Unique epitope–antibody interactions in the intrinsically disordered proteoglycan-like domain of human carbonic anhydrase IX defined by high-resolution NMR combined with yeast surface display

ABSTRACT

Carbonic anhydrase (CA)-IX is an extracellular enzyme that is essential in the adaptation of tumor cells to their increasingly more hypoxic and acidic microenvironment. Within the family of carbonic anhydrases, CA-IX is unique in that it is the only CA with an N-terminal intrinsically disordered region (IDR) containing a proteoglycan (PG)-like domain. This PG-like IDR has been described to be instrumental in CA-IX’s enzyme activity, as well as tumor cell motility and invasion. We have characterized the antibody–epitope interactions of two novel and unique antibodies (11H9 and 12H8) that are specific for the human CA-IX’s IDR. Binding interactions of these antibodies to the intact IDR were studied by surface plasmon resonance and high-resolution nuclear magnetic resonance (NMR) spectroscopy, while the specific epitopes were determined by both NMR and yeast surface display (YSD). Our data show that 12H8 binds to the N-terminus of CA-IX, while 11H9 has a high affinity for an epitope located in the central region of the IDR containing three GEEDLP repeats in a manner that is different from the previously described M75 antibody. Titration NMR spectroscopy using CA-IX’s entire IDR in addition identified a secondary epitope of 11H9 at the beginning of the PG-like domain that remains exposed and available for further binding events after the engagement at its primary epitope at the center of the PG-like domain. Transverse relaxation optimized NMR spectroscopy of 11H9-F(Ab) in complex with the CA-IX IDR outlines structural rigidification of a linear epitope, while the rest of the IDR remains largely unstructured upon complex formation. This study illustrates how high-resolution NMR and YSD are used as complementary tools for a comprehensive characterization of antibody–epitope interactions involving intrinsically unstructured antigen domains with highly repetitive sequences.

KEYWORDS:

• Antibodies
• antibody–antigen interactions
• carbonic anhydrase IX
• epitope mapping
• F(ab)
• intrinsically disordered proteins
• NMR spectroscopy
• yeast surface display

Introduction

Carbonic anhydrases (CA) are a family of 15 ubiquitous enzymes that catalyze the reversible hydration of CO₂, resulting in the generation of HCO₃⁻ and H⁺ with differing efficiency in addition to a variety of other reactions. CAs can be found intracellularly in several compartments and attached to cells via a transmembrane domain or glycosylphosphatidylinositol anchor. When secreted,^1,2 CAs are involved in pH homeostasis of normal cells and tissues and some are of critical importance for the survival of cancer cells.^3,4 As tumors grow the diffusion distances to the nearest blood vessels increase, and this not only limits access to oxygen and nutrients but also prevents the proper removal of metabolic waste. Low oxygen levels cause inhibition of oxidative phosphorylation and tumor cells survive by switching to a glycolytic metabolism, resulting in the intracellular accumulation and subsequent extrusion of lactate, which in turn leads to an acidic extracellular pH (pHe). Acidification of the tumor microenvironment (TME), also known as acidosis, disrupts the intracellular pH (pHi) homeostasis, which is detrimental to many cellular functions. Tumor cells adapt to acidosis by activating a very efficient pH-regulating system that uses CAs to prevent intracellular acidification. The hypoxia-driven upregulation of CA-IX together with the activation and upregulation of acid/base transporters (i.e., Na⁺/H⁺ exchangers, Na⁺/HCO3⁻ cotransporters, anion exchangers, and monocarboxylate transporters) form very efficient transport metabolons that are critical for maintaining the proper acid/base balance in tumor cells.⁵

Of the three exofacial CAs (CA-IV, CA-VII, and CA-IX), CA-IX is structurally unique in that it contains an N-terminal proteoglycan (PG)-like domain (residues 53–111) rich in negatively charged amino acids. It has been suggested that this PG-like domain aids the catalytic efficiency of CA-IX in the acidic microenvironment⁶ by acting as a proton buffer in solid tumors that optimizes the hydration of CO₂.⁷ The functional importance of CA-IX’s PG-like domain is further illustrated by the reduced adhesion and increased spreading of cancer cells expressing a mutant form of CA-IX that lacks the entire PG-like domain, or in the presence of antibodies that target a specific region of the PG-like domain.^8–10

The primary sequence of the PG-like domain contains a number of prolines which delineate a series of negatively charged glutamic and aspartic acids. This sequence pattern is very similar to that of the keratan sulfate enriched region of aggrecan, a large proteoglycan, hence the name PG-like domain.¹¹ It is interesting to note that a recent study identified CA-IX’s PG-like domain as a novel member in a large class of intrinsically disordered proteins and protein regions (IDPs/IDRs) that are devoid of a folded three-dimensional (3D) structure.¹² Another feature of CA-IX’s PG-like domain is the presence of a tandem repeat of three identical GEEDLP motifs, which is preceded by four motifs that are highly homologous to GEEDLP^11,12 with a repetitive sequence that is also characteristic of many other IDPs/IDRs.^13,14

In this study, we characterized the molecular interactions of two novel antibodies (11H9 and 12H8)¹⁵ that specifically target the intrinsically disordered region of CA-IX (residues 37–136; see Table 1). To this end we exploited surface plasmon resonance (SPR), yeast surface display (YSD) and NMR spectroscopy using double (¹⁵N and ¹³C) and triple (²H, ¹⁵N and ¹³C) isotope labeled recombinantly expressed CA-IX’s intact IDR to examine this IDR’s detailed molecular interactions and to determine the binding epitopes of the 11H9 and 12H8 monoclonal antibodies (mAbs). The data presented here identify the N-terminus of CA-IX’s IDR as the binding epitope for 12H8, while a region spanning two negatively charged GEEDLP tandem repeats at the center of the PG-like domain as the primary binding epitope for 11H9. Further titration studies using high-resolution NMR identified a secondary epitope located in the N-terminal region of the PG-like domain for the 11H9 mAb. Here we also show evidence that the 11H9 mAb is capable of simultaneously binding both epitopes in the CA-IX IDR only in an intermolecular fashion, leading to the formation of multi-molecular cross-linked antibody aggregates.

Table 1. Sequence of the ePG-1 polypeptide containing an N-terminal methionine (M) from the bacterial expression vector and a C-terminal linker-His-tag for purification purposes.

Peptide	CA-IX residues*
ePG-1	37–136	MPQRLPRMQEDSPLGGGSSGEDDPLGEEDLPSEEDSPREEDPPGEEDLPGEEDLPGEEDLPEVKPKSEEEGSLKLEDLPTVEAPGDPQEPQNNAHRDKEGDAAALEHHHHHHHH

*Proline (P) after the bacterial methionine is residue 37 in the amino acid sequence of full-length CA-IX and the aspartic acid (D) before the purification tag (in underlined italics) is CA-IX residue 136. In NMR assignments, these residues are labeled as Pro1 and Asp100, respectively (see Figure 2C).

Results

11H9 and 12H8 CA-IX mAbs specifically bind distinctive epitopes within hCA-IX’s intrinsically disordered region

Antibodies 12H8 and 11H9 were among a pool of six novel mouse mAbs that were generated by immunization of mice with the entire extracellular domain of human CA-IX as an immunogen and selected for their potential as the antibody component of an antibody–drug conjugate.¹⁵ MAbs 11H9 and 12H8 did not interfere with the catalytic activities of CA-IX and both exhibited specific binding to CA-IX’s N-terminal region as demonstrated through Pepscan evaluation.¹⁵ To further characterize the binding epitopes of these two antibodies, we recombinantly produced the ePG-1 and ePG-2 polypeptides (Table 1; SupplTable S1) which cover the IDR of CA-IX. Before initiating NMR experiments, we evaluated the mAb binding characteristics of these polypeptides by SPR. First, we used two control antibodies: M75 mAb, which is known to bind CA-IX’s PG-like domain,⁸ and 2D7, which is a catalytic domain-specific mAb^15,16 and therefore should not bind. Our data (Figure 1a and SupplFig.1A) clearly show the expected binding by M75 and the lack thereof by 2D7, indeed confirming that the mAb binding characteristics of CA-IX’s PG-like domain have been retained in the ePG-1 and ePG-2 polypeptides. We then evaluated the binding of the 11H9 and 12H8, in both a mAb and F(Ab) format and found that 11H9 binds both the ePG-1 (Figure 1b) and ePG-2 (SupplFig. S1B) polypeptides. However, we also observed that 12H8 only binds the ePG-1 (Figure 1c) and not the ePG-2 (SupplFig. S1C) polypeptide, indicating that 12H8 binding relies on the presence of four amino acids (PQRL) located at the exposed N-terminus of CA-IX that are present in the ePG-1 (Table 1) but not in the ePG-2 (SupplTable S1) polypeptide. The 12H8 and 11H9 F(Ab) fragments display binding profiles similar to the full-length mAb equivalents (Figure 1a,b,c and SupplFig. S1B, C). However, due to their monovalent character, F(Ab)s bind, in general, with reduced affinity resulting from their lack of avidity. This characteristic can be observed for both the 12H8 and 11H9 F(Ab) when binding to the ePG-1 polypeptide (Figure 1b,c and Table 2). Interestingly, the 11H9 F(Ab) has an apparently slow rate of dissociation from both surface-immobilized ePG-1 (Figure 1, panel b) and ePG-2 (SupplFig. S1, panel b) in contrast to the 12H8 F(Ab) (Figure 1, panel c).

