Candidate epitopes for measurement of hCG and related molecules: the second ISOBM TD-7 workshop

Participants of the Second International Workshop (WS) on human chorionic gonadotropin (hCG) of the International Society of Oncology and Biomarkers Tissue Differentiation 7 (ISOBM TD-7) have characterized in detail a panel of 69 antibodies (Abs) directed against hCG and hCG-related variants that were submitted by eight companies and research groups. Specificities of the Abs were determined using the First WHO International Reference Reagents for six hCG variants, i.e., hCG, hCGn, hCGβ, hCGβn, hCGβcf, and hCGα, which are calibrated in SI units, and hLH. Molecular epitope localizations were assigned to the ISOBM-mAbs by comparing ISOBM-Ab specificity, sandwich compatibility, and mutual inhibition profiles, to those of 17 reference monoclonal (m)Abs of known molecular epitope specificities. It appeared that 48 Abs recognized hCGβ-, 8 hCGα-, and 13 αβ-heterodimer-specific epitopes. Twenty-seven mAbs were of pan hCG specificity, two thereof with no (<0.1 %; epitope β1), 12 with low (<1.0 %; epitopes β2/4), and 13 with high (>>1 %; epitopes β3/5) hLH cross-reactivity. The majority of hCGβ epitopes recognized were located in two major antigenic domains, one on the peptide chain of the tips of β-sheet loops 1 and 3 (epitopes β2–6; 27 mAbs) and the second around the cystine knot (e.g., epitopes β1, β7, and β10; 9 mAbs). Four mAbs recognized epitopes on hCGβcf-only (e.g., epitopes β11 and β13) and six mAbs epitopes on the remote hCGβ-carboxyl-terminal peptide (epitopes β8 and β9 corresponding to amino acids 135–144 and 111–116, respectively). For routine diagnostic measurements, methods are used that either detect hCG-only, hCGβ-only, or hCG together with hCGβ or hCG together with hCGβ and hCGβcf. Sandwich assays that measure hCG plus hCGβ and eventually hCGβcf should recognize the protein backbone of the analytes preferably on an equimolar basis, should not cross-react with hLH and not be susceptible to blunting of signal by nonmeasured variants like hCGβcf. Such assays can be constructed using pairs of mAbs directed against the cystine knot-associated epitope β1 (Asp10, Asp60, and Gln89) in combination with epitopes β2 or β4 located at the top of β-sheet loops 1 + 3 of hCGβ involving aa hCGβ20-25 + 68-77. In summary, the results of the First and Second ISOBM TD-7 WSs on hCG provide the basis for harmonization of specificities and epitopes of mAbs to be used in multifunctional and selective diagnostic hCG methods for different clinical purposes. Electronic supplementary material The online version of this article (doi:10.1007/s13277-013-0994-6) contains supplementary material, which is available to authorized users.


Introduction
Physiology, protein structure, and posttranslational protein backbone variants of hCG The glycoprotein hormone hCG is essential for maintaining pregnancy. Physiologically, it is produced and secreted by the placental trophoblast and pathophysiologically by trophoblastic cancers and by germ cell tumors of the testis and ovary [2].
hCG is a protein heterodimer consisting of hCGα noncovalently linked to the hCGβ subunit. As all glycoprotein hormone (GPH) subunits, hCGα and hCGβ share structural homology with members of the cystine knot growth factor superfamily that includes nerve growth factor, platelet-derived growth factor and transforming growth factor beta [3]. The common structural cystine knot motif consists of two disulfide bridges that link adjacent antiparallel strands of the single peptide chain to form a ring that is axially permeated by a third disulfide bond. This central cystine knot determines the three-dimensional structure of hCGα and hCGβ. On one side of the knot, there are two neighboring hairpin-like peptide loops 1 and 3, which, in hCGβ, are stabilized by a disulfide bond between Cys 23 and Cys72. The single larger loop 2 is located on the opposite side of the knot [3].
The subunits are noncovalently linked in antiparallel, i.e., a head-to-toe fashion, such that loops 1+3 of one subunit are adjacent to loop 2 of the other subunit [3]. Loops 1 and 3 of either subunit and the hCGβ cystine knot, respectively are the most important antigenic regions [1].
The hCGβ genes have developed from an ancestral LHβ gene by gene duplications and mutations [4]. The hCGβ protein is 145 amino acids (aa) in length and encoded by 4 genes and 2 alleles (CGβ6/7, CGβ3/9, CGβ5, and CGβ8), while hLHβ is encoded by a single gene, CGβ4, on chromosome 19q13.3. Thus, hCGβ and hLHβ are highly similar in protein sequence (>85 %) and are immunologically closely related. Furthermore, LH and hCG activate the same receptor. The major structural difference between hCGβ and hLHβ is a carboxyl-terminal peptide extension of hCGβ (hCGβCTP) encompassing aa 113-145. hCGβCTP evolved through a read-through event due to a mutational loss of the stop codon at the genomic level and the incorporation of a hitherto untranslated gene sequence into the coding region [5]. Antibodies recognizing epitopes on hCGβCTP are used in a number of highly specific hCG assays [6]. A single gene on human chromosome 12q21.1-23 encodes the α-subunit, which is 92 aa in length and common to all four human GPHs [7].
hCG is heterogeneous with respect to protein backbone structure and carbohydrate content and is best considered as a complex family of hCG variants occurring in body fluids and tissues. The unambiguous nomenclature for the most important hCG forms of the protein backbone developed by the International Federation of Clinical Chemistry (IFCC) Working Group for Standardization of hCG Determinations is used here ( Table 1 and Fig. 1) [1,8,9].
Glycosylation isoforms hCG subunit folding, assembly, intracellular trafficking, secretion, receptor activation, and half-life in serum is dependent on glycosylation [10]. Both hCG subunits are glycosylated: hCGα contains two N-glycosylation sites at Asn52 and Asn78 that are either mono-, bi-, or triantennary or are sometimes missing. Most N-linked carbohydrate antennae at Asn13 and Asn30 of hCGβ are of the bi-antennary type, but malignancyassociated hCG increasingly carries triantennary carbohydrates at Asn30 and fucosylation at Asn13 (Fig. 2). Four putative O-glycosylation sites are located at Ser121 (core-2), Ser127 (core-1), Ser132 (core-1), and Ser138 (core-1) on the hCGβCTP. The pregnancy associated core-1 glycans on Ser127 and Ser132 are frequently replaced by core-2 glycans in hCG synthesized in early pregnancy and by tumors [11].
