A multiplatform strategy for the discovery of conventional monoclonal antibodies that inhibit the voltage-gated potassium channel Kv1.3

ABSTRACT

Identifying monoclonal antibodies that block human voltage-gated ion channels (VGICs) is a challenging endeavor exacerbated by difficulties in producing recombinant ion channel proteins in amounts that support drug discovery programs. We have developed a general strategy to address this challenge by combining high-level expression of recombinant VGICs in Tetrahymena thermophila with immunization of phylogenetically diverse species and unique screening tools that allow deep-mining for antibodies that could potentially bind functionally important regions of the protein. Using this approach, we targeted human Kv1.3, a voltage-gated potassium channel widely recognized as a therapeutic target for the treatment of a variety of T-cell mediated autoimmune diseases. Recombinant Kv1.3 was used to generate and recover 69 full-length anti-Kv1.3 mAbs from immunized chickens and llamas, of which 10 were able to inhibit Kv1.3 current. Select antibodies were shown to be potent (IC₅₀<10 nM) and specific for Kv1.3 over related Kv1 family members, hERG and hNav1.5.

Introduction

Despite the importance of voltage-gated ion channels (VGICs) in human health and disease, the development of therapeutic compounds that can modulate channel activity has proven difficult and been largely restricted to small molecule drugs or peptide toxins that often lack selectivity, have limited bioavailability and/or poor pharmacokinetics. Although monoclonal antibodies (mAbs) with their superior therapeutic profiles could potentially address these issues, discovery programs focused on identifying such antibodies present challenges of their own.

First and foremost, robust platforms for the production of recombinant ion channel proteins have been lacking. This has limited the amount of material available for animal immunizations, as well as the development of antibody screening tools, particularly those that allow deep mining of the immunoglobulin repertoire to identify rare mAbs. Secondly, the high degree of sequence conservation amongst ion channel orthologs, together with the comparatively small extracellular loops on many of these proteins, make ion channels particularly difficult targets for therapeutic antibody development.

To address the first challenge, we have employed the free-living ciliate, Tetrahymena thermophila, to overexpress recombinant human VGICs. Genomics studies in Tetrahymena have revealed expanded gene families for each of the principal classes of membrane transporters including P-type ATPases, major facilitator superfamily members, ABC transporters, and VGICs.¹ Indeed, the T. thermophila macronuclear genome sequence predicts more than three times the number of genes for VGICs relative to humans.¹ Presumably, these expanded gene families promote Tetrahymena's survival in a continually changing aquatic environment, and are key to physiological responses associated with mating, prey-predator interaction, and food-finding, many of which rely on membrane hyperpolarization and depolarization events that alter cell motility through their effects on ciliary beat.^2,3 This core technology platform has allowed us to express mammalian VGICs at levels considerably higher than previously possible in standard mammalian expression systems (P.C., personal communication). This has enabled a general strategy for the production of antibodies against human ion channels involving: 1) antigen formulations incorporating concentrated and highly purified channel proteins; 2) immunization of multiple antibody host species that are phylogenetically diverse; and 3) direct screening of B cells or large recombinant antibody libraries with proteins oriented on particle-based solid supports.

As demonstrated here, we successfully applied this strategy towards the discovery of modulating antibodies against the human voltage-gated K⁺ channel, Kv1.3 (hKv1.3), which regulates the membrane potential of both excitable and non-excitable cell types and plays a well-characterized role in regulating Ca²⁺ signaling in activated effector-memory T (T_EM) cells.^4,5 Up-regulation of Kv1.3 in auto-reactive T_EM cells is associated with sites of inflammation in autoimmune disease, and numerous studies have shown that specific Kv1.3 blockers are effective in a number of animal models of inflammation.^5-13 Therefore, Kv1.3 is regarded as an attractive target for therapeutic intervention, particularly for type 1 diabetes, multiple sclerosis, psoriasis, and other autoimmune disorders.^4,5

Here, we show that recombinant Kv1.3 expressed in Tetrahymena traffics to the cell surface, is correctly folded, adopts a native tetrameric conformation and is readily purified from membranes enabling the development of potent immunogens and screening tools. For these initial studies, chickens and llamas were chosen as phylogenetically diverse immune hosts since both have demonstrated success in producing antibodies against highly conserved mammalian proteins.^14,15 Using this approach, we identified 69 anti-Kv1.3 mAbs, 10 of which block Kv1.3 currents, yielding an ∼ 14% hit rate. Select antibodies demonstrate considerable potency (IC₅₀<10 nM) and selectivity (>166 fold) over related channel family members.

Results

Kv1.3 Tetrahymena expression construct design

To maximize expression of human Kv1.3 in Tetrahymena, we introduced a codon-optimized version of the KCNA3 gene driven by a T. thermophila cadmium-inducible promoter and linked to a neo2 drug resistance cassette into a high-copy ribosomal DNA vector designated pTRAS1 (Fig. 1a).¹⁶ Like its parent plasmid, pD5H8, pTRAS1 retains the ability to form highly amplified, 21 Kb palindromic chromosomes during macronuclear development, but contains a wild type rather than paromomycin-resistant version of the 17S ribosomal RNA gene, thus favoring retention of the transgene and its linked drug resistance marker that are typically lost through recombination (D.C-H, personal communication).¹⁷

Figure 1.

Expression of human Kv1.3 in Tetrahymena thermophila. A. Expression construct design. KCNA3, the gene encoding human Kv1.3, was modified with a C-terminal FLAG/10Xhis tag and placed under the control of the MTT5 and MTT1 promoter and terminator, respectively. The entire expression cassette was cloned as a NotI fragment into an rDNA vector, pTRAS1. The relative positions of chromosome breakage sites (CBS) and ribosomal genes (17s, 5.8s, and 26s) are shown. B. Single cell isolates maintain expression of recombinant Kv1.3. Anti-Kv1.3 Western analysis of single cells isolated from pooled Tetrahymena transformants and tested for their ability to express Kv1.3. Eight of nine single cell isolates expressed Kv1.3 at similar levels to the original pool (T1) with one clone (#117) expressing higher-levels of Kv1.3. A lysate from wild-type cells (WT) was included as a negative control. C. Tetrahymena-expressed Kv1.3 is phosphorylated. Purified Kv1.3 was incubated in the absence (−) and presence (+) of calf-intestinal alkaline phosphatase (CIP) and subsequently detected by anti-Kv1.3 Western analysis as described above. D. Comparison of Kv1.3 expression levels in Tetrahymena and CHO cells. Cell lysates generated from 25,000 Kv1.3 expressing Tetrahymena (Tth) or CHO cells were resolved by SDS-PAGE. Kv1.3 was detected by Western analysis using an anti-Kv1.3 antibody and an anti-guinea pig HRP conjugated antibody. E. Tetrahymena-expressed Kv1.3 binds both Agitoxin-2 (AgTX-2) and ShK. Mock-induced wild-type cells (WT) and Kv1.3-expressing Tetrahymena cells were fixed and labeled with either 10 nM Agitoxin-2-TAMRA (AgTX-2-TAMRA) or ShK-TAMRA and visualized by fluorescence confocal microscopy. Inset shows a close-image of a single Tetrahymena cell. White arrows highlight the Tetrahymena plasma (surface) membrane. F. Binding of ShK to Tetrahymena Kv1.3 is specific. Fixed Tetrahymena cells expressing Kv1.3 were incubated with 10 nM ShK-TAMRA in the presence of saturating (10X) amounts of Margatoxin (MgTx) or Iberiotoxin (IbTx) and examined by fluorescence confocal microscopy.