Figure 1. SPR binding studies of the CA-IX mAbs and F(Ab)s to immobilized ePG-1 polypeptide. SPR sensorgrams of (a) mAb M75 and mAb 2D7, (b) mAb 11H9 and its F(Ab) fragment, and (c) mAb 12H8 and its F(Ab) fragment to the immobilized ePG-1 polypeptide. MAb 2D7, which binds to CA-IX’s catalytic domain¹⁵, and mAb M75, which binds CA-IX’s PG domain⁸ were used as a negative and positive control, respectively. MAbs were flowed at 1.23, 3.7, 11.1, and 33.3 nM whereas F(Ab)s were flowed at 1.23, 3.7, 11.1, 33.3 and 100 nM over the chip surface with immobilized ePG-1.

Three panels of SPR sensorgrams of four monoclonal antibodies (mAbs), M75, 2D7, 11H9, and 12H8 and the F(Ab) fragments of 11H9 and 12H8. Left panel (a) has the sensorgram of M75 on top with the typical fast on (loading) phase and slow off (dissociation) phase and 2D7 at the bottom with no SPR signal. Middle panel (b) compares the sensorgrams of 11H9 with a very slow off phase (top) and its F(Ab) fragment with a faster, but still slow, off phase (bottom). Right panel (c) illustrates the sensorgrams of 12H8 with a faster off phase (top) compared to 11H9 and its F(Ab) fragment with a fast phase of dissociation.

Table 2. Phenomenological kinetic values obtained by fitting a 1:1 binding model to the SPR data for the 11H9, 12H8, and M75 anti-CA-IX mAbs and the F(Ab)s of 11H9 and 12H8 to the ePG-1 and ePG-2 polypeptides.

	e-PG1			e-PG2
mAb	ka	kd	KD	ka	kd	KD
11H9	1.70E + 06	LOD^a	LOD^a	1.82E + 06	LOD^a	LOD^a
12H8	4.78E + 06	2.71E–04	5.67E–11	NB^b	NB^b	NB^b
M75	2.54E + 03	2.70E–04	1.06E–07	2.90E + 03	2.57E–04	8.91E–08
F(Ab)
11H9	4.47E + 05	1.71E–03	3.83E–09	4.60E + 05	1.68E–03	3.66E–09
12H8	1.59E + 06	7.46E–02	4.70E–08	NB^b	NB^b	NB^b

^aThese values could not be determined from SPR sensorgrams as a result of the lack of dissociation of surface-bound 11H9 mAb in the timeframe (~10 minutes) of SPR experiments (Figure 1 and SupplFig. S1).

^bNB – No binding.

In-depth characterization of the interactions of 12H8 and 11H9 by NMR spectroscopy

Using high-resolution NMR spectroscopy, the ePG-1 polypeptide was, as expected, found to be unfolded without any significant secondary structure preferences. For example, the ³J_HN,HA coupling constants were determined to be 6–7 Hz for almost all (assigned) residues of the ePG-1 peptide (Figure 2a), which is indicative of random-coil conformations.¹⁷ Proton-¹⁵N heteronuclear NOEs (with ¹⁵N-labeled ePG-1) were very small (<0.3) and negative for many residues (Figure 2b), which are characteristic of unstructured/unfolded proteins.¹⁸ Such a lack of folding for ePG-1 is not surprising given the highly negatively charged nature of the PG-like domain and the low sequence complexity of the N- and C-terminal regions beyond the charged PG-like sequence.¹² Regardless, detailed NMR ¹H/¹⁵N-heteronuclear single quantum coherence (HSQC) assignments of ePG-1 (Figure 2c) and ePG-2 (SupplFig. S2A) showed a significant spectral resolution of 8 of the 11 Gly residues in the highly flexible conformations of these IDR fragments of CA-IX. Interestingly, the three remaining Gly residues with heavily overlapping NMR signals, i.e., Gly43, Gly49, and Gly55 (in ePG-1 NMR assignment numbering, see Figure 2c and SupplFig S2A), are located at the center of the PG-like domain, each at the beginning of the three tandem repeats of the GEEDLP sequence. Another feature of ePG-1 (and ePG-2) is the clustering of many residues at the center of the ¹H/¹⁵N HSQC spectra in which one finds some Asp and Glu residues and others, e.g., Asp21, Asp22, Leu24, Glu27, and Asp28, from the GEDDPLGEEDLP sequence containing the fourth GEEDLP motif (Figure 2c and SupplFig. S2A).

Figure 2. Solution conformation of ePG-1 characterized by NMR using backbone ³J_HN,HA coupling constants, ¹⁵N{¹H} heteronuclear NOE measurements and detailed resonance assignments. (a) essentially uniform ³J_HN,HA coupling constants (of 6–7 Hz) indicate largely random coil conformations for the ePG-1 backbone. (b) Small ¹⁵N{¹H} heteronuclear NOE values with excursions to negative values are also characteristic of unfolded polypeptide chains for ePG-1. NMR data for the central region of ePG-1 could not be analyzed due to overlaps of the ¹H/¹⁵N-HSQC signals of the highly repetitive sequence PPGEEDLPGEEDLPGEEDLP. The companion ¹⁵N longitudinal (T₁) and transverse (T₂) relaxation data are shown in SupplFig. S3B and SupplFig. S3C, respectively. (c) assigned HSQC spectrum of ePG-1. The amino acid sequence of ePG-1 (Table 1) is reproduced here with Pro (P37 of the hCA-IX sequence) as the first residue for the NMR assignment numbering. The italicized black font indicates the locations of the 11 Gly residues in the sequence of ePG-1. The inset details some signal assignments of the heavily overlapping region in the HSQC spectrum. A rectangular box outlines the side-chain amide HSQC signals of Asn and Gln residues. Two Gln residues, Q2 and Q8, and two Asn residues, N91 and N92 are located at the N-terminal and C-terminal regions of ePG-1, respectively.

Plots of the 3JHN,HA coupling constants at the top (a) and the 15N{1H} heteronuclear NOE values at the bottom (b) versus the assigned residues of 15N-labeled ePG-1. These plots have an empty region in the middle of the amino acid sequence of ePG-1 due to spectral overlaps and missing asignments. Sub-Figure c shows the 1H/15N-HSQC spectrum of 15N-labeled ePG-1 annotated by the assigned residues (on the right) and on the left displays the numbered amino acid sequence of ePG-1 used for NMR assignments.

Residue-specific interactions of the 12H8 and 11H9 mAbs with the CA-IX IDR were then followed by perturbations in the H/¹⁵N-HSQC spectra of ¹⁵N-labeled ePG-1 (Figure 3, SupplFig. S3 and SupplFig. S4) induced by adding the full-length antibodies and their respective F(Ab)s. Additions of the full-length 12H8 (SupplFig. S3A) or its F(Ab) (SupplFig. S3B) led to the disappearance of the same set of HSQC signals of ePG-1, demonstrating the equivalence of full-length 12H8 and its F(Ab) (Figure 1). These disappearing H/¹⁵N-HSQC signals were assignable to residues Leu4, Arg6/Asp10, Glu9, and Ser11 (Figure 3a using NMR assignment numbering; or residues L40, R42, E45, D46, and S47 using CA-IX’s sequence numbering in the single-letter aa code) along with large HSQC peak shifts of residues Gly14, Gly15, and Gly16 (i.e., CA-IX residues G50, G51, and G52) (Figure 3a and SupplFig. S3C). On the other hand, there are no backbone amide H/¹⁵N-HSQC signals for residue Pro1 (which has an N-C bond in place of an amide NH) and for residues Gln2 and Arg3, whose amide protons have weak, and therefore undetectable, proton NMR signals due to fast exchange at neutral pH conditions.¹⁹ Residues Gln2 and Gln8 also exhibited a strong response to 12H8 binding, involving their sidechain amide protons (Figure 3a). Binding of 12H8, therefore, is localized to the N-terminal region of ePG-1 (SupplFig. S3C) or within the CA-IX’s N-terminal sequence of PQRL₄₀PRMQE₄₅DSPLG₅₀GG.