Numbers represent positions of amino acid residues in the peptide chain. The metabolic product hCGβcf consists of two peptide fragments that are linked via five disulfide bonds (depicted by S-S), and its N-linked carbohydrate antennae are truncated. Open circles N-linked glycans, filled circles O-linked glycans biologically important domains to the molecular surface of hCG and hCG-related molecules. Several strategies were pursued to resolve epitope distribution and arrangement as well as identification of immunodominant regions. Epitope localization and sharing of epitopes among hCG, hCGvariants, subunits, and related hormones like LH and subunits were determined using molecular chimeras, hCG metabolites, homologous and heterologous glycoprotein hormones and subunits, chemically modified hormones, proteolytic hormone fragments and synthetic peptides, including peptide scanning, and most importantly by site-specific mutagenesis of hCGβ. It is important to mention that in three independent laboratories with different sets of mAbs and analytical techniques similar epitopes and antigenic domains were defined (for reviews, see [1,21]. In previous studies, 26 epitopes on hCG and hCG-related molecules were defined (for reviews, see [1,20,22]). Sixteen epitopes are located on the intact holo hormone hCG (epitopes β 1 -β 5 , β 8 , and α 9 ; α 1 -α 5 ; and c 1 -c 4 ). Seven of these are present on both free and assembled hCGβ (β 1 -β 5 , β 8 , and α 9 ; Fig. 1 and Table 2).
A number of epitopes are of restricted variant specificity. Abs against such epitopes are useful for variant-selective immunoassays designed to measure hCG, hCG+hCGn, hCGβ, hCGβ+hCGβcf, hCGβcf, or hCGα, respectively, in the presence of excess of other hCG protein backbone variants and GPHs.
Some additional Abs recognize epitopes defined only broadly at the molecular level, e.g., additional c-or β-mAbs. Within antigenic domains there seem to be epitopes that remain to be defined more precisely [1].

Algorithm to define hCG epitopes of ISOBM mAbs (INN)
For the Second ISOBM TD-7 WS on hCG, a previously reported epitope mapping algorithm [22], which was also used in similar form in the First ISOBM TD-7 hCG WS [1] has been further refined to reliably characterize the specificities and epitopes of the 69 ISOBM-Abs. This three-step algorithm involves:  [11]). In pregnancy-derived hCGβ, the N-linked carbohydrates are of the biantennary type. O-Glycosylation of hCGβ at Ser 121 always contains a biantennary core-2 and at Ser 138 a core-1 structure with one or two sialic acids. Malignancy-derived hCG and very early pregnancy hCG as compared to middle-to-late pregnancy hCGβ is characterized by increased content of triantennary complex-type N-linked carbohydrates attached to hCGβ Asn 30 and fucosylated carbohydrates attached to Asn 13. "Hyperglycosylated" hCGβ contains an increased proportion of triantennary N-linked carbohydrates (Asn 30); core-2 type O-glycans at Ser 127, Ser 132, and Ser 138; and fucosylated Asn 13-linked glycan. Some glycosylation sites were not glycosylated in some variants (Ser 138, Ser 121, and Asn 13). Immunoassays for hCG-h based on mAb B152 recognize the encircled glycan at Ser 132 and surrounding peptide structure. The major differences in carbohydrate antennae composition between early and mid-to-late pregnancy-and malignancy-derived hCG are depicted in red. Filled square GlcNAc, filled diamond Fuc, empty square GalNAc, empty circle Man, filled circle Gal, empty diamond NeuAc 1. Determination of intraspecies Ab specificities with hCG, hCG-related variants [six First International Reference Reagents (IRR) preparations], hLH and synthetic hCGβCTP peptides to enable grouping of the mAbs according to their main specificities (α-, β-, and c-mAbs), and tentative assignment of epitopes by comparing specificity profiles to those of reference mAbs with known epitope recognition 2. Confirmation of epitope recognition and spatial arrangement of epitopes using sandwich assays with mutual antigen recognition or inhibition by pairs of mAbs to provide information about epitope disparity or identity/vicinity 3. Cross-referencing of the ISOBM-Abs' reaction profiles in specificity and sandwich assays to those of reference mAbs recognizing previously defined epitopes.
This approach frequently enabled definitive assignment of epitopes at the molecular level. In rare cases where no exact molecular localization could be determined due to the lack of appropriate reference mAbs, additional circumstantial evidence, e.g., mutual steric inhibition in simultaneous antigen recognition with mAbs of known epitope localization or recognition of breakdown products like hCGβcf or inter-species cross-reactivity, was used to elucidate the epitope's antigenic domain [1].
The Abs were checked for purity by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), the protein content determined by measuring the absorbance at 280 nm (1 mg/mL=1.

First international reference reagents for hCG and hCG variants
The new international standards for hCG, nicked hCG (hCGn), hCGα, hCGβ, hCGβn, and hCGβcf were purified and characterized by the IFCC Working Group for Standardization of hCG Determinations [9] and adopted by the WHO as the First IRR for hCG and related variants [33]. The material is intended for use in the calibration of immunoassays in substance concentrations, i.e., moles per liter [6]. One milligram each of the six First IRRs for hCG and related molecules were kindly supplied by the NIBSC (Dr. Catharine Sturgeon, CS) to each of the participants and used to characterize the 69 ISOBM-Abs (Table 3). Determination of Ab specificity, affinity, and epitope localization (ABB, NRH) The main specificity profiles of mAbs were determined (1) by direct binding RIA (DB-RIA) with 125 I-labeled hormones and hormone fragments with excess Ab (Online Resources 7,8) and (2) with competitive ligand analysis (CLA), a RIA format, wherein the binding between 125 I-hCG and serial diluted Abs is competed with fixed concentrations of the six First IRRs of hCG and hCG-related molecules and hLH (75/552), respectively (Online Resource 9). Cross-reactivity of the ISOBM-Abs with hLH was determined by titration RIA (NRH) by comparing titers of 125 I-labeled hCG versus 125 I-labeled LH (Online Resource 10). Epitope recognition on the hCGβCTP by ISOBM-Abs was evaluated by competitive RIA with synthetic peptides (NRH; Online Resource  11). Ab affinities were determined by Forster Resonance Energy Transfer (FRET) (ABB; Online Resource 12) and by BIAcore® (NHD; Online Resource 13). For elucidation of the spatial arrangement of epitopes, Ab compatibility in antigen recognition was evaluated by sandwich RIA (NRH; Online Resource 14).