Expression and membrane localization of Kv1.3 in T. thermophila

In human T cells, Kv1.3 exists as homotetramer that can be modified by phosphorylation.¹⁸ Consistent with this, Western analysis shows that Tetrahymena expressed Kv1.3 resolves as two bands, a smaller 65 kDa band, which is in general agreement with the predicted mass of the Kv1.3 monomer (66 kDa), and a larger 85 kDa band that disappears following treatment with alkaline phosphatase (Figs. 1b & c). A standardized comparison of expression levels between Kv1.3 expressing Tetrahymena and CHO cells shows that Tetrahymena cells produce significantly more Kv1.3 on a per cell basis (Fig. 1d).

Evidence that the recombinant channel assembles into a tetramer was based on staining of fixed cells with K⁺ channel peptide toxins. Agitoxin-2 (AgTX-2), Margatoxin (MgTX), and Stichodactyla toxin (ShK) are potent Shaker potassium channel blockers derived from the venom of scorpions (AgTX-2; MgTX), and the sea anemone, Stichodactyla helianthus (ShK).^19-21 These peptide inhibitors bind with high-affinity to the outer vestibule of the tetrameric channel, occlude ion conductance through the pore, and have been used extensively for probing K⁺ channel structure.^22-27 Fluorescence microscopy showed that both AgTX-2- and ShK-TAMRA bind Tetrahymena cells expressing Kv1.3, but not wild-type negative control cells (Fig. 1e). At higher magnification, toxins were found to localize at the cell periphery or perinuclear areas enriched in endoplasmic reticulum (ER), suggesting that Kv1.3 assembles into tetramers in the ER before being transported to the cell surface (Fig. 1e; top right panel). To ensure binding of the labeled toxins was specific for Kv1.3, cells were co-incubated with 10-fold saturating amounts of competing non-labeled MgTX or Iberiotoxin (IbTX). Iberiotoxin is a potent blocker of large conductance Ca²⁺-activated K⁺ channels derived from Burthus tamulus scorpion venom that does not affect voltage-gated potassium channels such as Kv1.3.²⁸ Fig. 1f shows that MgTX, but not IbTX, competes with both AgTX-2- and ShK-TAMRA binding in Kv1.3 expressing cells. These results confirm that Kv1.3 traffics to the Tetrahymena cell surface as a correctly assembled tetramer.

Kv1.3 purification, reconstitution, and characterization

Kv1.3 containing a C-terminal His tag was extracted from Tetrahymena membranes with the detergent FC12 after pre-treatment with digitonin and purified by nickel affinity chromatography. Psora-4, a specific small molecule Kv1.3 inhibitor was maintained in extraction and purification buffers to stabilize the channel and generally resulted in increased yields. The majority of the purified protein resolved as a doublet of ∼65 kDa and ∼85 kDa on SDS-PAGE, with additional Coomassie stainable material being present at higher (>250 kDa) molecular mass (Fig. 2a). Both the doublet and higher mass species were confirmed to consist of Kv1.3 by anti-Kv1.3 Western analysis and mass spectroscopy analysis (data not shown). Multiple purifications (n = 7) resulted in a range of recovery of purified protein between 0.5–3 mg/L and an average recovery of 1.3 mg Kv1.3 per liter of culture. Purified Kv1.3 was either reconstituted into phosphatidylcholine-based proteoliposomes for long-term storage (Fig. 2a) or adhered to magnetic beads via the C-terminal His tag as described in Materials and Methods (Fig. 2c & d). Kv1.3-containing magnetic beads preferentially recovered antibodies that recognized extracellular epitopes (Fig. 2e), confirming that this formulation incorporated right-side out ion channel.

Figure 2.

Purification of recombinant Kv1.3. a. SDS-PAGE analysis of purified and reconstituted Kv1.3. Kv1.3 was purified as described in Materials, resolved by SDS-PAGE before and after reconstitution into liposomes and stained with SimplyBlue™ SafeStain™. B. Ligand Binding Analysis. Kv1.3-containing liposomes were incubated with FAM-ShK (3 nM) in the presence or absence of a 50-fold excess of either MgTx or IbTx. Top Panel is a representative experiment showing total binding expressed as Anisotropy measured by fluorescence polarization. Bottom Panel represents specific binding to FAM-ShK. Kd was estimated as 11.5 nM +/− 3.4 nM based on specific binding curves generated in three separate experiments c. Fluorescence microscopy analysis of Kv1.3 magnetic beads. Magnetic beads containing tethered Kv1.3 reconstituted into a lipid bilayer consisting of rhodamine-labeled PE and non-labeled PC were examined by fluorescence microscopy. Beads were examined under a rhodamine filter to detect labeled PE incorporation (Left Panel) and with a FITC-filter following labelling with an anti-Kv1.3 antibody that recognizes an epitope on the first extracellular loop and anti-guinea pig conjugated FITC (Middle Panel). Rhodamine and FITC- images were merged (Right Panel) to confirm co-localization of Kv1.3 and the lipid bilayer (yellow fluorescence). D. Schematic illustration of Kv1.3 magnetic beads. Shown is the magnetic bead surface; the lipid bilayer consisting of PC (yellow lipids) and Rhodamine-labeled PE (red lipids); the six transmembrane domains of the Kv1.3 monomer (S1-S6); the C-terminal engineered FLAG (orange Triangle) and 10Xhis (Green Box) tags; a star indicates the position of the epitope on the first extracellular loop that is recognized by the Kv1.3 antibody utilized in c and e. e. Kv1.3 magnetic beads preferentially precipitate an antibody that recognizes an extracellular epitope. Kv1.3 magnetic beads or control beads were incubated with antibodies (6.7 nM) that recognize either internal (FLAG and 10Xhis) or external (Kv1.3) epitopes. Beads were washed and bound IgG eluted directly in SDS-PAGE loading buffer. IgG was detected by Western analysis using either anti-mouse-HRP (anti-FLAG and -His) or anti-guinea pig-HRP (anti-Kv1.3).