Figure 3. (a) ¹H/¹⁵N-HSQC spectrum of ePG-1 responding to 12H8-F(Ab) binding with assignments of ePG-1 signals that perturbed and disappeared (shown in black) in the 1:1 complex. The inset shows the detailed responses of the 11 Gly signals with the disappearance of the signals of G14 and G16 of free ePG-1 (black) along with the appearance of two new HSQC signals for the complex (red). Signal movements in response to 12H8 binding were assigned tentatively as indicated by arrows in the inset, producing the chemical shift perturbations for G14, G15 and Gl6 in decreasing amplitudes. (b) Gly region of ePG-1 HSQC responding to 11H9-F(Ab) binding; the left-most panels show the superposition of the complex spectra with that of free ePG-1 (black) exemplifying the response of G25 while the right panels detailing Gly signal intensity evolutions from free ePG-1 (black) to the 1:1 (red), 1:2 (blue) and 1:5 (green) complexes, respectively. (c) Ser region of the ePG-1 HSQC spectrum responding to excess 11H9-F(Ab) binding, i.e., From 1:1 (red) to 1:2 (blue) and to 1:5 (green) for the ePG-1/F(Ab) ratio, respectively; free ePG-1 spectra are shown in black.

Read the detailed description of this figure

Select regions of the 1H/15N-HSQC spectra of free ePG-1 and in the presence of the F(Ab) fragments of 12H8 (a) and 11H9 (b and c). For (a), the spectrum of free ePG-1 is plotted in one color (black) while in another color (red) the HSQC spectrum of ePG-1 in the presence of the 12H8 F(Ab) fragment at the 1:1 molar ratio. The perturbed residues thereby appear as having black colors in the presence of F(Ab), henceforth labeled by the assigned residues of ePG-1. Also in plot (a), there is an inset showing the details of signal movements in the Gly region of the HSQC spectrum. Plot (b) displays Gly signal movements of ePG-1 with increasing concentrations of 11H9 F(Ab), from 1:0, to 1:1, to 1:2 to 1:5 for the molar rations of ePG-1 and F(Ab), respectively. Plot (c) shows details of signal movements in another HSQC spectral region of ePG-1 with the same increasing concentrations of 11H9 F(Ab), from 1:0, to 1:1, to 1:2 to 1:5. This plot outlines resonance perturbations of select Ser residues (indicated by signal labeling) of ePG-1 responding to 11H9 F(Ab) binding.

On the other hand, no significant perturbations were evident upon adding the 11H9 mAb to ¹⁵N-labeled ePG-1 (SupplFig. S3D). 11H9 F(Ab) titrations also induced limited perturbations in ePG-1 (SupplFig. S3E), mostly concentrated around the area of the first GEEDLP (residues Gly25-Pro30 in NMR assignment numbering; Figure 2c), with the most evident signal displacements for Gly25 in the presence of excess amounts of F(Ab) beyond the 1:1 stoichiometry (Figure 3b). Interestingly, the signal cluster for residues Gly43, Gly49, and Gly55 has a substantially decreased intensity for the 1:1 complex, which became much more pronounced for the 1:2 ePG-1:F(Ab) and 1:5 complexes (Figure 3b). These are accompanied by identifiable signal perturbations for some assignable Glu, Asp, and Leu residues from the GEEDLP sequence (SupplFig. S3F). The main binding site for 11H9 may therefore lie in this central region of the PG-like domain with three tandem repeats of the same GEEDLP sequence which created heavy resonance overlaps (Figure 2c). After this higher-affinity site is occupied (i.e., at the 1:1 complex), excess amounts of 11H9 F(Ab) become available for binding to lower-affinity sites, as shown by resonance perturbations at Gly25 only under such conditions (Figure 3b and SupplFig. S3F). Very importantly, the excess 11H9 F(Ab) also produced significant resonance perturbations for Ser31 and Ser35 and clear doubling of HSQC peaks for Ser17 and Ser18 (Figure 3c), all surrounding the first G₂₅EEDLP₃₀ sequence at the beginning of the PG-like domain.

TROSY NMR characterization of the ePG-1 IDR in complex with the 11H9 F(ab)

The observation of the 11H9-F(Ab)-induced perturbations at two sites in ePG-1, especially in the central region of the PG-like domain (Figure 3B) motivated us to perform a more detailed examination on how three identical ‘GEEDLP’ tandem repeats may interact with the 11H9 antibody. The ePG-1 protein was therefore deuterated (substituting all non-exchangeable carbon attached protons with deuterium) and labeled with the ¹⁵N isotopes at the same time. This double-labeling strategy enables the use of TROSY NMR at high magnetic fields, which can detect rigidly structured residues in large complexes (molecular weight of >60 kDa),²⁰ such as ePG-1 bound to the 11H9-F(Ab). Eight additional signals (crosspeaks) appeared in the H/¹⁵N-TROSY spectrum of ¹⁵N/¹³C/²H-labeled ePG-1 (the additional ¹³C-labeling for TROSY signal assignments, see Methods Section) in a 1:1 complex with the 11H9-F(Ab) (Figure 4). Interestingly, the rest of the TROSY spectrum of the 11H9-F(Ab)/ePG-1 complex was essentially identical (Figure 4) to the ¹H/¹⁵N-HSQC spectrum of free ePG-1 (Figure 2c), as already shown in HSQC titration experiments with selectively perturbed residues (Figure 3b/c, SupplFig. S3D and SupplFig. S3E). In other words, a 1:1 complexation of the 11H9-F(Ab) does not dramatically influence the disordered conformations of ePG-1 except for those residues that form and surround the direct structural contact between 11H9 and its epitope residues in ePG-1. 3D TROSY-HNCACB²¹ NMR data with ¹⁵N/¹³C/²H-labeled ePG-1 allowed the assignment of the few TROSY signals to the PGEEDLPG sequence, located within the central region of ePG-1 containing three identical GEEDLP repeats. Specifically, three well-separated TROSY signals were assigned to the E, D, and L residues of the EDLP unit, two assigned to D and L residues of the DLP sequence, and two as originating from two different PG motifs, all based on the unique ¹³C_α and ¹³C_β chemical shifts and the inter-residue correlations in the TROSY-HNCACB spectrum. In other words, TROSY assignments establish that the primary binding epitope of 11H9 is indeed located at the sequence of PGEEDLPGEEDLP with two PG sequences, one EDLP unit and one DLP unit. Furthermore, the resolution of EDLP and DLP by TROSY shows that 11H9 F(Ab) binding to this epitope creates an asymmetry to the two essentially identical EDLP units in the tandem repeats of two GEEDLP motifs. There are still two partially overlapping sequences of PGEEDLPGEEDLP in PGEEDLPGEEDLPGEEDLP with three GEEDLP repeats, but NMR data alone cannot distinguish between these two potential epitopes.

Figure 4. Amide ¹H/¹⁵N-TROSY NMR spectrum of ¹³C/¹⁵N/²H-labeled ePG-1 in a 1:1 stoichiometric complex with the 11H9-F(Ab) fragment. Well-dispersed signals are labeled with the respective residue assignments in the triple-labeled ePG-1. It should be noted that the remaining signals in the rest of this TROSY spectrum are similar in chemical shift positions to the ¹H/¹⁵N-HSQC spectra of free ePG-1 (see Fig. 2C).

Plot of a region of the TROSY spectrum of ePG-1 annotated by the assigned residues of ePG-1 in its complex with the 11H9 F(Ab). The TROSY spectrum also contains many HSQC signals distributed within a narrow band of (NH) chemical shifts from 7.6 to 8.6 ppm.

YSD epitope mapping of the 11H9 and 12H8 mAbs

Fragment scanning of the CA-IX N-terminal fragment: To gain further insights into the binding epitopes of the 11H9 and 12H8 mAbs, we examined the interactions of CA-IX’s N-terminal IDR with these and the M75 control antibody by using yeast S. cerevisiae displaying the full-length CA-IX IDR and overlapping peptides (Figure 5a). All three antibodies (11H9, 12H8, and M75) were able to bind the native and heat-denatured yeast-displayed full-length IDR, suggesting either linear/contiguous binding epitope(s) for these antibodies or the lack of a defined 3D structure for CA-IX’s N-terminal region, the latter of which is shown conclusively by NMR spectroscopy.

Figure 5. Epitope mapping by yeast surface display strategy. (a) Cartoon depicting the CA-IX N-terminal fragments that were expressed on the cell surface of yeast. (b) ELISA results showing the yeast-displayed CA-IX fragments binding to mAb 11H9 (red), 12H8 (blue) and M75 (gray).

