High-volume hybridoma sequencing on the NeuroMabSeq platform enables efficient generation of recombinant monoclonal antibodies and scFvs for neuroscience research

Abstract

The Neuroscience Monoclonal Antibody Sequencing Initiative (NeuroMabSeq) is a concerted effort to determine and make publicly available hybridoma-derived sequences of monoclonal antibodies (mAbs) valuable to neuroscience research. Over 30 years of research and development efforts including those at the UC Davis/NIH NeuroMab Facility have resulted in the generation of a large collection of mouse mAbs validated for neuroscience research. To enhance dissemination and increase the utility of this valuable resource, we applied a high-throughput DNA sequencing approach to determine immunoglobulin heavy and light chain variable domain sequences from source hybridoma cells. The resultant set of sequences was made publicly available as a searchable DNA sequence database (neuromabseq.ucdavis.edu) for sharing, analysis and use in downstream applications. We enhanced the utility, transparency, and reproducibility of the existing mAb collection by using these sequences to develop recombinant mAbs. This enabled their subsequent engineering into alternate forms with distinct utility, including alternate modes of detection in multiplexed labeling, and as miniaturized single chain variable fragments or scFvs. The NeuroMabSeq website and database and the corresponding recombinant antibody collection together serve as a public DNA sequence repository of mouse mAb heavy and light chain variable domain sequences and as an open resource for enhancing dissemination and utility of this valuable collection of validated mAbs.

Figure 1

Sequencing workflow and bioinformatics processing. Hybridomas of interest are sequenced using a workflow consisting of RNA extraction, cDNA synthesis, and semi-nested PCR amplification with IgG-specific primers followed by the addition of unique Illumina barcodes to each sample. Illumina libraries are then generated, and adapters are ligated for sequencing on the MiSeq platform. Bioinformatics processing is shown on the right panel. Reads from the Illumina sequencing are run through HTStream for base quality trimming and other read processing. Next, they are passed through DADA2 for amplicon denoising followed by SAbPred ANARCI tool based on the IMGT numbering scheme. All ASVs, metadata, and other quality metrics are uploaded to the NeuroMabSeq database and website where further information and tools are provided to the end users. This includes but is not limited to BlastIR results, BLAT searches across the database, and recommended high quality sequences for recombinant antibody design. Annotations of internally generated scores are provided in addition to other database information. Finally, high quality sequences are used in the design of gene fragments for generation of R-mAb and scFv expression plasmids.

Figure 2

Example of a NeuroMabSeq website sequence entry. (a) View of the query interface that the website provides where users can search by mAb ID, target, etc. Shown are sequence entries returned for a search for “L113/13”. This includes the mAb ID, the category of hybridoma, the protein target of the mAb, the number of light and heavy chain sequences returned, the star scores, and whether the sequences are identical to those used in the design of gene fragments that resulted in successful cloning of R-mAbs and scFvs. (b) The entry for the L113/13.5 light chain is shown with separate boxes for “Sequencing Information”, “Scoring Information” and “Amino Acid Information”. The “Sequencing Information” dropdown contains data such as the number of ASVs attributed to the obtained sequence and the number of total reads attributed to light chains or heavy chains for the sample, as well as the plate. The “Scoring Information” reveals the star rating assigned to each sequence, as well as the contribution of the ASV-based star and replicate-based star components of the scoring to the total score. The “Amino Acid Information” dropdown contains information such as the full amino acid sequence, the sequence corresponding to the ANARCI prediction of IMGT amino acid positions for the V_L domain, and within this the FR 1-4 and CDR 1-3 boundaries. The nucleotide sequence corresponding to the ANARCI prediction of IMGT amino acids is also shown to facilitate design of gene fragments for Gibson Assembly-based cloning of recombinant mAbs and scFvs. In addition, the “BLAT Sequence” feature is available to compare this sequence to all other sequences in the database. An analogous set of information is supplied for the heavy chain.