Three separate batches of Kv1.3 proteoliposomes were analyzed in independent experiments to estimate binding affinity constants using fluorescence polarization.^29-31 Fluorescein 6-carbonyl (FAM)-labeled ShK (FAM-ShK, 3 nM) was added to increasing amounts of Kv1.3-containing proteoliposomes in the absence or presence of a 50-100-fold excess of competing MgTx or IbTx. Fig. 2b is a representative experiment showing that total FAM-ShK binding increased with increasing amounts of protein and that MgTx, but not IbTx, competed with ShK binding (top panel). Specific binding curves (Fig. 2b, bottom panel) were used to determine an average binding affinity (K_d) of 11.5 nM +/− 3.4 nM (n = 3), ∼ 200-fold lower than the IC₅₀ (52 pM) reported for FAM-ShK using patch clamp electrophysiology.²⁷

Generation of anti-Kv1.3 antibodies

Two chickens were immunized with Kv1.3 proteoliposomes (with DNA/LPS adjuvant) as described in Materials and Methods. One chicken (19724) demonstrated a specific titer at week 6 and was chosen for further study. Anti-Kv1.3 antibodies were identified in gel encapsulated microenvironment (GEM) assays using splenocytes recovered from chicken 19724 and antigen reporters (that is, screening tools) composed of either polystyrene beads coated with Kv1.3 proteoliposomes; Kv1.3 magnetic beads containing right side out, oriented Kv1.3 protein (Fig. 2c, d & e); or, Kv1.3-expressing Tetrahymena cells. For each reporter strategy, GEMS were prepared from 10 million splenocytes and one 96-well plate of GEMS was picked, yielding 50 sequence-unique single-chain variable fragments (scFv)-Fc clones in total (see Table 1 for details). Each antibody bound Kv1.3 by ELISA, but clone ch_p1E6 demonstrated significantly less activity compared to other clones and ch_p1F8 also bound the non-related ion channel (NRIC) proteoliposomes control (the binding of a subset of these antibodies is shown in Fig. 3a). MAbs recovered from GEM screens incorporating Kv1.3 proteoliposomes or magnetic beads were additionally tested for their ability to bind Jurkat cells via flow cytometry. Jurkat cells are a human leukemia T-cell line that has been reported to express ∼ 200 active Kv1.3 ion channels per cell.³² Of 41 antibodies tested, Jurkat cell binding was detected with 6 antibodies (ch_p1A11, ch_p1H7, ch_p1D8, ch_p1E6, ch_p2A3 and ch_p2G9) comprising a range of median fluorescence intensity of 1157 to 9401 compared to an average MFI of 198 +/− 41 (n = 35) for all other antibodies analyzed.

Table 1.

Summary of anti-Kv1.3 antibody identification.

		ELISA positive	Functional
Antibody Platform	Screening Method	(unique sequences)	Antibodies Identified
Chicken	GEM Assay: Kv1.3 Proteoliposomes	24	7
	GEM Assay: Kv1.3 Magnetic Beads	16	2
	GEM Assay: Kv1.3 Whole Cells	10	0
Llama	Phage Display: Kv1.3 Magnetic Beads	19	1
	Total	69	10

Figure 3.

Identification of anti-Kv1.3 antibodies. A. Chicken anti-Kv1.3 antibodies. ScFv-Fc screening was carried out by ELISA using three-fold serial dilutions of antibody against proteoliposomes containing Kv1.3 or an irrelevant human VGIC also expressed in Tetrahymena. An isotype control (IC) for generated antibodies was also included. Shown are results for antibodies that inhibit Kv1.3 activity. Note the relative lack of reactivity of clone ch_p1E6 against Kv1.3 and clone p1F8 reactivity against the irrelevant proteoliposome control. B. Llama anti-Kv1.3 antibodies. ScFv-Fc screening was carried out using ten-fold serial dilutions of antibody on mesoscale against proteoliposomes containing Kv1.3 or an irrelevant human VGIC also expressed in Tetrahymena. IC1, isotype control for generated antibodies; IC2, isotype control for anti-Kv1.3 polyclonal antibody. Data is shown for 6 of 19 specific Kv1.3 binding scFv-Fc antibodies.

To generate antibodies in llamas, four animals were immunized and boosted with plasmid DNA encoding Kv1.3 six times, and additionally boosted with Kv1.3 proteoliposomes twice. Kv1.3-specific immune responses were measured for Kv1.3 proteoliposomes versus NRIC proteoliposomes. Three of four llamas showed an immune response against Kv1.3 proteoliposomes, but also to irrelevant proteoliposomes. Nevertheless, scFv libraries were constructed for two llamas (Armello and Lisandro) that exhibited the strongest responses against Kv1.3 proteoliposomes. Two phage libraries were created for each llama, comprising combinations of the immunoglobulin heavy chain repertoire and either the κ light chain, or λ light chain repertoires, resulting in library sizes of ∼ 3–8 × 10⁸ clones. Phage panning was performed on Kv1.3 magnetic beads. After two rounds of phage selections, an enrichment of 10-fold for library Armello κ, 18-fold for Armello λ, 43-fold for Lisandro κ and 100-fold for Lisandro λ over NRIC magnetic beads was achieved. Individual soluble scFv in periplasmic fractions were screened on mesoscale for specific binding to Kv1.3 proteoliposomes. Only a limited screening of 36 clones per library was performed, nevertheless 48 of the 144 screened clones displayed specific binding. Sequencing of the VH and VL regions of these specific clones showed that they could be divided into 19 different VH families based on VH complementarity-determining region (CDR)3 sequence length and similarity. The clones derived from the Armello κ scFv library consisted of 5 VH families, from Armello λ of 3 VH families, from Lisandro κ of 9 VH families and from Lisandro λ of 3 VH families. Nineteen clones representing the best performing clone on mesoscale of each different family were selected to be re-cloned as scFv-Fc fusion proteins. After transfection, production in HEK293E cells, and protein A purification, a new specificity screen on mesoscale was initiated. For all scFv-Fc molecules representing these 19 different VH families, specific binding to Kv1.3 could be reconfirmed (the binding of a subset of these antibodies is shown in Fig. 3b).

Identification of functionally inhibiting anti-Kv1.3 antibodies

Purified scFv-Fc anti-Kv1.3 antibody clones were tested by whole-cell patch-clamp at a single concentration (400 nM) for inhibition of human Kv1.3 channels transiently expressed in L929 mouse fibroblast cells. Nine chicken mAbs and one llama mAb were identified as inhibiting Kv1.3 activity (Fig. 4a & b). Each functional antibody was derived from antibody screens incorporating either Kv1.3 proteoliposomes and/or magnetic beads (Table 1) and blocked between 44–82% of Kv1.3 currents (Table 2) within 600 seconds of application (Fig. 4d). None of the antibodies tested, chicken or llama, under the assay conditions used, displayed agonist activity towards the Kv1.3 channel. Interestingly, while not all of the functional chicken antibodies bound Jurkat cells as detected by flow cytometry, each of those that did were functionally active.

Figure 4.

Identification of anti-Kv1.3 antibodies that functionally inhibit Kv1.3 channel activity. Purified scFv-Fc anti-Kv1.3 antibodies from either chickens or llamas were tested at a concentration of 400nM via electrophysiology for their ability to block current from human Kv1.3 expressed in L929 fibroblast cells. Shown are representative traces for each of the antibodies that blocked Kv1.3 current. Black lines represent control currents, red lines represent currents following addition of antibody. Inhibiting anti-Kv1.3 antibody clones derived from chickens are shown in a, the functional llama anti-Kv1.3 antibody is shown in b. An example of an antibody that was tested and shown not to modulate Kv1.3 activity is shown in c. d. Time-dependent development of current inhibition by monoclonal antibodies targeting Kv1.3. (Left Panel) Time-current plots showing current inhibition of three (3) individual cells expressing hKv1.3 channels by the monoclonal antibody ScFv-Fc L1A3. Antibodies were added after current stabilization at 0 second. Currents were elicited by pulsing to +40 mV for 200 ms from a holding potential of −80 mV every 30 seconds. (Right Panel) Means ± SD plot of the current inhibition of three individual cells in the left panel. Similar time-dependent profiles were observed for each of the blocking antibodies.