Graphic illustration of CA-IX fragments studied by yeast surface display. Panel (a) shows how the displayed fragments cover the N-terminal (i.e., PG) region of CA-IX and (b) the plot of yeast ELISA of three monoclonal antibodies, 11H9, 12H8, and M75, showing the differing binding behaviors of the three antibodies to different CA-IX fragments. 11H9 has large ELISA readings for fragments 77–91, 87–101, 67-111, and 37–140; 12H8 for fragments 37–71 and 37–140; M75 with large ELISA readings for 57–71, 77–91, 37–71 and weaker, but non-negligible ELISA signals for fragments 87–101 and 107–121.

To obtain a higher resolution epitope map, peptide fragments of 15 amino acid (aa) residues (with a five aa tiling overlap) and covering residues 37–140 (Fragment #14) of hCA-IX were expressed on the yeast cell surface display (Figure 5a). Cell ELISA data (Figure 5b and SupplTable S2) show that 12H8 binds to a distinct 15-aa epitope (Fragment #1; aa 37–51, PQRLPRMQEDSPLGG). 11H9, however, binds to an epitope whose sequence is shared by peptide fragments PPGEEDLPGEEDLPG (Fragment #5; aa 77–91) and EDLPGEEDLPEVKPK (Fragment #6; aa 87–101), while the commercial M75 antibody binds strongly to an epitope whose sequence is shared by peptides DDPLGEEDLPSEEDS (Fragment #3; aa 57–71) and PPGEEDLPGEEDLPG (Fragment #5; aa 77–91). Yeast display data also indicate some level of M75 binding with two additional peptide fragments, i.e., EDLPGEEDLPEVKPK (Fragment #6; aa 87–101) and SLKLEDLPTVEAPGD (Fragment #8; aa 107–121) with weaker ELISA signals (Figure 5b and SupplTable S2). Such different binding characteristics of 11H9 and M75 indicate that these two antibodies recognize distinct epitopes, albeit with some degree of sequence similarity, especially in the central region of CA-IX’s PG-like domain, while 12H8 binds to a 15-aa peptide fragment at the N-terminal segment of CA-IX (Figure 6a). The commercial antibody M75, therefore, emerged the most promiscuous, having multiple epitopes, in the central region with its known linear epitope GEEDLPG⁸, in the homologous repeat GEEDLPS preceding the primary epitope as well as two sequences GEEDLPE and KLEDLPT having four identical residues EDLP.

Figure 6. Fine epitope mapping using yeast surface display (YSD). (a) Summary of antibody binding of yeast-displayed 15-residue tiling fragments of the CA-IX N-terminal region, residues 37–140. The colored sequences (12H8, green; 11H9, red, M75, blue) are the fragments within CA-IX IDR that showed the highest binding to each of the indicated antibodies by yeast cell ELISA (see Fig. 5). (b) Fine epitope mapping on 15-residue fragments that were identified to bind the antibodies through N-and C-terminal truncations to determine the “minimal” epitope followed by mutagenic alanine scan. Residues in large font and bold are those that do not tolerate alanine substitution indicating their critical role in antibody binding (see text for details).

Display of the amino acid sequence of residues 37–140 along with the 12 15-aa overlapping peptide fragments, illustrating the differing epitope (binding) loci for the three antibodies, 12H8, 11H9, and M75. Panel (a) shows the binding epitope in the form of 15-residue fragments, while panel b outlines minimal epitope for each antibody identified by sequence truncation and alanine substitutions.

Fine epitope mapping of the CA-IX IDR-binding mAbs: We then carried out a ‘fine epitope mapping’ experiment, focusing on those fragments that had strong ELISA signals (Figure 5a), through truncation and mutagenic alanine scan (Figure 6). The truncation analysis defined the “minimal” peptides that are both necessary and sufficient for the binding to these antibodies (Figure 6b): LPRMQEDSP for 12H8; EDLPGEED for 11H9; and GEEDLP(S/G) for M75. This analysis also showed that replacing one of residues L, R, Q, E, and D, in the 12H8 epitope LPRMQEDSP with an alanine abolished binding of 12H8 (as defined by yeast ELISA readings). Intriguingly, none of the amino acids of the 11H9 epitope EDLPGEED could tolerate substitution for alanine, whereas in the M75 epitope GEEDLP(S/G), alanine replacement of only two residues (G and E) results in loss of binding (Figure 6b). It should be noted that all five epitope residues identified by epitope fine mapping as essential for 12H8 binding also showed severe NMR HSQC signal perturbations leading to obvious loss of signal intensities (i.e., L40, R42, E45, and D46 and the sidechain amide of Q44, which correspond to NMR assignments of L4, R6, E9, D10, and the sidechain amide of Q8, respectively, in Figure 3a) in NMR titration experiments. The primary epitope for 11H9 lies at the junction of two consecutive repeats of GEEDLP, which is in complete agreement with TROSY NMR data showing high structural rigidification of the PGEEDLPG sequence moiety upon binding of this mAb’s F(Ab) (Figure 4). It is interesting that YSD data identified the high-affinity (primary) binding epitope EDLPGEED for 11H9 from the central region of the PG-like domain but did not locate the weaker and secondary binding site around the first G₂₅EEDLP₃₀ sequence which was identified by NMR titration experiments (Figure 3b/c and SupplFig. S3F). Indeed, 11H9 showed no binding to yeast cells displaying peptide 57–71 (Fragment #3), 67–81 (Fragment #4) and 37–71 (Fragment #12) (Figure 5b), despite that these peptide fragments contain DDPLGEED and EDLPSEED (Fragment #3 and #12), and EDSPREED (Fragment #4), each being a variant of the high-affinity epitope EDLPGEED.

Higher-order interactions between the CA-IX IDR and the mAb 11H9

With TROSY NMR data showing a largely unstructured IDR domain in a complex with 11H9 F((Ab) (Figure 4), the question arises as to how the full-length, bivalent 11H9 mAb may interact with CA-IX’s IDR containing two 11H9 epitopes. The ¹H/¹⁵N-HSQC spectrum of ¹⁵N-labeled ePG-1 remained largely unperturbed in the presence of a less than stoichiometric concentration of the 11H9 mAb with a molar ratio of 6:1 for ePG-1 and 11H9, respectively (SuppFig. S3D), showing the lack of NMR exchange lineshapes due to the slow dissociation of the ePG-1/11H9 complex already identified in SPR studies (Figure 1a and Table 2). Very interestingly, decreasing the stoichiometry to 3:1 (PG:11H9) caused some intensity decreases of the HSQC signal cluster of the overlapped residues Gly43/Gly49/Gly55 (SupplFig. S4A) along with selective perturbations of some residues in the tail region of ePG-1, namely residues His94, Ala101, Ala102, Ala103, Leu104, and Glu105 (SupplFig. S4B). No further spectral changes were observed (spectra not shown) when a 3:1 complex was prepared with different concentrations, i.e., 65 µM and 22 µM for ePG-1 and 11H9, respectively, instead of the higher 160 uM and 54 uM concentrations, respectively, for the first 3:1 complex (i.e., as in SupplFig. S4A/B).

Surprisingly, a further decrease of the PG:11H9 molar ratio to (approximately) 2:1 led to turbidity of the complex solution (and precipitation, not shown) accompanied by more pronounced HSQC signal losses of the same Gly43/Gly49/Gly55 cluster and intensity losses for residues His94, Ala101, Ala102, Ala103, Leu104 and Glu105 (SupplFig. S4C/D). Together with these, the HSQC signal of Gly99 starts to broaden in this 2:1 complex (SupplFig. S4C). This is in dramatic contrast with the PG:12H8 complex, which had no signs of either sample turbidity or precipitations (SupplFig. S3A). With the molar ratio adjusted to 1:1, a more pronounced sample turbidity and precipitation occurred in an independently prepared sample of the PG/11H9 complex, which was accompanied by essentially a complete loss of the HSQC signal intensities for the Gly43/Gly49/Gly55 cluster and residues His94, Gly99, Ala101, Ala102, Ala103, Leu104, and Glu105 (SupplFig. S4E/F). In addition, residue Gly25 (which is part of the secondary epitope) exhibited HSQC signal perturbations in this 1:1 complex, along with now clearly visible perturbations on residues Gly19 (SupplFig. S4E) and Ser18, Ser31, Ser35, and Asp40, a cluster of residues around the secondary epitope identified in 11H9-F(Ab) titrations with F(Ab) excess (Figure 3c). This latter observation clearly indicates the availability of the secondary epitope in the soluble 2:1 (PG:11H9) molecular complex for interacting with the fraction of the free unoccupied 11H9 molecule (Figure 7a).

