High-throughput production of human proteins for crystallization: the SGC experience

Abstract

Producing purified human proteins with high yield and purity remains a considerable challenge. We describe the methods utilized in the Structural Genomics Consortium (SGC) in Oxford, resulting in successful purification of 48% of human proteins attempted; of those, the structures of approximately 40% were solved by X-ray crystallography. The main driver has been the parallel processing of multiple (typically 9-20) truncated constructs of each target; modest diversity in vectors and host systems; and standardized purification procedures. We provide method details as well as data on the properties of the constructs leading to crystallized proteins and the impact of methodological variants. These can be used to formulate guidelines for initial approaches to expression of new eukaryotic proteins.

Supplementary Figure 1

(A) Target diversity; (B) Crystallization class distribution of proteins solved by the SCG (Oxford).

Fig. 1

Cloning pipeline. Distribution of the number of constructs generated per each target (A) or each target domain combination (B). Productive targets/domains are those for which at least one construct led to soluble, purified protein; non-productive targets/domains could not be produced in substantial amounts from any of the constructs generated. (C) Distribution of cloned fragments generated per target/domain (same as panel B, but fragments cloned in different vectors are counted only once).

Fig. 2

Schematic domain organization and construct design. Fes tyrosine-protein kinase: the black line represents the full-length gene. The three PFAM-annotated domains (FCH, SH2, and P-kinase) are depicted in boxes; the blue and red lines represent the gene fragments that were all cloned and tested for expression and crystallization. Note that the blue and red lines represent distinct constructs. HSD11B1 (11-beta-hydroxysteroid dehydrogenase 1): a demonstration of different definitions of structured domains.

Fig. 3

Fraction of soluble constructs per target/domain. For each productive target/domain (as defined in Fig. 1), the fraction of constructs that tested positive for soluble expression was scored. The proportion of domains with different fractions of soluble constructs is displayed in histogram form.

Fig. 4

Ends of successful constructs relative to the predicted and observed structured regions. The constructs leading to 282 unique structures were compared with structured regions predicted by PSIPRED (A), the observed structured region in the crystal structure (B), and the annotated PFAM domains (C). The length of extensions in the N-termini (white) and C-termini (full) are depicted as histograms. Note the different scale in panel C.

Fig. 5

Presence of tags on crystallized proteins. Size distribution of proteins that were crystallized with the tag (open) or following tag cleavage (full).

Fig. 6

Multiconstruct testing in baculovirus. Ten constructs of one gene (A) and five constructs of a second gene (B) were cloned in baculovirus and expression was tested following infection of 3-ml suspension cultures of SF9 cells (Shrestha et al., 2008). The cells were lysed after 48 h, the lysates were clarified by centrifugation and the recombinant His6-tagged proteins were purified by Ni-affinity. Aliquots of the uncentrifuged total lysate (T) and the NiNTA-eluate (E) were analyzed by SDS–PAGE. The positions of the recombinant proteins are marked with (*).