Table 2.

Summary of anti-Kv1.3 antibody functional activity.

	Percent Kv1.3 current blocked at 400 nM
scFv-Fc clone	Average block (%)	STDEV (%)	n
ch_p1A1	50.28	13.02	8
ch_p1A11	56.51	10.1	6
ch_p1F8	44.34	12.12	3
ch_p1D8	54.47	1.52	3
ch_p1H4	45.99	12.38	6
ch_p2A3	45.02	3.5	2
ch_p2G9	75.35	11.65	11
ch_p1H7	47.32	12.42	2
ch_p1E6	82.28	10.45	12
L1A3	73.53	8.71	3

Based on the relative inhibition of current at 400 nM, the most potent chicken antibody (ch_p1E6) and the llama L1A3 antibody were characterized further. Dose response analysis determined IC₅₀ concentrations for ch_p1E6 and L1A3 as 6 and 109 nM, respectively (Fig. 5a & b). Selectivity for each of the antibodies was assessed on related Kv1.x family members (Kv1.1, Kv1.2, Kv1.5), hERG (Kv11.1) and Nav1.5. At 1 μM, neither antibody inhibited activity more than 5% for any of the channels tested (Fig. 5c & Table 3), indicating that the fold selectivity for Kv1.3 is, at minimum, greater than 166-fold for ch_p1E6 and 9-fold for L1A3.

Figure 5.

Anti-Kv1.3 antibody potency analysis. A. Dose response curves. Ten-fold serial dilutions of chicken antibody p1E6 (red traces) and llama antibody L1A3 (green traces) were used to generate dose-response curves. Black lines represent control currents. B. IC₅₀ Determinations. IC₅₀s for p1E6 (red curve) and L1A3 (green curve) were determined by fitting the calculated percentage of current block to a Hill equation. IC₅₀s for p1E6 and L1A3 were estimated as 6 and 109 nM, respectively c. Analysis of p1E6 and L1A3 selectivity. ScFv-Fc clones p1E6 and L1A3 were tested for their ability to block the activity of related Kv1.x family members (Kv1.1, Kv1.2, Kv1.5), hERG and Nav1.5 at a concentration of 1 μM. Shown are representative traces from each experiment. Black lines, control current; Red lines, currents following addition of antibody.

Table 3.

L1A3 and p1E6 selectivity profile.

	L1A3		p1E6
	% Blocked at 1 µM		% Blocked at 1 µM
Channels	(Mean ± SD)	n	(Mean ± SD)	n
Kv1.1	4.18 ± 2.65	2	3.87 ± 2.65	2
Kv1.2	0.68 ± 0.97	2	0.00 ± 0.00	2
Kv1.5	0.48 ± 0.69	2	1.28 ± 1.80	2
Kv11.1 (hERG)	2.11 ± 2.98	2	5.32 ± 0.45	2
Nav1.5	4.98 ± 3.09	3	0.00 ± 0.00	2

Epitope binning analysis

Owing to the large recovery of functional antibodies, particularly from chickens, we performed an epitope binning analysis to gain an understanding of the diversity of epitopes recognized by the anti-Kv1.3 antibodies and to determine if there was a correlation between functional activity and epitope bins. Thirty-six chicken mAbs, including eight of the nine functional antibodies (clone ch_p1E6 was excluded-see below) and the llama L1A3 antibody, were analyzed by array-based surface plasmon resonance (SPR) imaging. A heat map of the of the binning analysis indicates that antibodies can be divided into at least 5 epitope bins based on their ability to block competing antibody binding to the Kv1.3 antigen (Fig. 6a). A network plot (Fig. 6b) highlights the fact that a majority (56.7%) of the antibodies fall into one bin (Bin 1) that includes 7 of the 9 functional antibodies tested. Bin 2 contains 23% of the antibodies tested and includes 1 functional antibody (ch_p1F8), whereas Bins 3 and 4 contain 13.5 and 5.4% of antibodies tested, respectively, and do not contain any functional antibodies. Binding to Kv1.3 by clone ch_p1D8 was not blocked by any other antibodies tested, and therefore was assigned its own epitope bin (Bin 5). No obvious correlation was noted between epitope bin assignment and antibodies that, in addition to functionally inhibiting Kv1.3 activity, also bound Jurkat cells.

Figure 6.

Kv1.3 epitope binning analysis. A. Binning analysis heat map. Shown are antibodies colored by bin (1–5) and by functional data. Antibodies that block Kv1.3 current are highlighted in blue (Ephys only), those that additionally bind Jurkat cells are highlighted in red (Ephys + Jurkat) and those that do neither are highlighted in grey. Relative competition activity of each antibody is color coded and indicates strong competition (red boxes), intermediate/weak competition (yellow boxes) or no competition (green boxes). Dark red boxes indicate competition from the same antibody pair. B. Binning network plot. Antibodies are colored to identify those that inhibit channel activity (blue), additionally bind Jurkat cells (red) or do neither (grey). Note antibody L1A3 was not tested for its ability to bind Jurkat cells, however, for simplicity was denoted as inhibiting ion channel only (blue) in a & b.

Clone ch_p1E6 binding to Kv1.3 in the SPR format was insufficient to enable analysis, a result most likely related to the poor binding of this antibody to purified antigen in ELISAs. However, its potency in blocking mammalian expressed Kv1.3 and ability to bind Jurkat cells suggests that ch_p1E6 effectively binds native Kv1.3 in mammalian cells at an epitope that is partially lost during purification of the channel or its preparation for ELISA, and therefore represents a separate bin (i.e., Bin 6).

Discussion

A number of groups have successfully generated anti-Kv1.3 immunoglobulins that inhibit channel activity, including polyclonal antibodies generated in rabbits, heavy chain only VHH nanobodies in llamas (US Patent Application 20170137512), and engineered cowbodies that incorporate inhibitory Kv1.3 toxins into the ultralong CDR that are characteristic of these antibodies.^33,34 However, to our knowledge, this study describes the first set of conventional mAbs capable of blocking the activity of human Kv1.3. To achieve this, we applied a general strategy designed to mitigate several of the challenges associated with the production of functional antibodies against human VGICs. Importantly, we used Tetrahymena thermophila as an expression platform to generate sufficient quantities of recombinant Kv1.3 to support the development of different types of immunogens and antibody screening tools. While previous studies have suggested the utility of this system for the production of recombinant membrane proteins in general, our study clearly underscores this for voltage-gated ion channels in particular.¹ While the choice of a K⁺-selective channel for these initial studies was based partially on the fact that the vast majority (308 of 332) of annotated VGIC genes in Tetrahymena predict K⁺-selective channels, recent studies have shown this system works equally well for the production of a wide range of other human channels (P.C., personal communication).¹

Localization studies with fluorescently tagged peptide toxins suggest that recombinant Kv1.3 produced in these cells can assemble into homotetramers and traffic to the membrane surface (Fig. 1). However, Kv1.3 was not efficiently extracted from membranes with detergents (e.g., FC12 or DDM) without first pre-treating the membranes with digitonin, where its effect was presumably dependent on its ability to sequester sterols in the lipid membrane. Under these conditions FC12 was found to be the most efficient detergent tested for ion channel extraction. The affinity of Kv1.3 purified under these conditions was ∼ 200-fold lower for ShK than that reported for Kv1.3 expressed in mammalian cells. The reduced affinity of the purified channel is not entirely surprising and may be attributable to a number of factors, including subtle changes in protein folding, which may, in turn, be due to altered post-translational modifications or the artificial lipid environment of the purified channel. Nevertheless, the immunogen and antibody screening formulations described here were sufficient to generate and identify functionally inhibiting anti-Kv1.3 antibodies, thus confirming the utility of Tetrahymena for the expression of native or near-native mammalian channel proteins.