Figure 7. A schematic representation of the unique interactions of the 11H9 mAb with the ePG-1 polypeptide containing two epitopes. (a) the 2:1 (PG:11H9) complex is formed preferentially when the two combining sites of 11H9 are occupied by the high-affinity epitope EDLPGEED (shown in black circles), especially when the concentration of ePG-1 is higher than or equal to the 2:1 stoichiometry. With a concentration of ePG-1 at the 1:1 (PG:11H9) ratio, the same 2:1 (PG:11H9) complex is formed for 25% of the 11H9 concentration while 50% of 11H9 forms a 1:1 (PG:11H9) complex via the high-affinity epitope. The fraction of unoccupied 11H9 (25%) interacts with the low-affinity secondary epitope, e.g. DDPLGEED, exposed in the 2:1 (PG:11H9) complex (shown in pink), leading to differential HSQC signal perturbations of residues around this site (SupplFig. S4). Such a 2:1 (PG:11H9) complex can also be formed on SPR sensor chips with the ePG-1 polypeptide immobilized via its His-tag binding to surface-conjugated anti-His antibodies (see Materials and Methods), leading to the extremely slow rate of dissociation of 11H9 mAb from the sensor chip surface (Fig. 1b and SupplFig. S1B), (a) with a concentration of ePG-1 lower than the 2:1 stoichiometry, the insufficient spacing provided by ~ 22 residues between the high-affinity and low-affinity epitopes in ePG-1 favors intermolecular cross-linking or aggregation of the 1:1 PG:11H9 species, as evidenced by turbidity and precipitation in ePG-1/11H9 samples prepared at ~ 2:1 and ~ 1:1 ratios for the concentrations of ePG-1 and 11H9, respectively (see SupplFig. S4 for further details). The 2:1 (PG:11H9) complex species shown in (A) is also present in these (2:1 and 1:1) samples as detected by both ¹H/¹⁵N-HSQC (SupplFig. S4C/D/E/F) and by size-exclusion chromatography (SupplFig. S5) after removing the cross-linked and insoluble 11H9 aggregates by centrifugation.

Cartoon diagram illustrating the two possible binding modes between 11H9 and the ePG-1 polypeptide. In (a), the mAb 11H9 having two F(Ab)’s is shown to bind ePG-1 preferentially via the high-affinity binding site EDLPGEED, resulting in the 2:1 (PG:11H9) binding stoichiometry. (b) shows the cross-linking action of ePG-1 on 11H9 when their concentration ratio reaches 2:1 (PG:11H9) or below, leading to antibody aggregation.

HSQC signal losses for the Gly43/Gly49/Gly55 cluster in the 1:1 and 2:1 (PG:11H9) samples mirror similar responses in titration experiments using the 11H9 F(Ab) (Figure 3b/c), especially with excess concentrations of the F(Ab) (Figure 3b), showing that the observed HSQC signals in these complexes of ePG-1 with 11H9 mAb are now dominated by ¹⁵N-labeled ePG-1 bound to the mAb instead of being largely from the uncomplexed ePG-1 as in the 6:1 (PG:11H9) complex (SupplFig. S3D). The intermediate sample with a 3:1 (PG:11H9) molar ratio, leaving only ~1/3 of ePG-1 in the uncomplexed state, also showed similar HSQC spectra as the free ¹⁵N-labeled ePG-1 (SupplFig. S4A/B), affirming the largely unstructured nature of ePG-1 whether bound to either the full-length 11H9 mAb or to its F(Ab) (Figure 4). This lack of folding of ePG-1 whether free or bound to 11H9 can also lead to anomalous migration of PG¹² or the PG/11H9 complex through size-exclusion media (SupplFig. S5). PG bound to the 11H9 mAb does distinguish itself in that some tail residues, especially His94, Gly99, Ala101, Ala102, Ala103, Leu104, and Glu105 of ePG-1, appear to be involved in “binding” based on their HSQC signal perturbations (SupplFig. S4E/F). These added tail residues of ePG-1 (Table 1), however, may simply have altered conformational and/or dynamic properties in the large ePG-1/11H9 complex. As such, the 2:1 complex between ePG-1 and 11H9 (Figure 7a) is also very likely what would be captured on SPR sensor chips with each immobilized anti-His antibody binding two ePG-1 molecules (or ePG2, SupplFig. S1), resulting in further slowed rates of dissociation of bound 11H9 mAbs (top panels of Figure 1b and SupplFig. S1B) in addition to avidity effects related to antigen immobilization.

Discussion

Several antibodies that bind human CA-IX, including M75⁸ and cG250,²² were described during the past decades. G250 recognizes a determinant present in renal-cell carcinoma and absent from normal kidneys.²² These earlier CA-IX antibodies have been exploited for cancer detection and as candidates for therapy.¹⁵ More recently, other anti-CA-IX mAbs,^23,24 including the newly characterized 11H9, 12H8, 2D7, 2C7, 9B6, and 4A2,¹⁵ have been reported. The unique mAb 12H8 is being developed as a novel candidate immunoconjugate,¹⁵ especially as a promising radio-immunotherapeutic (RIT) agent (unpublished data). Several of these antibodies, i.e., M75,⁸ H7,²⁴ and 11H9 and 12H8,¹⁵ have been shown to specifically bind to CA-IX’s IDR, especially the PG-like domain. The epitope of the M75 mAb has been well characterized, binding the GEEDLP(S/G) consensus sequence,^8,25 whereas mAb H7 appears to interact with the larger 12-residue GEDDPLGEEDLP epitope.²⁴ Applying the PepScan technology (https://www.pepscan.com) to the entire CA-IX extracellular domain sequence suggested that the 12H8 epitope is located at the N-terminus of the CA-IX sequence, while that of 11H9 is likely to be somewhere in the center of the PG-like domain.¹⁵

In this study, we set out to examine the interactions of the entire IDR, residues 37–136 (i.e., ePG-1; Table 1), of hCA-IX by NMR spectroscopy and to characterize the precise binding epitopes of mAb 11H9 and 12H8 in conjunction with yeast surface display. Our comprehensive NMR data, utilizing the ¹⁵N/¹³C-labeled ePG-1 polypeptide, showed that hCA-IX’s N-terminal extension domain is without any 3D structure or even a secondary structure preference (Figure 2 and SupplFig. S2), which confirms recently published data on the same CA-IX domain using unassigned proton NMR spectroscopy and other biophysical methods.¹² This conformational property can be ascribed in part to the low sequence complexity of the N-terminal (aa 37–55) and the C-terminal residues (aa 107–136) of CA-IX’s N-terminal extension, qualifying them as ‘flexible linkers’ between the signal sequence and the PG-like domain and between the PG-like domain and CA-IX’s catalytic domain (Table 1). The lack of a 3D fold in the PG-like domain (Figure 2) is related to a highly negatively charged sequence that is rich in aspartic and glutamic acids (Table 1), as well as the presence of seven repeats consisting of the following six residue-related sequences: GEEDLP (4), GEDDPL (1), SEEDSP (1), and REEDPP (1). An IDR such as the PG-like domain could thus provide the conformational flexibility that may confer strong avidity effects in binding of antibodies to the native hCA-IX dimer.¹⁵ In addition, the negatively charged intrinsically disordered PG-like domain may also serve to modulate the aqueous structure and dynamic hydration at the extracellular membrane surface. This property has been linked to the non-catalytic extruding of lactate from the intracellular cancer cell compartment and its apparent “antenna” function in the pH regulation by CA-IX’s catalytic domain.²⁶ Our SPR data indicated that the M75, 11H9, and 12H8 mAbs all bind with high affinity to the immobilized ePG-1 polypeptide, i.e., residues 37–136 of hCA-IX’s entire N-terminal region (Figure 1 and Table 1).

Interestingly, the 11H9 mAb and F(Ab) were able to bind the ePG-2 polypeptide, but the 12H8 mAb and 12H8 F(Ab) were unable to bind ePG-2, a variant of ePG-1 that lacks the four N-terminal aa residues PQRL₄₀ (SupplFig. S1). We have shown that 12H8 can be used as an in vivo imaging agent for hCA-IX-expressing tumors (unpublished results), which implies that mature hCA-IX expressed by tumor cells thus contains these extreme N-terminal ‘PQRL’ residues. 12H8 titration data with ¹⁵N-labeled ePG-1 further delineated N-terminally located residues, especially L40, R42, Q44, E45, and D46 (Figure 3a and SupplFig. S3C) as the 12H8 binding loci, which is in complete agreement with positive binding of the mAb to the 15-residue peptide fragment PQRLPRMQEDSPLGG displayed on the yeast cell surface (Figure 5) and the earlier published PepScan data.¹⁵ Fine epitope mapping by alanine scanning of a 9-residue fragment, LPRMQEDSP, located within the 15-residue peptide fragment PPRLPRMQEDSPLGG, identified the core binding for 12H8 to be sequence LPRMQEDSP within which residues L40, R42, Q44, E45 and D46 cannot be replaced by alanine. It can therefore be concluded that of the four extreme N-terminal residues ‘PQRL’ only L40 is absolutely required for the high-affinity binding of 12H8 to hCA-IX’s N-terminus, which is an important property that would render the binding of 12H8 less sensitive to possible processing variations in signal peptide release.

