MAb WN1 222-5 in context

Monoclonal Antibodies to Endotoxin Core Segregate into Families of Specificity and Cross-Reactivity to Lipopolysaccharides in ELISA by Correlation Cluster Analysis.

G. Robin Barclay et al.

e-mail: r.barclay@virgin.net

Abstract

A series of 291 mouse monoclonal anti-endotoxin antibodies predominantly with specificity for epitopes in the LPS cores in various endotoxins were assayed in ELISA to a large panel (66) of rough and smooth LPS and lipid A from a variety of sources and species. The resultant data was sorted according to relationships between MAbs and between LPS antigens, so that similar MAbs congregated together and similar LPS congregated together. The MAbs segregated into 24 major family groups. The LPS segregated into clusters which reflected structural similarities. Some MAb families displayed reactivity for either Re, Re-Rd, or Rc chemotype LPS and did not react with more complete rough or smooth LPS. Some MAb families were specific for one or more core types from E.coli and Salmonella: within this category some MAbs apparently retained core-type specificity in rough (Ra) and smooth LPS, while others showed core-type cross-reactivities on R-LPS but more restricted specificity on S-LPS. Other MAbs families were identified with specificities for Rc or Re or lipid A minimal binding structures which showed wide cross-reactivity with larger LPS including S-LPS from different species and strains of Gram-negative bacteria. Some of these have potent endotoxin neutralising activity in vitro and in vivo.

return to homepage

Abbreviations

BSA, bovine serum albumin; ELISA, enzyme-linked immunosorbent assay; KDO, 3-deoxy-D-mannooctulosonic acid; LPS, lipopolysaccharide; MAb, monoclonal antibody; R-LPS, rough LPS; S-LPS, smooth-LPS;

Introduction

The endotoxins of Gram-negative bacteria are lipopolysaccharide (LPS) molecules with three distinct domains referred to as lipid A, core oligosaccharide and O-polysaccharide. The lipid A and core oligosaccharide comprise the endotoxin core and are relatively conserved among different Gram-negative bacterial species. The O-polysaccharides show wide structural diversity and give rise to the O-specificity of different strain and species serotypes. Lipid A is the toxic moiety of endotoxin, and is covalently linked to core oligosaccharide in all LPS [1]. Rough (R) bacteria lack the O-polysaccharide of their smooth (S) strain bacteria counterparts, and different rough mutant bacteria have been isolated expressing a range of incomplete core oligosaccharide structures ranging from complete core (Ra) to deep-rough (Re) mutants expressing only lipid A linked to inner-core KDO residues [2,3]. R-LPS from these mutants has been invaluable in analysis of specificity and cross-reactivity of anti-core antibodies.

Cross-reactive LPS-core specific antibodies are found in all healthy adults [4,5,6]. These antibodies approach adult levels by one year old in infants [7]; they become acutely depressed in life-threatening Gram-negative sepsis [8,9,10]; and they are elevated following immunogenic natural exposure to endotoxin [11,12]. The hypothesis that such antibodies provide an important host defence [13] has stimulated the pursuit of polyclonal [14,15] and monoclonal [16,17] antibodies with these cross-reactive endotoxin-neutralising properties for immunotherapeutic use, with disappointing results [15,18]. Many of these studies have employed lipid A or R-LPS (or R-mutant bacteria) as immunogens, in an effort to employ only the most structurally conserved regions of LPS to elicit cross-reactive antibodies and avoid elicitation of serovar specific antibodies. However, the LPS core region has proved immunochemically more complex than was hypothesised, and the MAbs which promised new therapeutic widely reactive anti-endotoxin reagents have shown, in general, exquisite but restricted specificity ranges [19].