Both chickens and llamas were used as phylogenetically distinct species for immunization. In all, 69 anti-Kv1.3 antibodies were identified (50 from chickens and 19 from llamas). Of those, 10 were capable of blocking Kv1.3 current, an ∼ 14% discovery success rate for functional antibodies. Further analysis of a select chicken antibody, ch_p1E6, and the functional L1A3 llama antibody, revealed potencies of 6 nM and 109 nM, respectively, while neither showed any appreciable activity on related K⁺ or Nav1.5 channels at a concentration of 1 μM, thus confirming that conventional mAbs can selectively and potently inhibit Kv1.3 function with IC₅₀s in the single-digit nM range. The observed functional selectivity of the antibodies is assumed to be linked to their binding to unique epitopes on Kv1.3 relative to other Kv family members based on specific amino acid sequence determinants, particularly in the extracellular domains of these proteins.

A majority (90%) of the antibodies that blocked Kv1.3 were derived from chickens, while a single functional antibody was isolated from phage libraries constructed from immunized llamas. One reason for the higher rate of recovery of functional chicken antibodies may be due to the overall sequence divergence between Kv1.3 in humans, chickens and llamas. Human Kv1.3 shares 97.5% and 97.7% identity with mouse and llama Kv1.3, respectively, but only 83% identity with chicken Kv1.3. Moreover, analysis of the three Kv1.3 extracellular cellular domains, where antibodies would be presumed to exert their effect, reveals further divergence with the human sequence, with the first extracellular loop showing 85.7%, 80.95%, and 52.5% identity with mouse, llama and chicken sequences, respectively (summarized in Table 4).

Table 4.

Kv1.3 ortholog sequence alignment analysis.

	Percent Identity (CLUTSAL Omega)
Query: H. sapiens	M. Musculus (mouse)	G. gallus (chicken)	V. pacos (llama)
P22001.3	NP_032444.2	AAP94028.1	XP_006197566
full-length	97.51	83.15	97.72
S1-S2 ECD	85.71	52.5	80.95
S3-S4 ECD	100	100	100
S5-S6 ECD/Pore	94.74	94.74	100

ECD, extra cellular domain.

A further, and perhaps critical consideration in our overall strategy was to avoid hybridoma-based antibody generation and screening methods to mitigate the effects of non-efficient fusion events and loss of rare B cell clones leading to a decrease in diversity of the screenable immune repertoire.³⁵ In this study, we identified functional antibodies in screens of B cells (chicken mAbs) or from phage display libraries (llama mAbs) using either Kv1.3 proteoliposomes applied to a solid support (polystyrene beads) or from formulations containing Kv1.3 oriented on magnetic beads directly via an engineered C-terminal tag. Seven of nine of the functional chicken mAbs were isolated using Kv1.3 proteoliposomes adhered to polystyrene beads, suggesting that this formulation was particularly efficient for recovering functional mAbs in the GEM assay. While it is difficult to draw definitive conclusions as to a preferred screening formulation based on one experiment, it is unlikely that Kv1.3 displayed in either of the screening formulations (polystyrene beads, magnetic beads or whole cells) was responsible for the efficiency of antibody discovery since functional antibodies isolated from polystyrene-based formulations were capable of recognizing Kv1.3 in both whole cells and magnetic beads (see Supplementary Fig. 1). Interestingly, llama phage display libraries were panned with Kv1.3 magnetic beads alone; therefore, the role of phylogenetic diversity notwithstanding, altering the screening methodology to incorporate Kv1.3 proteoliposomes could potentially increase the recovery of functional antibodies from these phage libraries. Furthermore, given the fact that chicken mAbs capable of recognizing the native channel expressed in Jurkat cells also blocked channel activity, a direct screen or potentially another round of phage selection on these cells could have delivered a higher frequency of functional llama antibodies.

A binning analysis of the chicken anti-Kv1.3 mAbs identified 5 epitope bins with a majority of the antibodies, including 6 of the 8 functional antibodies tested, segregating into one bin. Furthermore, the llama L1A3 mAb, used as a reference in the analysis, segregated to the same bin suggesting a single Kv1.3 epitope was responsible for generating a majority of the functional antibodies in both platforms. Additionally, ELISA analysis showed that select antibodies from epitope bin 1 do not recognize linear peptides comprising each of the three Kv1.3 extracellular domains, suggesting that epitopes are complex and non-linear if not fully conformational (see Supplementary Fig. 2).

Progress is being made in the clinic targeting Kv1.3 with the highly potent venom peptide analog ShK-186 for the treatment of plaque psoriasis.³⁶ In subsequent studies, it will be of considerable interest to determine if one or more of the inhibitory mAbs described here exhibit similar efficacy and additionally display the benefits of long immunoglobulin half-life in a clinical setting. We hypothesize that a stable and potent anti-Kv1.3 mAb may offer significant advantages in enabling a therapeutic intervention that halts the autoimmune cascade in a number of serious human diseases, including type 1 diabetes and multiple sclerosis.

Materials and methods

Cells and reagents

T. thermophila strains B2086 (mat1-2/mat1-2 [mat1-2; II) and CU428 (mpr1-1/mpr1-1 [MPR1; mp-s, VII]) were obtained from the Tetrahymena stock center (Cornell University). Restriction endonucleases were from New England Biolabs. Anti-FLAG antibody was from Thermo Fisher Scientific (#MA1-91878), anti-C-terminal His antibody was from Life Technologies (#R93025), anti-Kv1.3 antibody was from Alomone (#AGP-005). Anti-mouse horseradish peroxidase (HRP)-conjugated antibody was from Bio-Rad Laboratories (#1706516) and anti-guinea pig HRP-conjugated antibody was from Life Technologies (#A18775). Kv1.3 toxins, Margatoxin (MgTX, #STM-325), Agitoxin-2-TAMRA (AgTX-2-TAMRA, #RTA-420-T) and Iberiotoxin (IbTX, #STI-400) were from Alomone. ShK-TAMRA was from Smartox (#SAT001-00100) and FAM-ShK was obtained from Bachem (#H-6088). T. thermophila growth media reagents, protease peptone (#9000270), yeast extract (#9000726), glucose (#VWRBK876) were from VWR. FeCl₃ was from ThermoFisher Scientific (#ICN15349980). Psora-4 was from Sigma (#P9872).