Compared to the unique epitope of 12H8, those of 11H9 are superficially similar to that of the well-known antibody M75.^8,25 First, NMR resonance perturbations on residues Gly43, Gly49, and Gly55 and some Glu and Asp residues in the GEEDLP sequence (Figure 3b and SupplFig. S3F) indicate a primary binding epitope in the central region of the PG-like domain. However, high-resolution (TROSY) NMR of the ePG-1/F(Ab) complex using a fully deuterated and ¹⁵N-labeled ePG-1 polypeptide (Figure 4) revealed that this binding epitope of 11H9 must span two consecutive GEEDLP repeats or the GEEDLPGEEDLP sequence (Figure 4) instead of residing in the single GEEDLP unit like M75.²⁵ It is interesting to note that the antibody H7²⁴ also appears to bind to the tandem of two related 6-residue sequence, i.e., GE₂₀DDPLG₂₅EEDLP₃₀, which overlaps with 11H9‘s secondary or lower-affinity epitope around G₂₅EEDLP₃₀. Our yeast display data showed high-affinity binding of 11H9 with both the PPGEEDLPGEEDLPG and the EDLPGEEDLPEVKPK sequences (Figure 5, 6). Fine mapping of these putative epitopes identified a continuous stretch of negatively charged residues. i.e., EDLPGEED, as 11H9’s primary binding epitope. None of the amino acids in this sequence can be replaced by alanine (Figure 6), indicating the critical importance of the first and all glutamic acids (E) in this epitope for high-affinity 11H9 binding, thus setting this antibody apart from M75 and from the H7 antibody the latter of which binds only the GEDDLPGEED epitope. Given the additional residues perturbed by (excess concentrations of) 11H9-F(Ab) (Figure 3c), i.e., Ser17, Ser18, Ser31, and Ser35, there may be a number of low-affinity 11H9 binding sites, namely, D₂₀DLPG₂₅EED, EDLP₃₀SEED, and EDS₃₅PREED₄₀, all of which are variations of the high-affinity epitope EDLPGEED and whose weaker binding to 11H9 could not be detected directly by displaying on the yeast surface smaller fragments containing these sequences (Figure 5b). There are also two EDLPGEED motifs in the GEEDLPGEEDLPGEEDLP region only one of which should be able to bind at a time due to the overlap of two essential residues, i.e., the ED unit in the middle of this 18-residue segment. Either way, 11H9 binding may induce more substantial structural rigidifications in the central region of the otherwise flexible N-terminal IDR of CA-IX.

Antibody–IDR interactions with two complementary binding sites may lead to the formation of a 1:1 bimolecular complex, a 1:2 (11H9:IDR) tri-molecular complex or cross-linked multi-molecular oligomers and/or aggregates. However, a maximum of 22 residues separates the first low-affinity epitope D₂₀DLPG₂₅EED from the last high-affinity epitope EDLPG₅₅EED in the ePG-1 sequence (Figure 2c or Table 1), which is insufficient for spanning the two symmetrically structured combining sites, leaving the second combining site of 11H9 unoccupied in a bimolecular 1:1 PG:11H9 complex. It is known experimentally that bivalent binding requires a minimum of four residues to span ~1.8 nm,^27,28 while bivalent binding to an intact mAb is shown to occur at a spacing of ~16 nm,²⁹ which corresponds to a peptide linker of more than 35 residues. As such, PG-11H9 interactions can only involve the formation of a 2:1 stable complex with both combining sites occupied by the high-affinity epitopes of two separate PG molecules (Figure 7a), as is the case for samples with more than 3:1 excess in PG concentrations (SupplFigure 3c and SupplFig. S4A/B). Approaching and below the 2:1 molar ratio, unoccupied combining sites in a bivalent 11H9 mAb are also available for interactions with the secondary epitope presented by another PG-11H9 1:1 complex, in other words, the formation of cross-linked antibody aggregates (Figure 7b), as observed for both the 2:1 and 1:1 complex preparation in terms of sample turbidity and precipitations. Cross-linking interactions of a full-length (bivalent) antibody can also occur on the cell-surface or in SPR studies (commonly referred to as avidity effects) depending on the density,¹⁵ the chemistry of antigen immobilization (Figure 1 and SupplFig. S1), and the conformational flexibility of immobilized antigens. Higher-order molecular assembly, a functional feature of intrinsically disordered proteins with multiple binding sites,³⁰ may therefore be exploited in experiments studying the role of CA-IX’s IDR in maintaining pH homeostasis in the hypoxic/acidic cancer microenvironments.

In general, studying antibody–antigen interactions and specifically epitope mapping represents a critical step toward characterizing and understanding the mechanism of action in antibody discovery and therapeutic development.^31,32 Established methods for epitope mapping³² include structure determination of antibody-antigen complexes by x-ray crystallography, antigen fragmentation using peptides or peptide libraries, site-directed mutagenesis, hydrogen-deuterium exchange mass spectrometry (HDX-MS),^16,33 and to some extent NMR spectroscopy.³⁴ Structure-based methods can be limited,³⁵ often requiring the removal of non-binding sequences/regions, as shown for the M75 study²⁵ and in other systems.^36,37 Epitope mapping efforts also need to recognize the recent identification of a large number of intrinsically unfolded proteins and functionally important intrinsically unfolded protein regions.³⁸ As exemplified by previous work^33,37 and by this study, antibody binding to IDPs/IDRs frequently engage the so-called linear or contiguous epitopes,³² which points to potentially high success rates of peptide (fragment)-based methods including the yeast surface display for identifying high-affinity binding sequences. On the other hand, 3D structure determination of full antibody-antigen complexes may fail, especially when an IDP/IPR in complex with the antibody can remain largely flexible, as shown here by the use of TROSY NMR spectroscopy. In this regard, titration NMR spectroscopy has the potential for locating weaker (secondary) binding sites which may function cooperatively with primary epitopes to impart higher-order interactions such as concentration-dependent antibody cross-linking and aggregation. In all, the data presented here demonstrate that high-resolution NMR experiments in combination with the yeast surface display technology provides a powerful method for a comprehensive understanding of antibody–antigen interactions involving intrinsically unstructured proteins or protein regions that may remain largely unfolded and unstructured even in complex with a structure-stabilizing protein such as an antibody.

Materials and methods

Recombinant antibody and F(Ab) production and purification

The 11H9 and 12H8 anti-human CA-IX mouse mAbs were generated, functionally screened, and sequenced as described.¹⁵ V_H and V_L CDR1–3 regions of 11H9 and 12H8 were grafted on a human IgG1 framework. The 11H9 and 12H8 antibodies, and their respective F(Ab) fragments were produced in CHO-3E7 cells. Briefly, cells were grown to a density of 2 × 10⁶ cell/mL in 150 mL of F17 medium (Catalog No. A1383501, Thermo Fisher Scientific) supplemented with 4 mM L-glutamine and 0.1% Kolliphor P188 in 1.0 L shaker flasks and then transfected with 2 μg total DNA (containing 500 ng each of the heavy and light chain constructs) using PEI MAX^™ (Catalog No. 24765, Polysciences, Inc.), then kept at 37°C for 24 h, fed with Tryptone N1 (Catalog No 19,553, Organotechnie, 1% w/v), and maintained at 32°C for another 6 days. Cell culture supernatant was harvested and analyzed by SDS-PAGE. Antibodies and F(Ab)s were captured from the supernatant using MabSelect Sure (Catalog No. 17543802, Cytiva Life Sciences) or Capto L resin (Catalog No. 17547801, Cytivia Life Sciences), respectively, and eluted with elution buffer (mAbs: 100 mM Citrate, 50 mM NaCl, pH 3.0; F(Ab)s: 0.1 M sodium citrate, pH 3.0); eluates were immediately neutralized with 1 M HEPES. Finally, samples were desalted and subjected to a phosphate-buffered saline (PBS) buffer exchange and to UPLC-SEC to ensure that the F(Ab)s are monomeric and finally stored at −80°C.