Biochemical characterization of the mAbs (ABB; NRH)
Gel permeation chromatography-high performance chromatography (ABB) The homogeneity of the mAbs determined by GPC ranged from 57 to 99 % with varying degrees of aggregation and low MW contaminants ( Fig. 3a; Online Resource 15). All samples exhibited low MW peaks possibly caused by buffer components (azide, citrate, DTT, etc.).

Sodium dodecyl sulfate polyacrylamide gel electrophoresis (NRH, ABB)
SDS-PAGE analysis was performed under nonreducing (Online Resource 16a; NRH) and reducing conditions (Online Resource 16b; ABB). The purity of the mAbs determined as the proportion of heavy and light chains relative to all protein bands ranged from 84 to 100 % (Online Resource 16b). An example for slight albumin impurity is shown in Fig. 3b. The appearance of double chains could be due to glycosylation differences [34,35], amino acid residue issues [34,35], or the presence of more than one Ab in the sample (Online Resource 16b).
The smearing or absence of bands was probably due to low solubility of these Abs at the low ion strength in IEF.

Isotyping (ABB)
Isotyping was performed on samples displaying double heavy or light chains in SDS-PAGE analysis (Table 4).

Mass spectrometry analysis (ABB)
Apart from the following exceptions most ISOBM-Abs gave expected results. ISOBM-382 had double light chains, and more than one group of heavy chains present following deglycosylation, suggesting presence of two mAbs. ISOBM-385 had two groups of heavy chains due to glycosylation and ISOBM-388 two nonglycosylated light chains due to amino acid residue differences. ISOBM-400 and ISOBM-406 had more than one group of light chains due to differences in , and IEF (c). a Percent purity by GPC-HPLC analysis is 96 % with 7 % aggregation and a small amount (2 %) of low molecular weight (LMW) contaminants. The LMW material could be sodium azide and residual albumin, but this was not confirmed. b Reduced SDS-PAGE analysis shows a combined heavy and light chain purity of 96 %. The first strip has been enhanced to increase the image contrast to highlight a faint band (<1 %) that has a similar molecular weight as the albumin standard (66 kDa). The second strip utilized the auto-scale feature available with Quantity One software (Bio-Rad). A 1 % band is noted at approximately 76 kDa, and a 3 % band is also seen at approximately 47 kDa. c IEF reveals a tight pI range of 5.9-6.0.  I-labeled hCG, hCG-variants, and hLH tracers, the 69 ISOBM-Abs were categorized according to their main specificities (α-, β-, and c-mAbs) (Fig. 4): Antibodies either recognized (a) assembled and/or free hCGα (hCGα-mAbs, n=8; α epitopes were not determined) or (b) assembled and/or free hCGβ or hCGβ metabolites such as hCGβcf (hCGβ-mAbs, n=48; epitopes β 1 -β 13 ), or (c) exclusively the intact±nicked hCGαβ heterodimer, but not the free subunits or metabolic variants thereof (c-mAbs, n=13; epitopes c 1 -c 4 ).
Ab affinities and specificities as determined by FRET (ABB) The ISOBM-Abs were grouped according to their specificity profiles based on affinity for hCG, hCGβ, hCGβcf, and hLH ( Fig. 5) determined by FRET. Affinities for the major hCG variants and hLH, reported as dissociation constants (K d ), ranged from subpicomolar values (0.3 pmol/L for hCGβ of ISOBM-429) to ≥50 nmol/L. The latter value indicated that binding was very weak or not detectable.
Another interesting observation is that all three sheep mAbs, ISOBM-428, ISOBM-429 and ISOBM-430, were in the high affinity group. ISOBM-430 was not tested by the FRET technology but by titration RIA. All four sheep Abs (three mAbs and polyclonal ISOBM-431) were directed against hCGβ loops 1+3 epitope β 5 that is shared by hLH and, therefore, in principle, do not seem suitable for hCG measurement. Nevertheless, ISOBM-429 seems to have tolerably low cross-reactivity with hLH (Figs. 4a and 5 and Online Resource 20), but its suitability for use in hCG+hCGβ variant measurement might still be hampered by preferential recognition of hCGβ. Fig. 4 Specificity profiles of the ISOMB-Abs of the Second TD-7 WS recognizing hCG and hCGβ variants (a), hCG-only and hCGα, respectively (b) were determined by binding of iodinated tracers to excess of Ab (DB-RIA) (NRH). ISOBM-mAbs were classified according to their main specificities and their epitopes recognized on the basis of crossreactivity patterns with hCG, hCG-variants, and hLH: (1) β-mAbs corresponding to epitopes β 1 -β 13 , (2) c-mAbs recognizing epitopes c 1 -c 4 on holo-hCG only, and (3) α-mAbs. a MAbs directed against epitopes β 1 -β 5 are pan-hCG reagents recognizing hCG and hCGβvariants but differ in their cross-reactivity with hLH: β 1 mAbs are highly specific for hCG and show no hLH cross-reactivity (<0.1 %), β 2 and β 4 show very low hLH reactivity (<1 %), whereas β 3 and β 5 strongly cross-react (>>1 %). Epitopes β 6 and β 7 are specific for uncombined hCGβ, hCGβn, and hCGβcf. MAbs against epitope β 8 at the very carboxyl-terminal end of hCGβCTP do not cross-react with hCGβcf and hLH but recognize all other hCG variants except for those lacking the CTP. These mAbs constantly show a low bindable fraction of the tracers as only approximately 50 % of the tracers can be bound specifically. This is in contrast to the β 1 -β 5 mAbs. ISOBM-418 seems to be directed against epitope β 9 as already typed previously in the First WS (ISOBM-280, [1]. Epitopes β 10 -β 13 are specific for hCGβcf as no other hCG variants or hLH are recognized by the respective mAbs. b c-mAbs directed against epitopes determined by the quaternary structure of hCG either do not (c 1 and c 2 ) or do