Rough bacteria are serum-sensitive. Only complete-core (Ra) R-LPS, co-expressed with S-LPS on smooth bacteria, occurs commonly with S-LPS in the host, contributing to pathogenesis and acting as immunogens eliciting natural polyclonal anti-endotoxin antibodies. Large numbers of MAbs have been reported which appear to react predominantly with lipid-A-distal hydrophilic moieties which, in many cases, represents the exposed end of the oligosaccharide chain from the particular R-LPS chemotype employed for immunisation [19]. In incomplete-core R-LPS and lipid A, such moieties represent breaks in the complete core structure, and are therefore novel antigens when compared to the natural LPS primary structures seen by the host. Such MAbs fail to react convincingly with natural, clinically important complete LPS and show little or no endotoxin neutralising activity with these LPS. It is also probable that the incomplete core R-LPS express different secondary and tertiary structures from complete core R-LPS and the core of S-LPS, affecting shape and aggregation properties which may contribute to antigenic structures [20].

We have succeeded in producing widely cross-reactive, LPS-core-specific, endotoxin-neutralising MAbs by employing immunisation procedures which mimic natural exposure to endotoxins [21]. We predominantly used complete core (Ra chemotype) bacteria or LPS, and cyclical immunisations were carried out to expose animals to different LPS-core serovars, in an attempt to boost cross-reactive immunity to the common structures shared by the different immunogens. Primary hybridoma screens employed multiple ELISA analyses with different individual LPS and LPS cocktails, to preferentially select multiply-reactive MAbs. The study was carried out as a collaborative exercise in two geographically distant centres in Edinburgh and Basle, over four years, from which 291 MAb-secreting hybridoma lines were retained, some of whose properties have been reported [21, 22]. Although it has been possible to examine the specificity of some of these MAbs in detail, a comprehensive and systematic analysis of this bank of anti-LPS MAbs has proved intractable because of the number of MAbs involved, and because of the number of LPS antigens against which they were tested in ELISA, for which only minimal or inferred structural information was available concurrently. We now report a method we have devised for evaluating relationships in a complex ELISA system, which simultaneously permits systematic analysis of MAb:MAb interrelationships and antigen:antigen (LPS:LPS) serological interrelationships. The method is based on correlation between ELISA reactivities, and congregates MAbs (or LPS) of similar reactivities together into serologically related families without any prior knowledge of MAb specificity or presumption of LPS structure. The reaction patterns of MAb families and LPS families can then be retrospectively analysed systematically, for assignation of presumptive MAb family LPS specificity, and assignation of presumptive LPS family serological properties with reference to published structure information.

Methods

Bacterial strains, LPS and lipid A.

The 66 LPS used in ELISA are shown in Table 1. Where indicated, including lipid A, these were purchased from List Biologicals, CA, USA (List) or Sigma Chemicals, Poole, England (Sigma). The remaining LPS were produced from the respective bacterial strains in the Department of Medical Microbiology, University of Edinburgh, as previously reported [5]. Smooth E.coli of known O-serotype were obtained from Alan S Cross (Walter Reed Army Institute of Research, Washington DC, USA) and included 2 smooth variants of E.coli O18 (Bort) comprised of the original K1 (K1+) encapsulated strain and its unencapsulated (K1-) mutant: a complete core rough mutant of E.coli O18 K1+ (Ra ex O18) was also included, as previously described [21]. Examples of rough E.coli core type strains R1 (HF4704), R2 (EH100), R3 (F673) and R4 (F2513), marked (k), were obtained from Nils Carlin (Karolinska Institute, Stockholm, Sweden), as was the rough E.coli C62. Further examples of rough E.coli core type strains R1 (F470), R2 (F576), R3 (F653) R4 (F2513), K-12 (W3100) and Re (F515), marked (b), were obtained from Helmut Bräde (Forschungsinstitut Borstel, Germany), as was smooth E.coli O4. Salmonella typhimurium rough mutants Ra (1542), Rb (119), Rc (878), Rd (1032) and Re (1102) were obtained from Ian Sutherland (Dept of Microbiology, University of Edinburgh). Pseudomonas aeruginosa rough (Rc) (PAC605) was obtained from Pauline Meadow (University College, London). Klebsiella aerogenes Rb (M10b) was isolated by Ian Poxton.

Heat-killed or formalin-killed bacteria used in immunisations in Edinburgh and Basle were prepared in the Department of Medical Microbiology, University of Edinburgh.