Polypeptide production, triple isotope labeling and purification

Unlabeled and ¹³C/¹⁵N-labeled polypeptides: An oligonucleotide encoding a human CA-IX N-terminal fragment polypeptide, designated ePG-1 (Table 1) was codon-optimized, synthesized, and inserted into the pJ401 expression vector (by Atum), while the pJ404 expression vector was used for a variant of ePG-1, ePG-2 (Suppl. Table 1) for recombinant expression in E. coli (Catalog No. 69449, Novagen). Plasmids were electroporated into the BL21 E. coli strain [F – ompT lon hsdSB,rB-,mB-)gal dcm [malB+]K-12(λS)]. One milliliter cultures in LB media were inoculated with these bacterial cell stocks and were incubated for 6 h at 37°C. Cultures were then transferred to 100 mL of filter-sterilized defined media containing M9 salts, 1 mM MgSO₄, 5 mg/L thiamine (T4625-25 G, Sigma-Aldrich), 100 µM CaCl₂, 2 g/L of glucose (or glucose enriched uniformly with the ¹³C isotope, i.e., U-¹³C₆-99%, CLM-1396-25, Cambridge Isotope Laboratories, Inc. for ¹³C-labeling) and 1 g/L of ammonium sulfate (or ¹⁵N-labeled ammonium sulfate, i.e., ¹⁵N₂-99, MLM-713-50, Cambridge Isotope Laboratories, Inc. for ¹⁵N-labeling) as the sole carbon and nitrogen source, respectively; kanamycin (BP906–5, Fisher Scientific) (50 µg/mL) (or ampicillin, AMP201.25, Bioshop, at 100 µg/mL for ePG-2) was used as selection markers. After overnight incubation at 37°C, cell cultures were diluted 10-fold in defined media and incubated at 37°C until A₆₀₀ = 0.8 ~ 1.0. Polypeptide expression was induced by the addition of IPTG (BP1755–10, Fisher Scientific, 1 mM final concentration) and continued incubation for an additional 6 h at 37°C. Cells were then collected by centrifugation at 4,000 rpm for 20 min at 4°C.

To obtain¹³C/¹⁵N/²H-labeled ePG-1, bacteria were adapted to grow in the ¹⁵N/¹³C-labeled defined media (described above). One mL of LB media containing 50 µg/mL kanamycin was inoculated with bacterial cell stocks and incubated at 37°C. After 6 h, the cell cultures were diluted 20-fold in defined media prepared with 100% H₂O and incubated at 37°C until A₆₀₀ ~0.5. Cells were then further diluted fivefold into defined media containing 25% D₂O²H-99.9%, DLM-4-99-5000, Cambridge Isotope Laboratories, Inc.) and incubated at 37°C until an A₆₀₀ of ~0.5 was reached. This process was repeated with defined media containing 50%, 75%, and finally 100% D₂O. Cells harvested from defined media containing 100% D₂O were then aliquoted in a 30% (v/v) sterile glycerol solution and stored in −80°C. For each production of ²H/¹³C/¹⁵N-labeled ePG-1, stored cell stocks were diluted fivefold in 8 mL of 100% D₂O containing ¹⁵N/¹³C-labeled defined media and incubated at 37°C until A₆₀₀ reached ~0.5. This was followed by three additional steps of fivefold dilution to reach a final volume of 1 L cells with A₆₀₀ of ~0.8. One mL of 1 M IPTG was then added, and cultures were incubated for 6 h at 37°C. The extent of deuteration for non-exchangeable carbon-attached (CH) protons in ePG-1 was ~98% determined by comparison of proton NMR spectra of un-deuterated and ²H/¹⁵N/¹³C (or D/¹⁵N/¹³C)-labeled ePG-1 samples.

Polypeptide Purification: Bacterially expressed labeled and non-labeled His-tagged ePG-1 (and ePG-2, see Supplementary Materials) were purified in a two-step procedure, first using immobilized metal-affinity chromatography followed by reversed-phase high-performance liquid chromatography. Briefly, cell pellets were suspended in 100 ~ 150 mL of lysis buffer (50 mM Na₃PO₄, 300 mM NaCl; pH 8) and sonicated on ice for 2 min (10s on, 15s off; power = 6). Cell debris was removed by centrifugation (10,000 rpm, 4°C, 30 min). The cleared supernatant was then mixed with 3 mL of the Ni-NTA resin (30230, QIAGEN) by gentle shaking. After 30 min, the mixture was loaded into a column and washed with 100 mL of the lysis buffer containing 20 mM imidazole. His-tagged ePG-1 (and ePG-2, see Supplementary Materials) was eluted with 15 mL of the lysis buffer containing 250 mM imidazole. The pH of the eluted protein solution was lowered to <4 by adding 2 mL of 10% trifluoroacetic acid (TFA), after which the solution was passed through a 0.45-micron filter before loading onto a reversed-phase C₁₈ column equilibrated with 10% (v/v) aqueous acetonitrile containing 0.1% (v/v) TFA. Purified ePG-1 (and ePG-2) was eluted using a linear gradient of 10%−100% (v/v) aqueous acetonitrile containing 0.1% (v/v) TFA (30 min). Selected peak fractions (UV absorbance at 214 nm) were collected and lyophilized. The protein content of the fractions was confirmed by mass spectroscopy and finally re-dissolved in 450 uL of a mixture of 10% D₂O and 90% NMR buffer (50 mM Na₃PO₄, 50 mM NaCl, 0.2 mM EDTA, and 0.01%-w/v sodium azide; pH 6.8).

Surface plasmon resonance

All SPR assays were carried out at 25°C using a BioRad ProteOn XPR36 (Bio-Rad Laboratories Ltd.). GLC Sensor chips and coupling reagents [10 mM sodium acetate, pH 4.5; sulfo-N-hydroxysuccinimide (S-NHS); 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide hydrochloride (EDC) and ethanolamine] were purchased from BioRad Inc. PBS containing 3.4 mM EDTA, 0.05% Tween, and 500 nM NaCl was used as running buffer, while 0.85% phosphoric acid (18s, 100 μL/min) was used as regeneration buffer.

GLC sensor chip surfaces were activated by injecting S-NHS/EDC (1:10; 140 s, 100 μL/min). Immediately after, 10 µg/mL anti-His antibody (0.2 mg/mL; Qiagen) was injected (25 µL/min) in running buffer until ~400 to 600 RUs were captured. Remaining active groups were quenched by ethanolamine (1 M; 140s injection at 100 µL/min); mock-activated inter-spots were used for blank referencing. The ePG-1 polypeptide (1.38 mg/ml) was then injected (240s, 25 µL/min), after which serial 11H9 and 12H8 mAb and F(Ab) dilutions (1:3 serial dilutions of 100 nM), along with a buffer blank were injected (120s, 50 μL/min), followed by a buffer injection (300s, 50 μL/min). The M75 antibody (M75, Absolute Antibody Inc.), which is known to bind the human CA-IX PG-like domain was used as positive control, while the in-house generated anti-CA-IX c2D7 mAb, which binds CA-IX’s catalytic domain,¹⁵ was used as negative control. Antibody surfaces were regenerated by 0.85% phosphoric acid (18s, 100 μL/min). Sensorgrams were aligned and double-referenced using the buffer blank injection and interspots, analyzed (ProteOn Manager software v3.1) and fitted to the 1:1 binding model to calculate k_a, k_d, and K_D (Table 2).

NMR spectroscopy

All NMR experiments were carried out on Bruker Avance-III 600 MHz and 800 MHz NMR spectrometers equipped with a 5 mm PATXI probe with Z-field gradient accessories. Triple resonance experiments for backbone assignments and ³J_HNHA measurements^39,40 were performed at 298 K and 800 MHz. The ¹⁵N relaxation data, ¹⁵N-R1, ¹⁵N-R2 and ¹⁵N{¹H} NOE⁴¹ of ePG-1 were recorded at 288 K and 600 MHz. 11H9 titration, ePG-1/11H9-F(Ab) complex TROSY, and TROSY assignment data were recorded at 298 K and 800 MHz. NMR samples of ePG-1 were from 0.2 mM to 1 mM in 450 µL NMR buffer (50 mM in NaPi and 50 mM in NaCl at pH 6.8) with 50 µL D₂O. ePG-1 complexes were prepared by adding water-dialysed and dried antibodies or F(Ab) (see below) to the ¹⁵N-labeled and ²H/¹³C/¹⁵N-labeled ePG-1 polypeptide in the NMR buffer, respectively.

