Humanization of Yeast to Produce Complex Terminally Sialylated Glycoproteins

See allHide authors and affiliations

Science  08 Sep 2006:
Vol. 313, Issue 5792, pp. 1441-1443
DOI: 10.1126/science.1130256


Yeast is a widely used recombinant protein expression system. We expanded its utility by engineering the yeast Pichia pastoris to secrete human glycoproteins with fully complex terminally sialylated N-glycans. After the knockout of four genes to eliminate yeast-specific glycosylation, we introduced 14 heterologous genes, allowing us to replicate the sequential steps of human glycosylation. The reported cell lines produce complex glycoproteins with greater than 90% terminal sialylation. Finally, to demonstrate the utility of these yeast strains, functional recombinant erythropoietin was produced.

The half-life and therapeutic potency of most glycoproteins, with the notable exception of antibodies, is dependent on the presence of terminal sialic acid. The exposure of other terminal sugars such as mannose, N-acetylglucosamine, and galactose on a glycoprotein leads to clearance by sugar-specific receptors or lectins (1, 2). Because most therapeutic glycoproteins require sialylation, their production to date has relied on mammalian hosts, which are able to perform humanlike N-glycosylation, including the ability to add terminal sialic acid. Yeast and filamentous fungi offer numerous advantages as recombinant protein expression systems when compared with mammalian cell culture, including higher recombinant protein titers, shorter fermentation times, and the ability to grow in chemically defined media. However, wild-type yeast glycosylate proteins with high-mannose type N-glycans (3) (Fig. 1A), which reduce half-life and compromise therapeutic function. Our laboratory has been engineering human glycosylation pathways into fungal hosts, including the yeast P. pastoris (46). The core repertoire of human glycosylation reactions (Fig. 1A) requires the sequential removal of mannose by two distinct mannosidases (i.e., α-1,2-mannosidase and mannosidase II), the addition of N-acetylglucosamine (by N-acetylglucosaminyltransferase I and II), the addition of galactose (by β-1,4-galactosyltransferase), and finally the addition of sialic acid by sialyltransferase. Sialylation, the final step of human glycosylation, is particularly difficult to accomplish in yeast, because wild-type yeast lacks all four prerequisites: (i) the ability to produce the N-glycosylated precursors terminating in β-1,4-galactose, (ii) the biosynthetic capability to produce the sugar nucleotide precursor cytidine monophosphate (CMP)–sialic acid [specifically, CMP-N-acetylneuraminic acid (CMP-NANA)], (iii) the transporter to shuttle CMP–sialic acid into the Golgi, and (iv) a sialyltransferase to transfer sialic acid to terminal galactose on the nascent glycoprotein (fig. S1) (7). All of these elements must work at high efficiency to allow for the production of sialylated glycoproteins, and organelle-specific targeting of several elements is required to permit these functions to occur in concert.

Fig. 1.

N-linked glycosylation pathways and characterization of N-linked glycans released from recombinant rEPO. (A) Representative N-linked glycosylation pathways in humans and P. pastoris. Mns: α 1,2-mannosidase; MnsII: mannosidase II; GnTI: β 1,2-N-acetylglucosaminyltransferase I; GnTII: β 1,2-N-acetylglucosaminyltransferase II; GalT: β 1,4-galactosyltransferase; SiaT: α 2,6-sialyltransferase; MnT: mannosyltransferase. (B to D) rEPO secreted from P. pastoris strains YSH551 (B), YSH557 (C), and YSH597 (D) was purified from culture supernatants by Ni-affinity chromatography. Glycans were released by PNGase-F treatment and labeled with 2-AB before HPLC analysis. (E) Glycans secreted from YSH597 containing sialic acid were treated with α-2,3-/-2,6-/-2,8-sialidase. Elution times for commercial glycan standards corresponding to GS5.0, 5.5, and 6.0 were 24, 40, and 66 min, respectively.

Erythropoietin (EPO) is a hematopoietic glycoprotein that stimulates the differentiation of late erythroid progenitor cells to mature red blood cells (8). Since a recombinant source of the protein has become available, EPO has found wide therapeutic use in the treatment of anemia. EPO is a heavily glycosylated protein, with three N-glycosylation sites and up to 40% of the molecular mass attributable to its glycans (9, 10). As with many glycoproteins, the therapeutic efficacy and receptor affinity of EPO relies on the degree and composition of N-glycosylation (11).

A previously reported glycoengineered strain of P. pastoris produces terminally galactosylated biantennary glycans of the complex type Gal2GlcNAc2Man3GlcNAc2 (i.e., Glycan Structure 5.0, GS5.0) (12). This strain RDP750 [Δoch1, Δpno1, Δmnn4B, Δbmt2, Δhis1, Kluyveromyces lactis and Mus musculus uridine diphosphate (UDP)–GlcNAc transporters, Mus musculus α-1,2-MnsI, Homo sapiens β-1,2-GlcNAc transferase I, Rattus norvegicus β-1,2-GlcNAc transferase II, Drosophila melanogaster MnsII, Schizosaccharomyces pombe Gal epimerase, D. melanogaster UDP-Gal transporter, and H. sapiens β-1,4-galactosyltransferase] was transformed with an expression plasmid encoding for rat EPO (rEPO) to generate strains RDP762 and YSH557 (7). Secreted rEPO from YSH557 consisted predominantly of GS5.0 N-glycans (Fig. 1C). For comparison, the same construct was used to transform wild-type P. pastoris NRRL-Y11430, resulting in strain YSH551, which secreted rEPO with mostly high-mannose N-glycans that are typical for this yeast (3) (Fig. 1B). Both strains displayed similar growth characteristics and expressed rEPO at about the same level (∼20 mg/liter), although the proteins differed significantly with respect to their N-glycosylation (Fig. 1, B and C).