Expression construct design and cloning

A synthetic gene encoding the human KCNA3 gene (accession number P22001.3) modified with an in-frame C-terminal dual FLAG-10XHis (DYKDDDDKHHHHHHHHHH) affinity tag was codon-optimized for expression in Tetrahymena and synthesized by GenScript. The synthetic gene was cloned into a somatic expression vector, pTIEV5, via PacI and SacI restriction sites engineered into the 5′ and 3′ ends of the synthetic gene, respectively. Transcription of the transgene was under the control of the inducible Tetrahymena metallothionein-5 (MTT5) gene promoter. The expression construct comprising MTT5 promoter, Kv1.3 gene, MTT1 terminator and a neomycin resistance cassette was transferred as a single NotI fragment into a high-copy rDNA vector, pTRAS1.

Generation of Kv1.3 expression strains

The Kv1.3 pTRAS expression vector was introduced into conjugating B2086 and CU428 T. thermophila strains by biolistic transformation as described previously.³⁷ Briefly, 20 μl of M17 tungsten beads (Bio-Rad Laboratories) were coated with 8 μg of expression vector DNA. Transformations were carried out using a Biolistic PDS-1000/He Particle Delivery System (Bio-Rad Laboratories) ∼ 10 hours following induction of conjugation. Following particle bombardment and recovery in NEFF medium (0.25% proteose peptone, 0.25% yeast extract, 0.55% glucose, 33 μM FeCl₃) at 30°C for 18 h, cells were supplemented with paramomycin (100 μg/ml), aliquoted into 96-well microplates and incubated for 3–4 days at 30°C. Wells containing paromomycin-resistant cells were identified and transferred to stock tubes. To identify Kv1.3 expressing cell lines, drug-resistant clones were grown in 3 ml cultures to a cell density of at least 5 × 10⁵ cells/ml and Kv1.3 expression was induced by the addition of 2 μg/ml CdCl₂. Cells were harvested at 6 h post-induction and stored at −20°C. Cells expressing Kv1.3 were identified by Western blot analysis using anti-Kv1.3 and/or anti-tag antibodies. To establish clonal cell lines, single cells were isolated from Kv1.3-expressing stock tubes into drops of NEFF media containing 100 μg/ml paromomycin. Single cell clones were transferred to stock tubes and the best expressing cell lines were identified by Western blot analysis.

Kv1.3 localization

To determine cellular localization, T. thermophila wild-type (CU428) and Kv1.3 expressing cells were fixed with 4% paraformaldehyde for 30 min. Tetrahymena cells fixed under these conditions become permeable and allow passage of reagents through the plasma membrane. Cells were incubated in the presence of either 10 nM ShK-TAMRA or 10 nM Agitoxin-2-TAMRA for 30 min. Cells were washed three times in phosphate-buffered saline (PBS) buffer and visualized by fluorescence confocal microscopy using an SP5 Leica laser scanning confocal microscope with a 63X oil objective lens. To investigate specific-binding of Kv1.3 toxins, cells were co-incubated with 10 nM ShK-TAMRA and saturating amounts (100 nM) of either Margatoxin (ShK competitor) or Iberiotoxin (non-competing toxin).

Kv1.3 Purification

Cells expressing Kv1.3 were resuspended in ice-cold 1x PBS buffer and disrupted by a single pass through a M-110P Microfluidizer (Microfluidics) at 4,000 Psi. Disrupted cell membranes were collected by centrifugation at 100,000 × g for 1h at 4°C, resuspended in ice-cold PBS and stored at −80°C. Crude membrane pellets were resuspended in TKG buffer (30 mM Tris HCl pH 7.4, 300 mM KCl, 10% glycerol) containing 1% n-dodecyl-b-D-maltoside (DDM, Anatrace, #D310), 0.2 μg/ml Psora-4, 0.25% phosphatidylcholine (Sigma-Aldrich) and protease inhibitors (Roche #11836170001, cOmplete™ protease inhibitors, EDTA-free, 1 mini-tablet per 10 ml buffer) and incubated 1 h at 4°C to allow extraction of Tetrahymena membrane proteins (Kv1.3 is not extracted under these conditions). Pre-extracted membranes were harvested by centrifugation at 100,000 × g for 1h at 4°C and then resuspended in TKG buffer containing 0.02% digitonin, 0.2 μg/ml Psora-4, protease inhibitors and incubated for 1h at 4°C. An equal volume of extraction buffer (TKG containing 2% FOS-choline 12, 10 mM EDTA, 0.2 μg/ml Psora-4, 0.5% phosphatidylcholine) was added dropwise and the suspension was incubated for 1h at 4°C. The suspension was centrifuged (100,000 × g) for 1h at 4°C and a cleared lysate collected and supplemented with 6 mM MgCl₂ and 50 mM imidazole. Kv1.3 was purified using Ni Sepharose 6 Fast Flow (GE Healthcare) resin. Bound protein was washed first in TKG buffer containing 0.1% DDM, 0.2 μg/ml Psora-4, 0.025% phosphatidylcholine and 50 mM imidazole and then in TKG buffer containing 0.025% DDM, 0.2 μg/ml Psora-4, 0.00625% phosphatidylcholine and 200 mM imidazole. Kv1.3 was eluted in TKG buffer containing 0.025% DDM, 0.2 μg/ml Psora-4, 0.00625% phosphatidylcholine and 750 mM imidazole and dialysed in the same buffer lacking imidazole. Protein was concentrated in an Amicon stirred cell concentrator fitted with an Ultracel® 100 kDa ultrafiltration disk (Millipore, # PLHK02510) under a stream of nitrogen.

Reconstitution of purified Kv1.3 into liposomes

L-α-phosphatidylcholine (Anatrace, #131601P) was dried under nitrogen and resuspended in buffer containing 20 mM Tris-Cl, 100 mM KCl, pH7.4 to a final concentration of 40 mg/ml. Lipids were rehydrated for 2 h at room temperature with 30 second bursts in a sonicating water bath every 10–15 min. Lipids were further sonicated for 5 min, frozen in liquid N₂ and thawed in a 30°C water bath. This process was repeated 2 times. Liposomes were clarified with an Avanti mini-extruder fitted with a 0.1 micron filter. Purified protein (0.5-1.0 mg/ml) was added to liposomes at a ratio of 3:1 and incubated for 10 min at room temperature followed by sonication in a water bath for 10 min. The protein-liposome mix was then added to 150 mg hydrated Bio-Beads SM2 (Bio-Rad Laboratories, #152-8920) per ml of liposome mix and incubated overnight at 4°C. Proteoliposomes were separated from Bio-Beads using a pulled Pasteur pipette, sonicated in a water bath for 5 min, snap frozen in liquid N₂ and stored at −80°C.