recognize hCGn (c 3 and c 4 ) [56]. The apparent hCGn cross-reactivity of c 1 and c 2 mAbs is due to a cross-contamination of this preparation with non-nicked hCG (approximately 20 %) [1]. The presence of non-nicked hCG and recognition by the ISOBM-mAbs of the two-nicked forms in hCGn were investigated in detail by LC-MS/MS (see accompanying publication by H. Lund). Epitope c 3 (ISOBM-446=INN-hCG-45, reference mAb) is highly specific for hCG+hCGn. ISOBM-mAb 433 that has the same specificity pattern might be directed against a fifth sterically independent c-epitope as shown by sandwich assay. The exact molecular localization of epitope c 4 on hCG is not known, but it is remote from the other c-epitopes. In the First ISOBM TD-7 WS, ISOBM-424 has been characterized (ISOBM-279) and classified as c 4 specific [1]. The α-mAbs have not been investigated in detail as to their epitope recognition. As they readily recognized iodinated tracers (in contrast to α 3 -and α 5 -mAbs), they should be directed against the epitope cluster α 1 /α 2 /α 4 with the exception of ISOBM-404 that is free hCGαspecific and therefore presumably recognizing the subunit assembly region of hCGα (aa hCGα33-42). Minor apparent cross-reactivity with hCGn is owed to a cross-contamination of hCGα in that preparation. hLH crossreactivity of ISOBM-404 might be due to dissociation of highly purified hLH that is observed during testing (PB, personal observation). DB-RIA with 125 I-tracers: results are expressed as maximum specific binding in percent of the "bindable fraction" of added tracer (NRH) [26]. Most Abs, including all directed against the cystine knotassociated epitopes β 1 and β 7 , the hCGβCTP epitope β 8 types 1 and 2, showed moderate affinities (50 pmol/L-5 nmol/L) against their primary target hCG variant. Low affinities could be observed for three reasons: (1) the primary antigenic target hCG variant of the mAb in question was not among the antigens tested, as is the case of mAbs against uncombined hCGα (ISOBM-404), or (2) genuine low affinity to the primary target antigens, e.g., ISOBM-418/280 (hCGβCTP epitope β 9 , aa hCGβ113-116) and ISOBM-443 and 448 against hCGβcf, and (3) FRET labeling affected binding of Abs (ISOBM-445 to hCGβcf; ISOBM-399, ISOBM-436, ISOBM-412, and ISO BM-421 to hLH) (Fig. 5). This is also the case with 125 Ilabeling of epitopes α 3 and α 5 [37].
Ab affinities and specificities as determined by BIAcore® (NHD) The specificity patterns of the ISOBM-Abs were determined based on their affinity for hCG, hCGβ, and hCGβcf in BIAcore®. The affinities (dissociation constants; K d ) ranged from picomolar values (<10 pM for hCG of ISOBM-399) to >10 nM. An affinity of <100 nM was observed for 43 of the Abs for either a single or a combination of the three antigens tested (hCG, hCGβ, and hCGβcf). Five of these Abs, ISOBM-427 (epitope β 2 ), ISOBM-399, ISOBM-423, ISOBM-441 (all three epitope β 3 ), and ISOBM-428 (epitope β 5 ) had affinities of <10pM for the antigen. It is striking that all of these recognize the tops of hCGβ loops 1+3; thus, all epitopes were located within the same antigenic domain. Assignment of epitopes to the ISOBM-Abs was achieved by comparing their specificity profiles to those of reference mAbs (Online Resource 20).
The affinity of many of the ISOBM-Abs appeared to be higher than determined by FRET analysis. This could perhaps be a consequence of having the antigen in a bound form on the BIAcore® chip rather than in a fluid state. The affinity for this immobilized form of antigen may result in an overestimate of affinity.  hCGβ metabolites such as hCGβcf (hCGβ-mAbs, n=24; epitopes β 1 -β 9 ) or (b) exclusively hCG±hCGβ, but not the free subunits (c-mAbs, n=10; epitopes c 1 -c 4 ). Epitope assignment was achieved by comparing profiles of the reference mAbs with ISOBM-Abs (Online Resource 20).
No CLA analysis was possible for 35 ISOBM-Abs, which had very low or no signal, which suggested that the Ab did not recognize 125 I-hCG or could not be competed with the amounts utilized.
Epitope classification by sandwich assays (NRH) IRMAlike sandwich assays were performed to confirm preliminary Ab epitope classifications by specificity assays and to determine epitope localization by comparison with reference mAbs. Characteristic reaction patterns were observed when solid-phase bound ISOBM-mAbs were tested for their ability to sandwich hCG or hCGβ with the panel of reference mAbs directed against epitopes β 1 -β 9 (Fig. 6a) and c 1 -c 4 (Fig. 6b). Patterns observed agreed with previously determined epitope locations for the reference mAbs [28,38].
Compatibility of Ab pairs in sandwich assays indicated that their epitopes were spatially distinct, e.g., epitopes β 1 versus β 2 -β 6 and vice versa (Fig. 6a). Identical or highly similar compatibility patterns of Abs to that of reference or other mAbs indicated recognition of identical or of neighboring epitopes within the same antigenic domain: e.g., the cystine knot-related epitopes β 1 and β 7 , or epitopes β 2 , β 4 , and β 5 on hCGβ loops 1+3. Epitopes within a particular antigenic domain can be easily discerned by cross-reactivity patterns with hCG variants and LH from various species [1]. Thus, although β 1 -and β 7 -mAbs show identical compatibility patterns in sandwich assays and are not compatible with each other (Fig. 6a), they recognize different but spatially adjacent cystine knot-related epitopes reflected by differing variant recognition patterns: β 1 -mAbs recognize a broad spectrum of hCG-variants whereas β 7 -mAbs are highly selective for hCGβ, hCGβn±hCGβcf and do not recognize hCG (Fig. 4).