Monoclonal Antibodies.

Monoclonal antibodies were generated by standard methods. The particular details of methods used in the Basle [21] and Edinburgh [23] laboratories have been previously described. A variety of immunisation procedures were employed. In most cases heat-killed bacteria were used. In many cases cyclical immunisations were carried out (see introduction) with complete-core (Ra) bacteria of different core structure. In other cases, different R-LPS variants were used (Ra to Re), and in some cases R-LPS from different species were used, in cyclical immunisations. Other immunisations, e.g. with purified R-LPS or with lipid A coated on mouse red cells, were carried out. In general, mice received immunisations at 4-week intervals, and tail-bleeds were taken at weekly intervals to investigate development of multiple LPS specificity by ELISA to selected LPS panels. Final immunisations of selected immune mice prior to spleen recovery varied from fusion to fusion, but were generally cocktails of the different immunising bacteria when cyclical immunisations were employed.

The ELISAs for hybridoma screening have been previously described [21, 23]. The Basle ELISA employed purified LPS or killed bacteria on PVC microtitre plates [21], while the Edinburgh ELISA employed polymyxin B-complexed purified LPS on polystyrene microtitre plates [4,5]. Primary screens of hybridoma supernatants employed multiple analyses on selected single or cocktail LPS, which varied from fusion to fusion. In most cases only hybridomas whose supernatants showed multiple or unexpected LPS reactivity patterns were retained for cloning and extended specificity analyses. Retained clones were grown beyond maximum supportable cell density for harvesting supernatants for detailed ELISA analysis. Samples of supernatants from all 291 retained cloned hybridoma cell lines were collected in Edinburgh for standardised ELISA assay, and stored at -40ºC until tested.

ELISA.

The standardised ELISA used to obtain the results presented here employed purified LPS complexed with polymyxin B and coated on 8-well polystyrene microtitre plate strips as previously described [5]. Batches of strips were coated with LPS-polymyxin, blocked with BSA, and stored at -40ºC until required. Strips were assembled in microtitre strip-plate frames to give 11 different LPS antigens in each frame: the last (12th) strip in each frame was coated with BSA only and was used as a negative control. The full ELISA required a set of 6 plates. Any empty spaces in the last (6th) frame were filled with uncoated strips. Initially, 59 different LPS antigens were used (Table 1); later a further 7 LPS antigens, E.coli R1, R2, R3, R4, Re (F515) and O4 (Borstel strains) and E.coli Ra (ex O18), were added to fill the last (6th) plate of the set. The first 162 MAbs were tested in the 59 LPS ELISA, and the last 129 MAbs (fusions SZ-43, SZ-44, J5V, WN, VB, VN, X1SW) were tested in the full 66 LPS ELISA.

MAbs (hybridoma supernatants) were diluted 1:20 in phosphate-buffered saline pH 7.2 (Dulbecco A, Sigma) containing 0.01% (v/v) Tween 20, 0.05% sodium azide and 4% polyethylene glycol (PEG 2000, Sigma) and dispensed in 100m l aliquots in single rows across the 6 plate set. 8 MAbs were used in rows A to H respectively of each set of 6 plates, so that MAbs reacted only with a single well of each LPS-coated strip. Plates were incubated overnight (16h) at room temperature to allow maximal binding of weak antibodies, washed, and 100m l of a 1:200 dilution of urease-conjugated sheep anti-mouse Ig (IgG, IgM, IgA reactive) (SeraLab, England) was added to each well. Plates were incubated for 90min at 37°C, washed, rinsed in distilled water, and 100m l of urease substrate (SeraLab, England) was added. Colour was developed over 90 min at room temperature and changed from yellow (no reaction) to deep purple (strong positive), and was stopped by addition of thimerosal solution. Plates were read on a Multiskan MC (Labsystems, Finland) reader at 590nm, and results were collected by output from the reader to an IBM-compatible PC for further processing.