Fluorescence polarization ligand binding analysis

Two-fold serial dilutions of Kv1.3 proteoliposomes were mixed with 3 nM FAM-ShK in duplicate in 20 mM Tris pH 7.4, 2.5 mM KCl, 0.1% bovine serum albumin (BSA), in the presence or absence of a 50-100-fold excess of MgTx (ShK competitor) or IbTx (irrelevant ligand), for 1 h at room temperature. Assays were carried out in 150 μl volumes per well in flat-bottom, non-binding surface 96-well black plates (Corning, #3991). Fluorescence anisotropy readings were taken using a Tecan Infinite M1000 Pro at an excitation wavelength of 470 nm and an emission wavelength of 535 nm. The specific binding (in the presence of MgTx) was determined and used to calculate the K_d by nonlinear regression using GraphPad Prism (GraphPad Software, La Jolla, CA) to the equation (which assumes protein depletion):

$$A=A_f+\left(A_b-A_f\right)\times \frac{\left(L_T+K_d+R_T\right)-\sqrt{{\left(-L_T-K_d-R_T\right)}^2-4L_T{R}_T}}{2L_T}$$

where

A =

anisotropy (measured)

R_T =

total protein concentration (known)

L_T =

fluorescent ligand concentration (3 nM)

A_f =

anisotropy of free ligand (calculated)

A_b =

anisotropy of bound ligand (calculated)

K_d =

dissociation constant (calculated)

Kv1.3 magnetic beads

Purified Kv1.3 was incubated with a slurry of magnetic His tag Dynabeads® (Thermo Fisher Scientific, #10103D) pre-equilibrated in buffer A (30 mM Tris-Cl, 300 mM KCl, 10% glycerol, 0.025% DDM, 0.00625% phosphatidylcholine, pH 7.4) for 2 h at 4°C. The Kv1.3 to bead ratio was ∼ 1 mg protein to 1 ml slurry. Beads were washed with buffer A, mixed with rehydrated lipids (10 mg/ml egg L-α-phosphatidylcholine [Anatrace, #131601P], +/− 0.05 mg/ml rhodamine-conjugated 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine [Anatrace, 810150P]) and incubated for 15 minutes at 4°C. The magnetic bead lipid mixture was incubated in a sonicating water bath for 10 minutes, then transferred to a new glass vial containing 200 mg hydrated Bio-Beads and rotated overnight at 4°C. Magnetic beads were separated from Bio-Beads with a pulled Pasteur pipette, washed and resuspended in buffer B (20 mM Tris-Cl, 100 mM KCl, pH 7.4).

Immunizations and antibody screens

Chicken

Immunization. Two female 6-wk old White Leghorn chickens were immunized intramuscularly initially with 112 μg purified recombinant Kv1.3 reconstituted into proteoliposomes. Boosts continued every other week using 85 μg, 112 μg, 280 μg and 390 μg of protein, with a final intravenous (IV) boost of 280 μg. Fifteen μg bacterial plasmid DNA and 10 μl of E. coli lipopolysaccharide (LPS, Alexis #581-012-L002) were added as adjuvant to the liposomes in a final volume of 0.5 ml for the 2^nd, 3^rd and 4^th boosts, and Lipid A was also added at the 4^th boost. The immunization protocol was approved for Crystal Bioscience by the Institutional Animal Care and Use Committee. During the course of the immunizations, chickens were monitored for signs of discomfort and illness; neither bird showed any such signs. Titers were determined by ELISA as follows: 96-well plates were coated overnight at 4°C with 2 μg/ml Kv1.3 or proteoliposomes containing a NRIC, washed 3 times with PBS, and blocked with 3% milk in PBS (PBSM) for 1 h. Dilutions of the sera were added in PBSM and incubated for 1 h, then the plate was washed 3 times with PBS, and anti-chicken IgY-HRP (Sigma, #A9046) was added in PBSM for 1 h. The plate was washed 3 times with PBSM, developed using TMB (Thermo Fisher Scientific, #2023) for 10 min, quenched with 1 N HCl, and read at 450 nm. Final titer reached 1:250 for bird 19718, and greater than 1:30,000 for bird 19724.

Screening splenic B cells for Kv1.3 reactive antibodies. Splenocytes were recovered from bird 19724 four days after the final antigen IV boost and incorporated into Gel Encapsulated Microenvironments (GEMs) according to Izquierdo et al.³⁸ GEMs were prepared using three strategies.

• 1.Polystyrene bead reporters: 5 μm polystyrene beads (Thermo Fisher Scientific, #A37306) were coated with Kv1.3 or NRIC proteoliposomes at a 5x mass ratio overnight. NRIC was coated onto beads dyed with coumarin to render them blue by fluorescence. Beads were washed and blocked with PBSM. The two beads were used in GEMS with chicken splenocytes, the blue NRIC bead serving as a specificity control.
• 2.Magnetic Dynabead particles: Kv1.3 magnetic beads were prepared as described above. The magnetic particles were incorporated into GEMS with chicken splenocytes.
• 3.Tetrahymena cells: Tetrahymena expressing Kv.1.3 were fixed with 4% paraformaldehyde, washed, and incorporated into GEMS with chicken splenocytes.

Cloning and Expression of Kv1.3 recombinant antibodies. As described in Izquierdo et al., chicken VH and VL gene segments were PCR amplified from individual positive B cells identified in the GEM assay, cloned into a scFv-Fc expression vector (human Fc), and transiently expressed in Expi293 cells grown in Expi293 Expression Medium using Expifectamine (Thermo Fisher Scientific, #A14524), in 96-well format.³⁸ Supernatants were tested for binding by ELISA, as described above, using anti-human-Fc-HRP (VWR #RL609-4303) secondary antibody. Unique clones were re-transfected in a 2 ml culture to confirm binding and for further evaluation. Supernatant antibody concentrations were quantitated by sandwich ELISA using anti-human-Fc and a scFv-Fc control. To purify antibodies for electrophysiology, 40 ml cultures were used for transfections, and antibody purified using Protein G Agarose (Thermo Fisher Scientific, #20399B) according to the manufacturer's protocol.

Flow cytometry: Jurkat cells (clone E6, ATCC) were grown in RPMI (Thermo Fisher Scientific, #450000-404) supplemented with 10% fetal calf serum (Hyclone, #SH30071.03) and Glutamax (Thermo Fisher Scientific, #35050061). Cells (1.5e5 /well) were incubated at 4°C for 1h with 10 μg/ml of each antibody diluted in PBS/1% BSA/0.1% sodium azide (FACS buffer), washed 2 times with FACS buffer, incubated with anti-human IgG-PE (Jackson Immunoresearch, #109-115-098) at 5 μg/ml at 4°C for 1h, washed 2 times with FACS buffer, resuspended in FACS buffer and read on an Attune Cytometer (Thermo Fisher Scientific).

Llama

Llama immunization and library construction: Llama immunization, peripheral blood lymphocytes (PBLs) isolation, RNA extraction and the construction of scFv libraries were performed as described previously.³⁹ All llamas used in this study were farmed outdoors in the Ardèche region of France according to French animal welfare legislation. Animal handling included immunization by intramuscular injection and peripheral blood collection. As a rule, experimental use of llamas is limited to a single immunization protocol; following the last blood drawing, animals are employed for recreational use. Within the framework of the guidelines four llamas were immunized six times biweekly with 2 mg/ml plasmid DNA encoding human Kv1.3 under control of a CMV promoter. A boost with 50 μg Kv1.3 proteoliposomes was given two times biweekly. Before the start and a week after completion of the immunizations, serum was prepared from 10 ml blood in EDTA tubes. PBLs were isolated from 400 ml blood collected five days after the last immunization, and RNA extraction and cDNA amplification was done as formerly described.⁴⁰ Immune responses were measured on pre- and post-immunization serum samples by mesoscale against both Kv1.3 and NRIC proteoliposomes, coated at 1 μg/ml. Detection of serum titrations was performed using a mouse anti-llama IgG-CH1 (in-house antibody) and an anti-mouse sulpho antibody (MSD, #R32AC-5). Subsequently, scFv libraries were made for two of the four immunized llamas as previously described.³⁹ In brief, VH, Vκ and Vλ libraries were independently constructed into the pSc vector, which was derived from the pCB3 phagemid vector. Re-cloning the light chain libraries in the VH library was performed, to obtain the combinatorial κ and λ library for each llama.