Antigenic domains and epitope maps of hCG and hCGβ (INN)
Results of the three approaches for epitope typing are summarized in Table 5. In Fig. 7, the ISOBM-mAbs are assigned to the three-dimensional epitope maps of hCGβ (a) and hCG (b), which were established with the reference mAbs previously [1]. The Abs grouped according to epitope recognition are listed in Table 8 in Appendix.
In the ISOBM panel, 48 out of 69 were β-Abs. The major antigenic domain on hCGβ located on the tips of the neighboring β-sheet loops 1 and +3 encompassing aa hCGβ20-25 and 68-77 (epitopes β 2 -β 6 ) was recognized by 27 of the β-Abs. Of these, 12 Abs recognize epitopes β 2 or β 4 (β 2/4 ), 13 epitopes β 3 or β 5 (β 3/5 ), and 2 epitope β 6 . Epitopes β 2 -β 5 Fig. 6 Classification and spatial relationship of ISOBM-mAb epitopes. Two-site IRMA-like sandwich assay experiments with a chessboardlike matrix of antibody pairs tested for their ability to simultaneously bind hCGβ (99/650) for hCGβ-mAbs (a) and hCG (99/688) for holo hCG-mAbs (b) (NRH). Reference Abs for epitopes β 1 -β 9 and c 1 -c 4 served as 125 I-labeled detection reagents, respectively. Reaction profiles of the solid-phase ISOBMii mAbs with the detection reference mAbs were cross-matched to that of solid-phase reference mAbs the molecular epitope specificity of which had previously been defined [1]. Similar reaction profiles were interpreted as epitope identity or neighborhood of mAbs. a The compatibility patterns of pairs of mAbs do not only reveal epitope affiliation of single mAbs but also disclose hCGβ epitope arrangement in larger antigenic domains consisting of one or more epitopes. Abs directed against epitopes located within the same antigenic domain are generally mutually exclusive in hCGβ recognition, whereas those the epitopes of which are located in different domains are compatible. Three major antigenic domains were identified on hCGβ: (1) the domain on the tips of hCGβ loops 1+3 encompassing epitopes β 2 -β 6 (2) the cystine knot associated domain including hCG specific epitope β 1 , hCGβ+hCGβcf specific epitope β 7 , and a structurally related hCGβ-only specific epitope epitope β 14 located on core hCGβ1-112 and characterized by a single mAb, and (3) hCGβCTP epitopes β 8 and β 9 remote from the other domains. MAbs against all hCGβ loops 1+3 associated epitopes β 2 -β 6 are compatible with the hCG-specific cystine knot-associated epitope β 1 and vice versa. Within antigenic domains not all epitopes can be discerned by distinct reaction profiles. As an example, although β 1 and β 7 show identical patterns in sandwich assays and are not compatible with each other, they are definitely recognizing different but adjacent epitopes as β 1 -mAbs are pan-hCGβ-mAbs recognizing a broad spectrum of hCG-variants and in contrast β 7 -mAbs are highly selective for hCGβ+hCGβcf and would not recognize, e.g., hCG (see, e.g., DB-RIA, Fig. 4). A second example are mAbs against epitopes β 4 (ISOBM-419 and ISOBM-445) and β 5 (ISOBM-428, ISOBM-429, ISOBM-430, ISOBM-431, and ISOBM-442) having an identical compatibility profile, i.e., nicely work with mAbs against epitopes β 1 and β 7-9 but not with β 2 -β 6 . These epitopes can be discerned by their variant recognition profiles whereby β 4 mAbs are specific for hCG (≤1 % cross-reactivity with hLH) and β 5 mAbs strongly cross-react with hLH (>>1 %) in titration and competitive RIA (Fig. 4). β 3 -mAbs, although showing a similar reaction pattern as other mAbs directed to hCGβ loops 1+3 associated epitopes (β 2 , β 4 , β 5 , and β 6 ), seems to be remote from the free subunit specific epitope β 6 and not compatible with hCGβCTP113-116 located epitope β 9 at the beginning of hCGβCTP. Such spatial vicinity between the hCGβCTP and hCGβ loop 3 has already been postulated previously [72]. As expected, the epitope of β 8 -mAbs located at the very carboxyl-terminal end of hCGβ (aa hCGβ141-144; [24]) is compatible with all other epitopes. In the first ISOBM TD-7 WS, a new epitope β 14 was observed represented by a single mAb (ISOBM-267) that exclusively recognized core hCGβ [1] and that now appeared compatible with all hCGβ located epitopes except for epitope β 1 , thus seems to be remote from any other core hCGβ epitope. ISOBM-406 according to its sandwich pattern (no compatibility with cystine knot epitopes β 1 and β 7 ) seems to be cystine knot associated. b c-mAbs show variant reaction patterns among themselves. The heterodimeric epitopes c 1 -c 3 are located in the same antigenic domain thus are not compatible with each other. c 4 is clearly remote from that domain as it is compatible with c 1 to c 3 -mAbs. ISOBM-433 recognizes a previously structurally not defined epitope that is highly hCG specific as is ISOBM-446 (epitope c 3 ) (Fig. 4). ISOBM-389 a highly hCG specific c-mAb that according to BIAcore® analyses rapidly dissociates (K d =13E−03), ISOBM-397 and ISOBM-418 (very low affinity in FRET analyses) did not perform well as capture mAbs in this type of assay and were negative throughout (not shown). Reactions classified as positive (mean+2 standard deviations) are depicted as closed squares. Noncompatible mAb pairs are shown as white squares a are pan hCG specific, i.e., present on hCG, hCGn, hCGβ hCGβn, and hCGβcf (Fig. 4a), whereas β 6 is present only on hCGβ, hCGβn, and hCGβcf.
Eight out of 69 mAbs are directed against hCGα. Six recognize assembled and one, ISOBM-404, which has been prepared by immunization with hCGα (Stenman et al., unpublished data), recognizes only free hCGα. The exact molecular localization of the hCGα mAbs was not elucidated (Figs. 4b and 7b).