To accomplish the final step of human glycosylation, the addition of terminal sialic acid, we transformed P. pastoris strain RDP762 with a range of DNA constructs encoding for enzymes involved in CMP–sialic acid biosynthesis, CMP–sialic acid transport, and sialic acid transfer to the nascent glycoprotein. In total, five enzymes were selected: H. sapiens UDP-N-acetylglucosamine-2-epimerase/N-acetylmannosamine kinase (GNE), H. sapiens N-acetylneuraminate-9-phosphate synthase (SPS), H. sapiens CMP–sialic acid synthase (CSS), M. musculus CMP–sialic acid transporter (CST), and a library of chimeric sialyltransferases linked to yeast type-II transmembrane localization peptides (ST).

More than 120 permutations of alternative CMP–sialic acid pathways, CSTs, and STs were screened. From this screen, we identified a small number of combinations displaying significant sialyltransferase activity and producing predominantly complex glycan structures. However, taking this approach we were unable to obtain glycan compositions containing >60% GS6.0 (Sia2Gal2GlcNAc2Man3GlcNAc2).

To further improve the efficiency of sialylation, we codon-optimized GNE, SPS, CSS, and CST, and screened additional ST/leader fusions. The catalytic domain of mouse α-2,6-ST, fused to the Saccharomyces cerevisiae mannosyltransferase 1 (Mnt1) targeting signal was particularly effective. To consolidate these efforts, we cloned all five genes into a single expression vector, pSH926 (7). Transformation of this vector into RDP762 complemented the histidine auxotrophy of the host, while targeting the gene cluster to the TRP2 locus of the Pichia genome. The resulting strain, designated YSH597, was cultured in shake flasks to secrete rEPO. Analysis of the N-glycans isolated from rEPO displayed a glycan composition that consisted predominantly of sialylated glycan structures GS6.0 (90.5%) and GS5.5 (7.9%, SiaGal2GlcNAc2Man3GlcNAc2) (Fig. 1D). Subsequent treatment of this sample with sialidase showed quantitative conversion to GS5.0 (Fig. 1E), confirming that GS6.0 and GS5.5 were terminally sialylated glycans.

To assess the activity of these two different rEPO glycoforms in vivo, we purified material from wild-type P. pastoris (YSH551) and YSH597 (7). The purified protein was characterized by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 2A), and rEPO from wild-type P. pastoris showed extensive heterogeneity consistent with hyperglycosylation and the range of high-mannose structures found by high-performance liquid chromatography (HPLC) (Fig. 1B). In contrast, rEPO from YSH597 showed a more uniform migration pattern, consistent with the glycan uniformity found by HPLC (Fig. 1D). As expected, when both samples were treated with peptide N-glycosidase F (PNGase-F), to remove the N-glycans, the mass and uniformity of the deglycosylated material appeared similar (Fig. 2A). To compare the functionality of these two vastly different glycoforms, animal studies were performed to determine their respective erythropoietic function. As expected, rEPO produced in wild-type yeast had no measurable erythropoietic function, whereas rEPO produced in YSH597 showed a dose-dependent response consistent with a biologically active form of the protein (Fig. 2B).

Fig. 2.

Characterization of recombinant rEPO obtained from P. pastoris. (A) SDS-PAGE analysis of recombinant rEPO (2.5 μg, postpurification) secreted from YSH551 and YSH597 strains after incubation in the presence (+) or absence (–) of PNGase-F. (B) Comparative hematocrit analysis of recombinant rEPO secreted from YSH551 (blue and green bars) and YSH597 (red and yellow bars). Values correspond to days 8 (blue and red bars) and 15 (green and yellow bars) after initial injection. Data presented as mean ± SD of n = 5 mice per dose.

We report the generation of yeast cell lines of P. pastoris with a substantially reengineered secretory pathway. These cell lines secrete terminally sialylated, complex, bi-antennary glycoproteins as exemplified by rEPO, as well as other recombinant proteins tested. The availability of such yeast cell lines may eliminate the need for mammalian cell culture in the future and allow for the production of therapeutic glycoproteins in a nonmammalian host. While significantly reducing production time and viral containment issues, this will also provide improvements in product uniformity and overall production economics. Previously, a panel of glycoengineered yeast cell lines displaying a limited repertoire of human glycosylation reactions allowed us to elucidate glycosylation-dependent structure activity relationships (12). Here, we have engineered into yeast the most complex step of human N-glycosylation, terminal sialylation, which will expand our ability to conduct structure-function investigations.

Supporting Online Material

Materials and Methods

Figs. S1 and S2


References and Notes

View Abstract

Navigate This Article