Synthesis of granulocyte–macrophage colony- stimulating factor as homogeneous glycoforms and early comparisons with yeast cell-derived material Qiang Zhanga,1, Eric V. Johnstona,1, Jae-Hung Shiehb, Malcolm A. S. Mooreb, and Samuel J. Danishefskya,c,2 aLaboratory for Bio-Organic Chemistry and bCell Biology Program, Sloan-Kettering Institute for Cancer Research, New York, NY 10065; and cDepartment of Chemistry, Columbia University, New York, NY 10027 Contributed by Samuel J. Danishefsky, January 7, 2014 (sent for review December 18, 2013) Granulocyte–macrophage colony-stimulating factor (GM-CSF) is bearing complex glycan domains at each native position, is illus- a medicinally important glycoprotein, used as an immunostimulant trative of the growing power of chemical synthesis to facilitate following bone-marrow transplant. On the basis of reports of its the study of medicinally relevant biologic targets (12–14). We potential utility as an anticancer vaccine adjuvant, we undertook describe herein the synthesis of homogeneous GM-CSF glyco- to develop a synthetic route toward single-glycoform GM-CSF. We proteins through a convergent route that offers ready access to describe herein a convergent total synthesis of GM-CSF aglycone a menu of glycoforms for further study. and two homogeneous glycoforms. Analytical and biological studies Structurally, GM-CSF consists of 127 aa with multiple sites of 27 confirm the structure and activity of these synthetic congeners. glycosylation (Fig. 1). Two N-linked glycans are located at Asn and Asn37 (15, 16). Interestingly, neither the position nor the glycoprotein synthesis | peptide ligation | alanine ligation extent of O-linked glycosylation has been unambiguously es- tablished. Studies suggest a range from two (Ser7 and Ser9/Thr10) 5 7 9 10 – to four (Ser , Ser , Ser , and Thr ) sites of glycosylation (17). ranulocyte macrophage colony-stimulating factor (GM- The tertiary structure is strongly influenced by two cross-linked GCSF) is a secreted glycoprotein that promotes cellular disulfide bonds. growth and proliferation. GM-CSF signaling initiates a cascade that culminates in the production of white blood cells through Results and Discussion stem-cell stimulation in the bone marrow (Fig. 1) (1). Thera- Our strategy for reaching homogeneous GM-CSF constructs peutically, this glycoprotein is valued for its properties as an relies upon powerful techniques developed for the synthesis of immunostimulant; GM-CSF impacts the production, differenti- complex biologics: namely, solid-phase peptide synthesis (SPPS), ation, and function of dendritic cells through potentiation of the + native chemical ligation (NCL), and metal-free dethiylation CD4 T-cell response (2, 3) and is regularly administered to (MFD). Merrifield’s SPPS method permits stepwise elongation patients undergoing chemotherapy or autologous bone-marrow of a growing peptide (18, 19). Convergence in the construction of CHEMISTRY transplant. GM-CSF is approved in Europe as the aglycone longer polypeptides, including small proteins, may be achieved (Molgramostim) (4) and in the United States as a glycosylated, through the use of the seminal NCL technology of Kent and mutant form (Sargramostim). At present, glycosylated GM-CSF coworkers, whereby fragments are merged at N-terminal cysteine is obtained exclusively via recombinant technologies using residues (20, 21). The possibility of expanding the scope of NCL yeast (Sargramostim) or Chinese hamster ovary (CHO) cell to encompass, for instance, alanine ligation was demonstrated in a key (Regramostim) technologies, which yield complex mixtures of paper by Yan and Dawson (22), who accomplished dethiylation glycoforms. The glycan heterogeneity reflects a lack of specificity in CHO-cell posttranslational glycosylation. The degree of Significance GM-CSF glycosylation has been reported to affect the in vivo properties of the glycoprotein (5, 6); the aglycone may cause As biologically active glycoproteins are increasingly investigated increased adverse side effects, perhaps due to its enhanced as potential therapeutic agents, there is a growing demand for susceptibility to truncation pathways (7). the development of strategies for the synthesis of homoge- In light of our longstanding interest in synthetic anticancer and neous, single-glycoform constructs for the purposes of rigorous HIV vaccines (8), we took note of reports suggesting that GM- evaluation. Currently, most glycoproteins are accessed through CSF might serve as a useful vaccine immunoadjuvant. Admin- recombinant methods as complex mixtures of glycoforms. istration of GM-CSF results in robust potentiation of the im- – mune response, and clinical studies suggest that the glycoprotein Granulocyte macrophage colony-stimulating factor (GM-CSF) is may hold promise as an adjuvant for anticancer vaccines (9, 10). an important glycoprotein therapeutic, used to stimulate the However, studies to date have used recombinantly derived GM- immune system following bone-marrow transplant and che- CSF mixtures, and results have been