Epitopes and antibodies
By definition, epitopes are molecular structures dependent on the existence of complementary Abs. Not the entire surface of a glycoprotein like hCG is antigenic. Against certain molecular areas no Abs exist as they are immunologically inert, e.g., due to insufficient T cell help, or sterically not accessible due to protein folding or shielding by glycans. In contrast, other areas representing structurally inherent epitopes, which are characterized by high solvent accessibility and high protrusion indices, are often sites of Ab recognition [40]. hCGβ cystine knot-associated residues Arg10 and Gln89 (epitopes β 1 and β 7 ), hCGα loop 1 residues Pro16, Phe17, and Phe 18 (epitopes α 1 , α 2 ,and α 4 ) and the antigenic domain on hCGβloops 1+3 comprising aa 20-25+68-75 (epitopes β 2 -β 6 ) all bulge away from the molecule forming prominent surfaces that are the major antigenic domains of hCG [3,22,37,[41][42][43]. There is a good chance that irrespective of the immunized species these molecular structures will be recognized as epitopes [38]. For example, the immunodominant antigenic domain on top of hCGβ beta-sheet loops 1 and 3 is recognized by Abs derived from mice and sheep as shown in the present study and interestingly by Abs from humans and rabbits (PB, unpublished observations). Moreover, hLH cross-reactive mAb B206 directed against an epitope within this cluster, presumably epitope β 3/5 , inhibited 40-90 % of the binding of human antisera to hCG [44]. The definition of epitopes by Abs and recognition of the multitudes of possible amino acid combinations within an inherently antigenic structure/domain is dependent on and restricted by the combinatorial repertoire of the VDJ and VJ immunoglobulin heavy and light chains gene segments, respectively, and the cellular capacity to mature the paratope of a given Ab to optimally fit the antigenic surface. This repertoire of Ab specificity varies with individual immune responses, haplotypes, and species. Not every amino acid combination within an antigenic domain will therefore be recognized by Abs of any individual or species. Thus, the repertoire of Ab specificities and corresponding epitopes within an antigenic domain is very large but still somewhat restricted as shown by the present and previous studies. For example, the antigenic domain on hCGβ loops 1+3 is recognized by large panels of Abs that differ slightly in hCG variant recognition, hLH cross-reactivity, affinity, etc. This has been shown to be due to variability in amino acid recognition within the antigenic domain [1].
It is striking that this antigenic region, aa hCGβ20-25+68-75 on the tips of loops 1+3, comprises 16 amino acids, a number that reasonably well corresponds to the surface covered by a single complementary paratope of an Ab whereby two to three amino acids that vary from Ab to Ab provide most of the binding energy and fine specificity [45]. Consequently, dozens of ISOBM-mAbs and Abs of other panels directed against hCGβ loops 1+3 epitopes β 2 -β 5 do not behave uniformly in their recognition of the approximately 15 potential contact amino acids composing discontinuous epitopes, even though they cover more or less the same surface with their paratope [43]. Thus, all differences in affinity, specificity, and hLH cross-reactivity of numerous antibodies directed against this major antigenic region seem to have their basis in variability of preferential recognition of a few amino acids, providing binding energy within very similar or even identical sets of amino acids covered by the Abs' paratopes.
The surface area of an epitope that is covered by a cylinder-like antigen binding site of an Ab is approximately 700 Å 2 in size [38,46], whereby the radius of the antibody   binding domain is 8-10 Å and the radius of the epitope covering area is 15 Å irrespective of Ab specificity [45]. Xray crystallography studies revealed that core hCG, i.e., hCG without hCGβCTP, has a length of 75 Å and a width of 30-35 Å [3, 47] corresponding to a surface area of approximately 8,200 Å 2 . As some regions on assembled hCGβ, such as the stems of β-sheet loops 1+3, are not recognized by any anti-hCG-mAbs [1,18,48], the total epitope-covered area on core hCG could be in the range of 5,000 Å 2 theoretically accommodating simultaneous binding of up to seven Abs to spatially independent epitopes. The minimal spatial requirement for sterical compatibility of two mAbs is that the respective epitopes are approximately 20-30 Å apart. In fact preliminary experiments showed that at least five radiolabeled mAbs against epitopes β 1 +β 3 +α 2 +α 3 +c 4 were able to bind to core hCG simultaneously [38].

Glycosylation and epitopes
With two exceptions, glycosylation has little effect on hCG's immunological make-up, although the glycans, which are hydrophilic in nature and thus surface exposed, represent approximately 30-35 % of its total molecular mass. The exceptions are glycans at the very end of hCGβCTP and in the stem region of hCGβ loop 1. The 14 epitopes on core hCG, which is lacking hCGβCTP, are dependent on the protein backbone. Neither desialylation, deglycosylation [48], partial natural deglycosylation as in the case of the metabolic product hCGβcf [49], nor intense glycosylation as shown with highly acidic pI variants of pregnancy-and tumor-derived hCG have essential effects on Ab recognition by the reference mAbs [17,18]. In addition, the number and the relative spatial location of epitopes do not differ between the isoforms [1,18,48].
The peptidic stem region of assembled hCGβ loop 1, which accommodates the two large N-linked glycans at hCGβAsn13 and Asn30 that are spatially near the hCGα glycan at Asn52 [3], is not recognized by any mAb in the panels of anti-hCG-mAbs of the previous and the present study. Thus, the immune response seems to be attenuated by the N-linked glycans in this region of hCGβ loop 1 [1,18,48].
A mAb (B152) that was not included in this study recognizes hCG with a core-2 O-glycan at Ser 132 and surrounding peptide structures [50,51]. Its epitope, which we termed β 8,3 , is spatially related to epitope β 8,2 that also seems to be influenced by the glycans on Ser 132 and/or Ser 138 [1,29].
Some hCG assays have been claimed to underestimate hCG-h [52]. However, these results have been obtained with an hCG-h preparation that also was completely nicked (C5) [39]. Thus, the results most probably reflected failure to recognize hCGn rather than hyperglycosylated hCG.