Figure 3

NeuroMabSeq workflow with details of samples sequenced. This Sankey diagram depicts the details of hybridoma samples sequenced to date. The 8642 novel (i.e., non-control hybridoma) samples sequenced included 1903 that did not yield usable sequences and were designated as “dropouts”. 6739 samples returned usable sequences that yielded 15,064 total V_L and V_H sequences. 13,401 of these remained after eliminating any sequences that had insufficient support, did not conform to ANARCI conventions of valid antibody sequences, or were duplicates; these were subjected to the star scoring system. Of these, 11,226 had sequencing quality scores greater than 3, while 2175 samples had sequencing quality scores less than 3. The number of unique mAb IDs after grouping all biological and technical replicates is also provided (1931 high quality and 331 low quality).

Figure 4

ASV Percent Score reported for the V_L and VH sequencing for each of the samples sequenced. Based on these distributions, and assumptions as to unlikely biological contributions of very minor transcripts, a cut off was applied so that only ASVs representing > 10% of total reads (black line) were considered.

Figure 5

Score distribution of “Stars” awarded to each of the V_L and V_H sequences reported. Due to the tendency of V_L sequence entries to report more ASVs we see a tendency for a left skewed distribution for V_L sequence entries compared to V_H sequence entries. The sections of high density shown are due to the scoring system which counts the number of matches of biological and technical replicates for a given sequence.

Figure 6

R-mAb and scFv Gibson Assembly cloning strategy. (a) R-mAb cloning strategy showing a schematic of the four-piece Gibson Assembly based construction of the R-mAb expression plasmid and at the bottom a schematic of the V_L and V_H gene fragments. (b) scFv cloning strategy showing a schematic of the two-piece Gibson Assembly based construction of the scFv expression plasmid and at the bottom a schematic of the V_H -linker- V_L gene fragment.

Figure 7

R-mAb evaluation. (a) Immunofluorescence immunocytochemistry on transiently transfected COS-1 cells. Cells were transfected with a plasmid encoding mouse THIK-2 and double immunolabeled with the progenitor N468/37 IgG1 mAb (green) and the N468/37R IgG2a R-mAb (red). Hoechst nuclear labelling is shown in blue. (b) Immunoperoxidase/DAB immunohistochemistry on rat brain sections comparing immunolabeling with the anti-HCN4 hybridoma-generated N114/10 mAb to the N114/10R R-mAb. (c) Strip immunoblots on rat brain samples showing side-by-side comparisons of hybridoma-generated conventional progenitor mAb samples “C” and transfected cell generated R-mAb samples “R”. Numbers to the left of each set denote mobility of molecular weight standards in kD. (d) Immunofluorescence immunohistochemistry on rat brain sections. Multiplex immunolabeling with four mouse mAbs including the subclass-switched R-mAb L113/130R in adult rat hippocampal CA1.

Figure 8

scFv evaluation. (a) Immunofluorescence immunocytochemistry on transiently transfected COS-1 cells. Cells were transfected with a plasmid encoding the mouse GABAB R1 receptor and double immunolabeled with the progenitor N93A/49 mouse mAb (green) and the N93A/49 scFv (red). Hoechst nuclear labelling is shown in blue. (b) Immunoperoxidase/DAB immunohistochemistry on rat brain sagittal sections. The top pair of images show immunolabelling with the anti-GABA_A receptor α4 hybridoma-generated N398A/34 mAb (top section) and the N398A/34 scFv (bottom section). The bottom pair of images show immunolabelling in neocortex with the anti-Ankyrin-G hybridoma-generated N106/20 mAb (left section) and the N106/20 scFv (right section). (c) Immunofluorescence immunohistochemistry on rat brain sagittal sections. The top set of images show multiplex immunolabeling of brain (top two rows) and hippocampus (bottom row) with the anti-VGlut1 hybridoma-generated N28/9 mAb (IgG1) (left section) and the N28/9 scFv (right section) in green. The sections were also labelled with the mouse mAbs anti-Parvalbumin L114/3 mAb (IgG2a) in red and the anti-Amigo-1 L86A/37 mAb (IgG2b) in purple. Hoechst nuclear labelling is shown in blue. The bottom pair of images show multiplex immunolabeling of rat cerebellum with the anti-Kv1.2K⁺ channel hybridoma-generated K14/16 mAb (IgG2b) (left section) and the K14/16 scFv (right section) in green. The sections were also labelled with the mouse anti-Calbindin L109/39 mAb (IgG2b). Hoechst nuclear labelling is shown in blue.