inconclusive (9); perhaps motherapy. GM-CSF is also being explored as a potential im- such translational issues might be addressed in a more in- munoadjuvant for anticancer vaccines. We have completed formative fashion with homogeneous GM-CSF agents. In a more a chemical synthesis of homogeneous, single-glycoform GM-CSF. speculative line of inquiry, we also wonder whether appendage of Through adaptation of this modular synthetic route, it will be tumor-associated carbohydrate antigens to the protein backbone possible to gain access to a menu of single-glycoform GM-CSF might yield powerful new anticancer vaccine candidates (11). congeners for a wide range of biological studies. Even beyond these fascinating medicinal questions, we recog- nized in GM-CSF a synthetically compelling glycoprotein target. Author contributions: Q.Z., E.V.J., J.-H.S., M.A.S.M., and S.J.D. designed research; Q.Z., E.V.J., With these considerations in mind, we undertook to design a and J.-H.S. performed research; Q.Z., E.V.J., J.-H.S., M.A.S.M., and S.J.D. analyzed data; and modular route to homogeneous GM-CSF glycoforms. In this Q.Z., E.V.J., J.-H.S., M.A.S.M., and S.J.D. wrote the paper. endeavor, we would rely upon key methodological and strategic The authors declare no conflict of interest. advances from many groups, including ours, in the synthesis of 1Q.Z. and E.V.J. contributed equally to this work. complex glycoproteins. The past decade has indeed witnessed 2To whom correspondence should be addressed. E-mail: [email protected]. a dramatic maturation of the field of “biologics” through chemical This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. synthesis. Our recent synthesis of homogeneous erythropoietin, 1073/pnas.1400140111/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1400140111 PNAS | February 25, 2014 | vol. 111 | no. 8 | 2885–2890 Downloaded by guest on October 1, 2021 Fig. 1. GM-CSF 3D structure and fully glycosylated GM-CSF sequence (1). The ribbon structure of GM-CSF is based on the X-ray of GM-CSF aglycone expressed by Escherichia coli. via Raney nickel mediation (22, 23). In 2007, our laboratory de- With the component fragments in hand, methods to accom- scribed mild MFD conditions that enable major expansion of the plish their merger were evaluated (Fig. 4). In the event, coupling menu of NCL beyond cysteine residues (24). Thus, subsequent to of fragments 1a (1) and 2a (2) in pH 7.0 kinetic NCL buffer cysteine-type ligation, the free thiol group may be readily removed afforded 3. The crude reaction mixture was directly subjected to through MFD to reveal either native alanines or a range of amino MFD. Happily, Cys34 was smoothly reduced to Ala34 (3→4)whereas 54 acid residues at the site of ligation (Fig. 2). the Thr -SEt functional group remained intact. NCL of fragments 3a (6) and 4a (7) followed by treatment with MeONH2•HCl Synthesis of the GM-CSF Aglycone provided polypeptide 8 with all of the cysteine residues depro- 4 8 We first set out to accomplish a synthesis of the GM-CSF agly- tected. The NCL-based coupling of and afforded the GM-CSF 9 9 cone. We envisioned dissection of our target into four fragments, aglycone primary sequence ( ). Processing of construct ,asshown 10 each hopefully accessible via SPPS. The key connections would in Fig. 4, produced fully synthetic, folded GM-CSF aglycone . [In one of the few papers that report yield (27), the authors include a kinetic alanine ligation between fragments 1a and 2a, – as well as two cysteine ligations to join fragments 2a, 3a, and 4a disclose a comparably low yield range, from 14 42%. That article (Fig. 3). As a critical design element, we envisioned that kinetic attributes the low yield to the aggregation-prone properties of ligation between fragments 1a and 2a would be followed by MFD the protein. In our estimation, the modest folding yield is due to at the ligation site (Cys34→Ala34) in the presence of a thioester at the aggregation as well as unspecific binding to the folding cassette.] The properties of aglycone 10 are described in Fully Synthetic C-terminal Thr53 residue. The cysteine residues of fragments 3a GM-CSF: Analytical Characterization and Biological Studies (28). and 4a would be protected as t-butyl thioethers; the thiols would be liberated in the course of the ligation. Adoption of this approach Synthesis of Glycosylated GM-CSF Analogs would obviate the need for extensive protection and deprotection of We now directed our attention to our long-term goal of reaching the cysteines slated for “survival” in the final target. single glycoform versions of GM-CSF. Although our central ob- Our synthetic vision for the GM-CSF aglycone was reduced jective was to install glycan domains of a high-complexity order, to practice (Fig. 4). The synthesis commenced with the prepa- 6 54 simulating those likely to be found in glycoprotein cytokines ration of fragment 3a ( ). Thus, protected polypeptide Thr - (29), we thought it prudent to begin with a more modest level of Pro94 (5) was retrieved from the SPPS resin by treatment with 95 glycosylation.