The cystine knot-associated antigenic domain includes epitope β 1 involving aa hCGβArg10+Arg60 and possibly Gln89 that sterically are in close proximity to each other [42,43]. hCGβArg10 and Gln89 are unique to hCG and not shared by hLH. This presumably explains why epitope β 1 is highly specific for hCG and its variants and therefore is not Fig. 7 Epitope maps of hCG, hCGβ, and variants (INN) (modified according to [1], with permission) were previously constructed based on the epitopes recognized by the reference mAbs. The identification of reference mAb epitopes was performed by direct binding, competitive and sandwich RIA and ELISA with hormones of various species, hormones subunits, metabolic breakdown products, and synthetic peptides (for reviews, see [1,22]). Furthermore, on the basis of molecular modeling of crystallographic data of hCG and subsequent mutational analyses to assign epitopes to particular amino acids, epitopes of reference mAbs and, by comparison, epitopes of ISOBM-mAbs could be superimposed on the molecular model of hCGβ. a Assignment of ISOBM-mAbs to epitopes on the molecular model of hCGβ/hCGβn/hCGβcf/hCGβCTP. Reaction profiles of the ISOBMii mAbs in specificity and sandwich assays were compared to that of reference mAbs. It appeared that the most immunogenic region of hCGβ is determined by the peptide sequences that correspond to hCGβcf. In particular, the tips of beta-sheet loops 1+3 corresponding to hCGβ20-25+68-77 comprise the major antigenic domain (epitopes β 2 -β 6 ) that is recognized by high affinity mAbs. The only hCG-specific epitope on core hCGβ is β 1 located around the center of the molecule corresponding to part of the cystine knot (aa hCGβ10,60,89). Adjacent to epitope β 1 , the hCGβ/hCGβcf-specific epitope β 7 is also located in this region (aa hCGβ61,89) [43]. Thus, pairs of antibodies against these two epitopes are not compatible in sandwich type assays (Fig. 6a) [24]. hCGβCTP epitopes β 9 and β 8 are located at either end of the hCGβCTP, whereby β 9 might be close to epitope β 3 (Fig. 6a) [72]. b Epitope map of hCG. ISOBM-mAbs were assigned to epitopes on a ribbon representation of the molecular model of hCG [3]. hCGα and epitopes thereon are depicted in blue, hCGβ and its epitopes in green. Conformationally (c) dependent epitopes determined by the quaternary structure of hCG are shown in red. Note the major antigenic clusters of epitopes on the top of beta sheet loops 1 and 3 of hCGα (α 1 / α 2 /α 4 and α 3 /α 5 ) and of hCGβ (β 2 -β 5 ), the central cystine knot-based epitope cluster encompassing highly hCG-specific β 1 and c-epitopes (c 3 ), the latter having a share on loop 2 of hCGβ, that in turn are confluent with the α 1 /α 2 /α 4 epitope cluster. The hCGβCTP epitopes are located on both of its ends at aa hCGβ113-116 (epitope β 9 ) and aa hCGβ133-144 (epitope β 8 ) present on hLH or hLHβ [26]. Due to its superior specificity, it is highly valuable for hCG/hCGβ-variant measurement by immunoassay with no interference by hLH or hLHβ [1].
The assumed location of epitope β 7 on hCGβ, hCGβn, and hCGβcf is based both on mutational analyses and vicinity analysis by sandwich assays: It is associated with the cystine knot, present on hCGβcf, and Asp61 and Gln89 have a role in this epitope. Thus, in sandwich type assays, β 7 -mAbs are not compatible with β 1 -mAbs (Fig. 6a) [1,22,24].
MAbs against the cystine knot epitope β 7 recognize hCGβcf in addition to hCGβ. ISOBM-407 is an exception to this, although other parameters match with epitope β 7 , it shows an exceptionally low cross-reactivity with hCGβcf ( Fig. 4) and thus seems to be suitable for measurement of hCGβ in urine in the presence of high levels of hCGβcf. The assignment of hCGβ specific epitope β 14 to the cystine knot antigenic domain is based on circumstantial evidence as mAb ISOBM-267 defined in the First ISOBM TD-7 WS to recognize epitope β 14 is not compatible with hCGβcystine knot-related epitope β 1 but with all other hCGβ-related epitopes (Fig. 6a). Two hCGβcf epitopes β 10 and β 12 are also cystine knot-associated (PB, unpublished data). An additional cystine knot-related epitope is represented by mAb ISOBM-406.
In summary, β-epitopes located on the protein core hCGβ1-112 are discontinuous in nature, determined by the tertiary protein structure, present on hCGβcf, and arranged in antigenic domains associated with the cystine knot and on the tips of loops 1+3. MAbs directed against these epitopes are of adequate affinity and suitable for immunoassay applications.
When immunizing with the glycoprotein hCG, the vast majority of antibodies will be generated against composite epitopes on hCGα or the core region of hCGβ (aa 1-112) but only rarely against linear peptide sequences of low structural order like the hCGβCTP. MAbs against hCGβCTP are generally of fairly low affinity. Nevertheless, they are used in diagnostic sandwich-type immunoassays as they do not cross-react with hLH (Fig. 4).
hCGα epitopes (α 1 -α 7 ) In the panel of ISOBM-mAbs, 8 of 69 recognize hCGα epitopes. One of these mAbs, ISOBM-404, seems to be specific for free hCGα, and it is speculated that it might recognize the sequence hCGα33-42 on the single loop 2. As no reference hCGα-mAbs (Table 2) were included, a detailed assignment of epitopes was not possible.
MAbs against epitope c 3 are sterically related to epitope c 2 , highly specific versus hLH as well as non-combined intact and modified subunits (<1 % cross-reactivity), not influenced by nicking of assembled hCGβ loop 2, and thus recognize hCGn and hCG equally [1,56] (Fig. 4). They are therefore highly suitable for simultaneous measurement hCG and hCGn ( Table 6).
The exact molecular localization of epitope c 4 has not been resolved yet. It is present on hCGn and hLH, remote from and thus sterically compatible with all other c-epitopes and to a minor extent determined by hCGβ as shown by low cross-reactivity [1,26,38]. A variant of the c 4 -epitope represented by ISOBM-424 (= ISOBM-279, First ISOBM TD-7 WS) that is not shared with hLH (cross-reactivity <0.1 %) seems to exist. A presumably fifth highly specific c-epitope has been observed in sandwich assays wherein mAb ISOBM-433 is compatible with mAbs to c 1 -c 4 (Fig. 6b). Its molecular localization is unknown. ISOBM-389 is a c-mAb that could not be epitope typed but, according to its specificity profile analyzed by LC-MS/MS, might be a c 2 mAb (see accompanying publication by H. Lund).

