-incubated with murine 225 or when the IgG is pre-incubated with soluble 404SG, a yeast produced version of the extracellular domain of EGFR (). Additionally, the sm3E IgG labels the CEA-expressing LS174T cell line but is blocked by pre-incubating the cells with sm3E scFv (). Functional expression of these three antibodies, derived from different mouse variable genes and further altered by chimerization, suggests that this system can robustly express full-length antibodies regardless of their family or origin.
Figure 4 (a) Labeling of fluorescein-conjugated yeast with purified 4m5.3 IgG followed by an anti-human PE-conjugated antibody and analysis by flow cytometry. A type-matched, non-specific, human IgG1 kappa polyclonal antibody (Hu IgG1 polyclonal) and cells labeled (more ...)
The αpp8 and αppS4 leaders have been shown to improve the secretion of a range of proteins. The isolated leaders (listed in ) also exhibit some common sequence motifs. The two most common types of mutations in the improved leaders are the A22V mutation in the early pro region and a tendency toward more polar residues in the WTαpp LLFI motif found in the 63rd to 66th residues of the prepro. To examine the effects of the V22A mutation, a mutagenic cycle was performed on the leader at this position (). In this cycle, the V22A mutation that is ubiquitous in the improved leaders was also mutated in the WTαpp. Additionally, the alanine at the 22nd position of the mutant leader αpp8 was reverted back to the wild-type valine. Looking at the secretory output of the mutants in the V22A mutagenesis cycle (), it is apparent how important the V22A mutation is to the αpp8 leader, yet this mutation does not benefit the WTαpp leader at all. Although the V22A mutation is essential to the enhanced activity of the αpp8 leader, the improvement is only fully realized within the context of the other mutations. This dependence is an important realization as it implies that the mutations found in the αpp8 leader work in concert with one another. Therefore, it may be supposed that toggling the 22nd residue back and forth between valine and alanine could have different effects on protein trafficking depending on whether it is done in the context of the WTαpp or αpp8 leader. Perhaps the V22A mutation overcomes a secretory bottleneck that is only encountered when the other αpp8 mutations are present. Alternatively, the V22A mutation might interact directly with another mutated residue on the leader in order to be effective. Because of this synergy between the various mutations, it is more relevant to look at individual mutations in the αpp8 leader in hopes of ascertaining the contribution of each mutation to the secretory improvement imparted by the leader as a whole. In following this strategy, the polar LSST motif in the αpp8 leader was reverted back to the wild-type LFFI or mutated to the more polar but bulkier residues LEDE. Together, these data show that there is a functional preference for more polar residues in place of the native LLFI motif in the mutant leaders.
Figure 5 (a) Mutational analysis of improved leaders. The WT Val22 was mutated to alanine, and the Ala22 in αpp8 was mutated back to the WT valine. The effect of the hydrophobicity of the 63rd-66th residues on secretion was also examined by mutating αpp8(LSST) (more ...)
Although the secreted forms from both wild-type and mutant leaders are consistent with mature 4m5.3 scFv in size, companion intracellular Western blots show that there are two forms of intracellular scFv: a glycosylated, higher molecular weight, pre-Kex2 proprotein form in addition to a full-length, mature form (). Intracellular immunofluorescent microscopy suggests that a considerable amount of intracellular protein is localized to a large organelle consistent with the vacuole in size and structure (data not shown). The vacuole has previously been shown to be a common destination for heterologous protein that has been mis-sorted from the Golgi (Coughlan et al. 2004
; Rakestraw and Wittrup 2006
; Steube et al. 1991
; Zhang et al. 2001
). A survey of intracellular 4m5.3 scFv and its distribution between the pre-Kex2 form and the mature form was made by analyzing the band intensity of Western blots for two separate sets of experiments (). The amount of protein found in the pre-Kex2 and mature forms as well as the total intracellular protein (pre-Kex2 plus mature) for the WTαpp, αpp8, and two of the αpp8 mutagenic cycle leaders are shown. For easier comparison, all of the data have been normalized to the total intracellular protein found from the WTαpp leader. The percent of total intracellular protein made of the mature form for each leader is also shown. This analysis of intracellular 4m5.3 considered with the effects of the leaders on secretion in show that on the whole, the αpp8 leader redistributes total protein from the intracellular forms to the supernatant when compared to WTαpp. Furthermore, changing the alanine at the 22nd
position in the αpp8 leader back to the wild-type valine causes a drastic reduction in pre-Kex2 protein accumulation compared to αpp8. This result suggests that this alanine in the αpp8 leader may reduce Kex2 cleavage efficiency somehow resulting in preferential trafficking to the surface when the pro region is finally cleaved (no proprotein was found in the supernatant). Alternatively, an alanine at the 22nd
position may give the αpp8-led protein more secretory pathway residence time, which is a characteristic previously correlated with improved secretion (Kjeldsen et al. 1997
; Luo 2002
). In any case, the αpp8(A22V) mutation reveals a secretory bottleneck characterized, in part, by relatively less pre-Kex2 intermediate than the unmutated αpp8 leader. The alanine in the 22nd
position of the αpp8 leader is a key part of overcoming this bottleneck. The αpp8 LLFI reversion has little effect on total intracellular protein accumulation or intracellular protein distribution compared to αpp8.
Because vacuolar mis-targeting seems to be a common fate for heterologous protein including the WTαpp 4m5.3 scFv here, it may be that the wild-type leader itself contains sequence elements that make it more likely to engage one of the vacuolar sorting proteins. This engagement would cause the protein to be trafficked to the vacuole instead of the surface. We hypothesized that the αpp8 leader removes the sequence elements responsible for this vacuolar mistrafficking causing more protein to be secreted. To test this hypothesis, chromosomal deletions of three vacuolar sorting proteins often implicated in heterologous protein trafficking, VPS8, VSP10, and VPS13 (Bowers and Stevens 2005
; Cowles et al. 1997
; Harsay and Bretscher 1995
; Harsay and Schekman 2002
; Zhang et al. 2001
) were made in the expression strain. The VPS8 and VPS13 mutations do improve the secretion of wild-type leader scFv almost two-fold, suggesting that vacuolar sorting could play some role in the retention of protein directed by the WTαpp (). However, the VPS mutations do not account for all of the secretory improvement seen in the αpp8 leader. Deletions of the ER associated degradation (ERAD) protein DER1 (Knop 1996
) or the vacuolar sorting protein VPS10 had no effect on secretion.
Figure 6 Three vacuolar sorting genes, VPS8, VPS10, VPS13, as well as an ERAD associated gene, DER1, were deleted, and the resulting 4m5.3 secretory competency under the WT and αpp8 leaders was tested. The data are normalized to the secretion in the WT (more ...)