From deep sequencing to actual clones

Protein Engineering, Design & Selection vol. 27 no. 10 pp. 301–307, 2014 Published online September 1, 2014 doi:10.1093/protein/gzu032 From deep sequencing to actual clones Sara D’Angelo1,4, Sandeep Kumar2,3, Leslie Naranjo2, Fortunato Ferrara1, Csaba Kiss2 and Andrew R.M. Bradbury2,4 1 New Mexico Consortium, Los Alamos, NM, USA, 2Bioscience Division, Los Alamos National Laboratory, Los Alamos, NM, USA and 3Present address: Compugen USA, Inc., San Francisco, CA 94080, USA 4 To whom correspondence should be addressed. E-mail: (A.R.M.B.); (S.D.) Edited by Valerie Daggett The application of deep sequencing to in vitro display technologies has been invaluable for the straightforward analysis of enriched clones. After sequencing in vitro selected populations, clones are binned into identical or similar groups and ordered by abundance, allowing identification of those that are most enriched. However, the greatest strength of deep sequencing is also its greatest weakness: clones are easily identified by their DNA sequences, but are not physically available for testing without a laborious multistep process involving several rounds of polymerization chain reaction (PCR), assembly and cloning. Here, using the isolation of antibody genes from a phage and yeast display selection as an example, we show the power of a rapid and simple inverse PCR-based method to easily isolate clones identified by deep sequencing. Once primers have been received, clone isolation can be carried out in a single day, rather than two days. Furthermore the reduced number of PCRs required will reduce PCR mutations correspondingly. We have observed a 100% success rate in amplifying clones with an abundance as low as 0.5% in a polyclonal population. This approach allows us to obtain full-length clones even when an incomplete sequence is available, and greatly simplifies the subcloning process. Moreover, rarer, but functional clones missed by traditional screening can be easily isolated using this method, and the approach can be extended to any selected library (scFv, cDNA, libraries based on scaffold proteins) where a unique sequence signature for the desired clones of interest is available. Keywords: antibody/inverse PCR/deep sequencing/phage display/yeast display Introduction Next-generation sequencing (NGS) (Niedringhaus et al., 2011) has been widely implemented in projects that go beyond genome sequencing, for which it was primarily developed: protein evolution (Schlinkmann et al., 2012), computationally designed drug screening (Whitehead et al., 2012), in vitro (Ravn et al., 2010) and in vivo (Reddy et al., 2010) antibody selections (reviewed in Georgiou et al., 2014), interactome (Di Niro et al., 2010) analysis and immune repertoire profiling # The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: 301 Received June 18, 2014; revised June 18, 2014; accepted July 3, 2014 (Glanville et al., 2011) have all benefited greatly from the ability to sequence massive numbers of clones. In vitro display technologies (reviewed in (Rothe et al., 2006; Bradbury et al., 2011)) represent the common denominator for most of the selection/evolution approaches in that they allow the isolation of biomolecules with desired properties from large libraries, using multiple rounds of selection. Traditionally, candidate binders are identified by extensive individual clone screening using microtiter plates. This is an inefficient approach beset by redundancy for abundant clones, and sparse, or absent, representation of clones present at lower frequencies within the selected population. Our group (Di Niro et al., 2010), along with others (Ravn et al., 2010), has pioneered an alternative approach that exploits the use of NGS in in vitro display selection analyses. The entire selection output (usually 105 – 6 clones) is analyzed by deep sequencing. Sequences are binned, ranked, and a rapid assessment of the abundance and identity of positive clones is easily obtained. In addition, rarer clones that would not have been identified by standard screening may be found (D’Angelo et al., 2014) as well as potentially cross-reactive or polyspecific clones (Ferrara et al., 2013). One of the most challenging examples in the in vitro selection field is represented by antibodies. Their simplest recombinant format, the scFv (single-chain fragment variable) (Huston et al., 1988), has been widely used to select targetspecific binders (Bradbury et al., 2011). In the scFv, the variable domains (VH and VL), responsible for antigen-binding activity, are connected by a flexible linker. Ideally, the complete VH and VL genes would be sequenced by NGS. However, the technology is presently limited by attainable read lengths. Consequently, when selection outputs of a scFv library are characterized by NGS, analysis is usually restricted to HCDR3 (Heavy Chain Complementarity Determining Region 3), the most variable of the six hyper-variable loops present in the VH and VL chains. HCDR3 shows wide variations in length, structure, shape and sequence (Morea et al., 1998), as well as intrinsic conformational diversity (James et al., 2003), reflecting the importance of HCDR3s in antibody-binding specificity (Xu and Davis, 2000). Although HCDR3 is the most diverse, variability is also found in the five other complementarity determining region (CDRs), as well as, to a lesser extent, the four framework regions flanking the CDRs (Fig. 1a). Altogether, these features contribute to antibody diversity and consequent antigen-binding activity. While the complexity and depth of an antibody library/selection can be assessed from deep sequencing using appropriate algorithms (AbMining Toolbox (D’Angelo et al., 2014), VDJFasta (Glanville et al., 2009)), the rescue of identified clones is another matter. Deep sequencing provides vast amounts of useful information, but positive clones need to be separately isolated using the obtained sequence information. A number of rescue strategies have been reported for antibodybased constructs, including correlation with randomly picked clone (Ravn et al., 2010), fragment assembly (Ravn et al., 2010) and gene synthesis (Saggy et al., 2012). The approach of random clone picking involves correlating the full (Sanger) S.D’Angelo et al. Fig. 1. Inverse PCR strategy applied to antibodies. (a) The structure of the scFv gene. The variable light (VL) and heavy (VH) chain CDRs are indicated by white boxes. The portion of the VH covered by deep sequencing is identified and magnified: the HCDR3, as identified by the AbMining tool, is boxed in gray and the inverse PCR primers are shown. A black circle on the forward primer indicates phosphorylation. (b) Schematic representation of the specific HCDR3 rescue strategy: the desired HCDR3-specific back-to-back primers are designed on the HCDR3 sequence as obtained from deep sequencing. Plasmid DNA obtained from the selection output is (...truncated)