Resonance assignments: Seven heteronuclear 3D experiments HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, HNCO, HN(CA)CO,⁴⁰ and H(NCA)NNH⁴² were first carried out with a uniformly ¹⁵N/¹³C-labeled ePG-2 polypeptide (SupplTable S1). Backbone ¹H and ¹⁵N resonance assignments (SupplFig. S2A) of the shorter ePG-2 polypeptide were obtained using six 3D spectra of HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO. All the sequential correlations were verified using a 3D-H(NCA)NNH spectrum that establishes inter-residue connections between those residues separated by Prolines. This H(HCA)NNH spectrum is critical for distinguishing the Gly residue in the first PGEEDL fragment from the following two PGEEDL repeats in the sequence segment PPGEEDLPGEEDLPGEEDL because there are two consecutive Pro residues at the beginning of the first PGEEDL fragment. Based on the sequence-specific resonance assignment of ePG-2 (see Supplementary Materials), backbone resonances of ePG-1 were assigned using HNCA, HN(CO)CA, HNCACB, and CBCA(CO)NH spectra. The TROSY²⁰ spectrum of the ePG-1/11H9-F(aAb) complex was assigned using TROSY-HNCA and TROSY-HNCACB²¹ recorded with the uniformly ²H/¹³C/¹⁵N-labeled ePG-1 in complex with non-labeled 11H9-F(Ab)

¹⁵N NMR relaxation measurements: ¹⁵N-R1, ¹⁵N-R2, and ¹⁵N{¹H} NOE⁴¹ were acquired on a uniformly ¹⁵N-labeled ePG-1 at the lower temperature of 288 K. NMR pulse sequences from the (Bruker) spectrometer library, named hsqct2etf3gpsi, hsqct1etf3gpsi, and hsqcnoef3gpsi, were used for the measurement of ¹⁵N-R1, ¹⁵N-R2, and ¹⁵N{¹H} NOE, respectively. The ¹⁵N-R1 measurements were performed with 10 relaxation delays in the order of 5, 1500, 60, 1200, 140, 950, 220, 700, 330, and 500 ms, and three repetition delays at 60, 700, 1200 ms. The ¹⁵N-R2 measurements were performed with 18 relaxation delays in the order of 115.2, 14.4, 345.6, 28.8, 316.8, 43.2, 288.0, 57.6, 259.2, 72.0, 230.4, 86.4, 210.6, 100.8, 172.8, 115.2, 144.0, and 129.6 ms, and three repetition delays at 14.4, 316.8, and 201.6 ms. In ¹⁵N{¹H} NOE measurements, the reference spectrum was acquired with a relaxation delay of 6s and the ¹H saturated spectrum was recorded with a 3s delay followed by a 3s saturation. Residue-specific assignments of the ¹⁵N relaxation data for ePG-1 at 288 K were obtained by tracing movements with temperature variations of six ¹H/¹⁵N-HSQC spectra recorded at 298 K, 296 K, 294 K, 292 K, 290 K, and 288 K, respectively. The 3D-HNHA spectrum recorded with Bruker pulse sequence named hnhagp3d at 288 K and 600 MHz was used to determine the backbone coupling constants of the ePG-1 polypeptide. All 3D spectra and NMR relaxation data were processed with nmrPipe.⁴³ Spectra were analyzed by using nmrView.⁴⁴ The cross-peak intensities of NMR relaxation measurements were extracted using Non-Linear N-Dimensional Spectral Modeling, a module of NMRPipe.^{43 15}N-R1 and ¹⁵N-R2 values were fitted using in-house developed software programs.

NMR titration experiments: Interactions of the ePG-1 polypeptide with antibodies 12H8 and 11H9 were followed by NMR signal perturbations in the ¹H-¹⁵N HSQC spectrum (at 800 MHz) of ¹⁵N-labeled (including ¹⁵N/¹³C-labeled and ¹⁵N/¹³C/²H-labeled) ePG-1 samples after addition of less than stoichiometric amounts of the respective antibody. Both full-length mAbs 12H8 and 11H9 and their F(Ab) fragments were used for the titration experiments. The F(Ab) titrations proceeded beyond stoichiometric (1:1) ratios to excess for the respective F(Ab) in order to determine the existence of secondary binding sites and/or relative binding affinities in the case of multiple binding epitopes (i.e., for 11H9). For all experiments, the antibodies were dialyzed into pure water followed by lyophilization and reconstitution either in pure water (to ~100–150 mg/mL) or in the phosphate buffer of 50 mM in NaPi and 50 mM in NaCl at pH 6.8 (or NMR buffer). Small volumes of the concentrated antibody solutions were introduced into prepared samples of ePG-1 under the appropriate buffer and pH conditions. For 11H9 mAb titrations specifically, a sample of the ¹⁵N-labeled ePG-1 was first prepared at 1.6 mM in a volume of 100 µL with 60 µL of D₂O and 40 uL of the NMR buffer and the 11H9 mAb solution at 120 µM by dissolving the dried 11H9 powder (see above) in 450 µL of the NMR buffer. Appropriate portions of these two solutions were mixed to produce the various samples of the PG:11H9 complexes with approximate molar ratios of 6:1, 3:1, 2:1, and 1:1, respectively.

Epitope mapping by yeast surface display

The CA-IX fragments covering the PG-like domain were cloned in the pPNL6 vector (provided by the Pacific Northwest National Laboratory, USA), expressed as fusion proteins [Aga2-HA-(CA-IX)-MYC] and covalently displayed on the cell wall of the yeast surface.⁴⁵ The relative amount of displayed fusion protein was measured by probing yeast cells with a chicken polyclonal Anti-c-Myc IgY antibody (ab19233; Abcam), followed by a horseradish peroxidase (HRP)-conjugated goat anti-chicken IgY (H+L) secondary antibody (ab97135, Abcam), and used to normalize the binding signal for anti-hCA-IX mAbs. Binding of the 11H9, 12H8, and M75 mAbs to the yeast cells was assessed using a whole yeast cell ELISA.⁴⁶ Briefly, various clones of the induced yeast cells were loaded into a 96-well filter plate (1×10⁷ cells per well), blocked with PBS containing 1 mg/ml bovine serum albumin, incubated with the anti-hCAIX antibodies, after which the binding of 11H9, 12H8, and M75 mAbs was revealed by using an HRP-conjugated secondary antibody of either goat anti-mouse IgG (ab6789, Abcam) or goat anti-human IgG-Fc (ab97225, Abcam) followed by incubation with the HRP substrate tetramethyl benzidine, according to the manufacturer’s conditions, and read at OD₄₅₀. To determine whether the binding epitopes are linear or structured, yeast cells were in addition heated at 80°C for 30 min to denature displayed hCA-IX fragments, then chilled on ice for 20 min prior to incubation with the antibodies.

Fine epitope mapping: Tiling fragments of 15 amino acids in length with five amino acids overlapping on each end covering the entire CA-IX N-terminal fragment (residues 37–140) were cloned, expressed, and displayed on the yeast surface. The binding of the 11H9, 12H8, and M75 mAbs to these yeast cell-expressed fragments was analyzed by ELISA as described above. Systematic deletion mutagenesis from either the N- or C-terminus of the mAb-binding fragment(s) was applied to delineate the “minimal” binding epitope, and alanine scanning mutagenesis was applied to assess the contribution of each individual amino acid within the minimal epitope to the binding of the mAbs.

List of Abbreviations:

Table

3D	Three-dimensional
CA	Carbonic anhydrase
CHO	Chinese hamster ovary cells
D2O	Deuterium oxide
EDC	1-ethyl-3-(3-dimethylaminporpyl)- carbodiimide hydrochloride
EDTA	Ethylenediaminetetraacetic acid
ELISA	Enzyme-linked immunosorbent assay
HEPES	Hydroxyethyl piperazineethanesulfonic acid
HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, HNCO, HN(CA)CO, and H(NCA)NNH	heteronuclear 3D NMR experiments for backbone resonance assignments of ^15N/^Citation13C-labeled proteins
HNHA	A 3D NMR experiment to measure the three-bond JHN-HA coupling constant
HSQC	Heteronuclear single quantum coherence
IDP	intrinsically disordered protein
IDR	intrinsically disordered protein region
LB	Luria-Bertani broth
mAbs	Monoclonal antibodies
NMR	Nuclear magnetic resonance
^Citation15N-R1	^Citation15N longitudinal relaxation rate
^Citation15N-R2	^Citation15N transverse relaxation rate
^Citation15N{^Citation1H} NOE	Heteronuclear {^Citation1H}^Citation15N nuclear Overhauser effect
PBS	Phosphate-buffered saline
PG	Proteoglycan
SDS-PAGE	Sodium dodecyl sulfate – polyacrylamide gel electrophoresis
S-NHS	Sulfo-N-hydroxysuccinimide
SPR	Surface plasmon resonance
TFA	Trifluoroacetic acid
TROSY	Transverse relaxation optimized spectroscopy
TROSY-HNCACB	TROSY-based 3D NMR experiment for backbone resonance assignments of large proteins with ^15N/^Citation13C/^Citation2H triple labeling
UPLC-SEC	Ultra-high performance liquid chromatography-size exclusion chromatography
YSD	Yeast surface display

Acknowledgments

We thank Mimi Simmons and Maurizio Achione of the Quality Attributes Characterization Team for collecting the size-exclusion chromatography elution profiles of ePG-1, 11H9 and their 2:1 and 1:1 molecular complexes.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2023.2248672.

Additional information

Funding

This work was supported by the Biologics and Biomanufacturing and Monoclonal Antibody Therapeutics Programs of the Human Health Therapeutics Research Centre, National Research Council Canada (NRCC Publication NRC-HHT_53602A).