Method-specific recognition of hCG and hCG variants
Sandwich-type assays measuring hCG alone or in combination with free hCGβ and metabolites are used for detection of pregnancy, pregnancy-related disorders, trophoblastic disease, and various other female and male tumors [2]. Detailed knowledge of the epitopes recognized by the Abs used facilitates development of assays providing better comparability of the results between methods. It has been suggested that assays that are multifunctional with respect to clinical use should (1) recognize in an equimolar fashion hCG and hCGβ protein backbone and glycosylation variants, (2) not cross-react with hLH or derivatives, and (3) not be prone to signal blunting by non-measured variants, e.g., caused by excess hCGβcf leading to false low results [36]. This is a problem when hCG in urine is measured with sandwich assays utilizing a mAb against core hCGβ1-112 in combination with an anti-hCGβCTP mAb [57].
While assays measuring hCG and all hCGβ-related variants are useful as first line methods, for diagnosis of pregnancy and cancer, it is often advantageous to specifically measure only selected variants [2]. Thus, specific hCGβ assays are used for first trimester Down's syndrome screening and also for diagnosis of testicular [58,59] and nontrophoblastic cancers, 20-50 % of which produce only hCGβ but not hCG [60][61][62]. However, the concentrations are mostly low, and the assays used need to be highly sensitive. Assays for hCGβ that are intended for first trimester screening of Down's syndrome need to be insensitive to interferences by an approximately 100-fold excess of hCG and tuned to measure fairly high concentrations. They are therefore of limited utility for the diagnosis of nontrophoblastic cancers.
Elevated plasma concentrations of hCGβ are reflected by high levels of hCGβcf in urine [63], and specific assay of this form has been used for diagnosis of nontrophoblastic cancer [62,64,65] and for the characterization of the First IRR for hCGβcf [9]. However, commercial assays are not available presently.
Candidate epitopes for measurement of hCG and hCGβ Assays specifically recognizing hCG, hCGβ, and related variants can be constructed using a combination of two pan hCGβ mAbs with identical specificity profiles [66], i.e., with one partner directed against epitopes β 2 or β 4 (the hCGβloops 1 and 3 domain) combined with a mAb-recognizing epitope β 1 (the hCGβ cystine knot domain; Table 6).
In the two ISOBM TD-7 WSs, 50 of 96 Abs were shown to recognize hCG+hCGβ and 23 of these did not recognize hLH. Theoretically, any of the five mAbs directed against epitope β 1 around the cystine knot could be combined with any of the 18 mAbs against epitopes β 2 or β 4 on loops 1+3 for construction of multifunctional assays. Epitopes β 1 and β 2/4 are shared by all important hCG and hCGβ protein backbone variants and glycosylation isoforms including hCG-h and hCGβ-h [17,18]. MAbs against these two discrete epitopes are highly specific for hCG with <0.1 and <1 % cross-reaction for hLH for epitopes β 1 and β 2/4 , respectively. No other epitope combination provided assays with equally wide and identical recognition of hCG and hCGβ variants and high specificity versus hLH.
While Abs recognizing these epitopes provide desirable specificity, variable affinity for hCG variants (Fig. 5) may cause nonequimolar recognition of hCG and hCG variants in different methods [6]. Although assay specificity can be predicted on the basis of mAb specificity profiles and epitope recognition [66], ultimate performance can only be evaluated with the final assay. An additional source of method variability in hCG measurement that cannot be fully predicted is that of Ab synergy, which may vary between different Ab pairs [67].
Alternative epitopes for measurement of hCG and/or hCGβ and variants Few manufacturers provide information about the epitope specificities of Abs used in their assays, but due to variable recognition of the First IRR preparations for hCG and variants, it is obvious that different epitope combinations are used in the major commercial assays [6]. In addition to the epitope combination β 1 -β 2/4 , other combinations are possible for the construction of assays for hCG and variants, e.g., epitopes β 8,1 and β 2 , β 1 -α 5 , α 4 -β 2 , etc., but none of them will fulfill all three above-mentioned criteria. However, the frequently used β 8 and β 2 combination does not pose problems as long as serum specimen are measured that do not contain hCGβcf, truncated hCG or truncated hCGβ, or clipped hCGβCTP.
For selective measurement of hCG or free hCGβ or hCGβcf certain epitope combinations can be suggested: for hCG (no recognition of hLH or noncombined subunits), a mAb against epitope c 2 or c 3 can be combined with one against β 2/4 (Table 6). Alternatively, β 1 -α 3 combinations [66] or β 2/4 combined with a tracer mAb against an α epitope are possible [68]. These designs eliminate cross-reactions with free subunits but are sensitive to interferences by free subunits and hCGβcf. For measurement of free hCGβ, a mAb to epitope β 7 or β 14 , and for hCGβcf, a mAb to epitope β 11 can be combined with one to epitope β 2/4 . For hCGα, combinations of mAbs against epitopes α 6 and α 5 are recommended [69] (Table 6).
A unique mAb coded B152 is used for the measurement of hCG-h that carries a core-2 glycan on Ser132 located on hCGβCTP [20]. However, the clinical utility of assays using this mAb remains to be established [70].
Future perspectives: harmonization of hCG and/or hCGβ and variant measurement Considerable reduction in between-method and betweenlaboratory variability in results can be achieved by a number of measures: (1) the establishment and usage of a clear nomenclature of hCG and its variants [1,8]; (2) endorsement of that nomenclature to define what hCG-assay measure [1,8,60]; (3) characterization of diagnostic assays with the new six First IRRs calibrated in SI units that were adopted by WHO for immunoassay standardization [6]; (4) standardization of methods with the highly pure new WHO Fifth IS for hCG encoded 07/36,4 which is identical to the First IRR for hCG 99/688; (5) harmonization of mAb epitopes used in diagnostic methods for hCG, hCGβ, and their variants; and (6) the establishment of reference methods for the various forms of hCG [8], which will be supported by the detailed knowledge on Ab epitope recognition reported in the present study.