Phage selection: Phage production and selection were performed according to standard protocols.^39-41 Phages expressing scFv were allowed to bind for 2 h at 4°C to rotating Kv1.3 or NRIC magnetic beads. Washing steps were performed using PBS containing 0.05% Tween and elution of phages was done by using 10 mg/ml trypsin (Sigma-Aldrich, #T1426). Two rounds of selection were carried out using 10 μg Kv1.3 magnetic beads and 20 μl input phages in the first round and 2 μg and 2 μl in the second round, respectively, to enrich for phages expressing scFv binding specifically to Kv1.3. The Kv1.3-binding clones were isolated as individual colonies and expression of soluble scFv in the periplasmic space was induced using isopropyl β-D-1 thiogalactopyranoside (Thermo Fisher Scientific, #R0391).

Screening binding assays: Individual soluble scFv in periplasmic fractions were screened for binding to 1 μg/ml Kv1.3 and NRIC proteoliposomes using mesoscale. The detection was carried out using mouse anti-myc antibody 9E10 (Bethyl Laboratories, #A190-105P) and an anti-mouse sulpho antibody. The cut-off ratio for binding to Kv1.3 over NRIC proteoliposomes was set to 2. The positive clones were sent for sequencing (LGC Genomics) to determine the number of families based on VH CDR3 sequence length and similarity.⁴² VH and VL sequences from selected clones were re-cloned in the scFv-Fc backbone where the scFv was fused to the hinge-CH2-CH3 domain of a human IgG1. Transfection, production in HEK293E cells (ATCC, #CRL-10852) and protein A purification were performed as previously described.⁴⁰ scFv-Fc screening using a 1/10 dilution series starting from 1 μg/ml was carried out to confirm binding to Kv1.3. By mesoscale, screening for binding to Kv1.3 and irrelevant proteoliposomes, coated at 1 μg/ml was done using an anti-human sulpho detection antibody (MSD, #R32AJ-5). As a positive control, a polyclonal anti-Kv1.3-biotin antibody was also assessed (Alomone, #APC-101-B).

Antibody screening and potency (IC₅₀)

Antibodies were tested on human Kv1.3 channels that were transiently expressed in L929 fibroblast cells (ATCC). Stable expression and sources of the mKv1.1, rKv1.2, hKv1.5, hNav1.5 and hKv11.1 (hERG) channels were described previously.^43,44 Cells were maintained in a culture medium of Dulbecco's modified Eagle's medium (DMEM) supplemented with L-glutamine and 10% heat-inactivated fetal calf serum in a humidified, 5% CO₂ incubator at 37°C. For transient expression, hKv1.3 cells were transfected in conjunction with pEGFP, using FuGENE® 6 transfection reagent (Promega, #E2691) in OptiMEM reduced serum medium (Life Technologies, #31985088), for identification of transfected cells under the epifluorescence microscope after 24 to 72h post-transfection. All cells were detached by trypsinization and plated on cover-slips coated with poly-L-lysine for whole-cell patch-clamp recording using the EPC-10 HEKA amplifier. All currents were recorded in normal Ringer’ solution containing (in mM): 160 NaCl, 4.5 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES (adjusted to pH 7.4 and 290–310 mOsm). Patch pipettes were pulled from soda lime glass (micro-hematocrit tubes, Kimble Chase, Rochester, NY) to resistances of 2–3 MΩ and filled with an internal pipette solution containing (in mM): 145 KF, 2 MgCl₂, 10 HEPES, 10 EGTA (pH 7.2, 290–310 mOsm) for Kv1.1, Kv1.2, Kv1.3 and Kv1.5 recordings; 120 KCl, 10 HEPES, 4 Na₂ATP, 10 EGTA, 5.374 CaCl₂, 1.75 MgCl₂ (pH 7.2, 290–310 mOsm) for hERG recordings; and 10 NaF, 110 CsF, 20 CsCl, 2 ethylene glycol tetraacetic acid, 10 HEPES (pH 7.35, 300 mOsm) for Nav1.5 recordings. Outward K⁺ currents were elicited using repeated 200-ms long depolarizing pulses to 40 mV with a 30-s interpulse interval for Kv1.3 and a 10-s interpulse interval for Kv1.1, Kv1.2 and Kv1.5. hERG currents were recorded with a two-step pulse from −80 mV first to 20 mV for 3 seconds and then to −50 mV for 2 seconds. Inward Na+ currents were elicited by 50-ms pulse to 0 mV from −90 mV applied every 10-s. Cell capacitance, a direct measurement of cell surface area, and access resistance were continuously monitored during recordings to ensure minimal current rundown. Test antibody solutions were freshly made in normal Ringer's solution immediately prior to bath perfusion, which was performed only once currents had stabilized from the preceding control condition or testing dose. IC₅₀ values were determined by fitting the calculated percentage of current blocked to a Hill equation using the software Origin 9.0 (OriginLab). Data are presented as means ± standard deviations.

Epitope binning

A panel of 36 sequence unique anti-Kv1.3 chicken mAbs (scFv-Fc constructs) supplied as crude transfectant supernatants, was epitope binned using a “premix” assay format by array-based SPR imaging (Array SPRi) as described previously.⁴⁵ In addition, select clones were also supplied in purified form (1D8, 1H7, 2A3, and 2G9) along with a functionally inhibiting llama anti-Kv1.3 reference antibody (L1A3) and coupled as a 48-spot array. The analysis was performed at 25°C on a Xantec CMD 200 M sensor chip and in a running buffer of 10 mM Hepes pH7.4, 150 mM NaCl, 3 mM EDTA, 0.05% Tween-20 (HBS-EP+) with 1 g/l BSA using premixed samples containing 33 nM detergent-solubilized Kv1.3 and 480 nM antibody (in binding sites) and regenerating each binding cycle with 75 mM phosphoric acid.

Funding Statement

This work was supported, in part, by the Juvenile Diabetes Research Foundation T1D Fund-PI Paul Colussi, PhD, pcolussi@tetragenetics.com.

Abbreviations

AgTX-2

Agitoxin-2

BSA

bovine serum albumin

CIP

calf-intestinal alkaline phosphatase

CDR

complementarity-determining region

DDM

n-dodecyl-b-D-maltoside

DMEM

Dulbecco's modified Eagle's medium

ER

endoplasmic reticulum

FC12

Fos-Choline-12

FAM

Fluorescein 6-carbonyl

GEM

gel encapsulated microenvironment

IbTX

Iberiotoxin

HRP

horse-radish peroxidase

LPS

lipopolysaccharide

MTT

metallothionein

MgTX

Margatoxin

NRIC

non-related ion channel

PBS

phosphate-buffered saline

SPR

surface plasmon resonance

Shk

Stichodactyla toxin

scFv

single chain variable fragment

VGIC

voltage-gated ion channel

Tth

Tetrahymena thermophila

IV

intravenous