ELISA results were processed by a QuickBASIC (Microsoft) programme written in the laboratory which subtracted optical density (OD) readings of the negative control (BSA coated well) from each test (LPS coated well), and formatted results for transfer to a database (DBase IV, Borland) as net OD, where each record was comprised of a field for the antibody identifier and fields for each of the 66 LPS antigens. Results for each MAb were stored as sequential records, with blank fields to identify any LPS not assayed for a given MAb.

Correlation cluster segregation analysis and visualisation.

A program was written in QuickBASIC (Microsoft) to directly extract the stored data from the database, count and store the MAb identifiers (291) and LPS identifiers (66) as arrays, and assemble the OD results in a matrix (291x66). The correlation (Pearson's correlation coefficient, r) of pairs of MAbs over the set of 66 LPS antigens was determined for all MAb pairings and assembled in a 291x291 matrix. Similarly, the correlations of all pairings of LPS were determined over the 291 MAbs and assembled in a 66x66 matrix. Correlations excluded results where any pair combination included blank result fields, so that for certain MAbs correlations were calculated only on the 59 LPS antigens of the original ELISA set, and the 7 late-added LPS were only correlated over the later tested MAbs. Although their resolution is lessened by this, these results have been included for completeness.

Each set of MAb correlations was rearranged in its matrix as follows. Starting with the origin at the top, left position (row-1, column-1) (MAb-1 correlation with itself, r=1.0), the remaining MAb correlations to the right in row 1 were arranged in order of descending correlation, so that those most like MAb-1 were closest to it (high r) and those least like it were furthest from it (low r). The new order of MAbs in row 1 was noted, and the remaining whole rows (and the MAb identifiers) were rearranged so that column 1 reflected row 1 in order. The origin was then moved to row-2, column-2, and the remaining correlations to the right and below this position were rearranged as before, but correlations to the left and above were not rearranged. The process was repeated moving the origin stepwise down the diagonal from top left (1,1) to bottom right (291,291), rearranging correlations below and right of the origin but leaving arrangements of correlations above and left of the origin. This resulted into segregation along the diagonal of discrete clusters of "similar" MAbs.

The same process was applied to the matrix of LPS antigen correlations, from top left (1,1) to bottom right (66,66), resulting in clustering of serologically "similar" LPS.

Finally, the OD results matrix can be rearranged directly and/or a database can be constructed with the fields of each record arranged in order of the 66 rearranged LPS antigens and ELISA results re-read in according to the new order of rearranged MAbs, to complete the correlation-cluster segregation sort.

It is useful to construct maps of the rearranged correlations and ELISA (OD) results to visualise the relationships which have emerged. Both correlations and OD are continuous data, but can be digitised or banded into discrete subdivisions, and represented by colours (Figure 3) or shades of grey (Figures 4 and 5) for graphic representation. In this case, these were translated into a bitmaps which can be imported by various widely-used computer graphics software packages. In the figures shown the data was digitised and banded by a QuickBASIC programme written for this purpose, the bitmaps were pre-processed in Picture Publisher 3.1 (Micrografx), and the final figures were composed in Corel Draw 4 (Corel).

Results

MAb and LPS correlations.

As an example of MAb correlation relationships, the pairing of MAb SZ-043-27.07.03 with two other MAbs from the same fusion are shown in Figure 1 as a two-dimensional scatter of ELISA reactivities with different LPS, and the resultant correlations (r) and regression lines. The highest correlation of MAb SZ-043-27.07.03 is with MAb SZ-043-27.11.02 (r=0.85), and the ELISA reactivities are distributed around the regression line: these MAbs appear similar, and may be subclones of the same hybridoma since they arose from the same well (well 27) of the primary hybridoma screen. A high but lesser correlation (r=0.63) was obtained between MAb SZ-043-27.07.03 and MAb SZ-043-5.01.04: these antibodies are evidently different, both reacting strongly with a sub-group of the LPS, but MAb SZ-043-27.07.03 reacting more strongly with most other LPS than MAb SZ-043-5.01.04.

As an example of LPS correlation relationships, the pairing of LPS from E.coli O18 (K1-encapsulated, Bort) is shown with the LPS of E.coli R1 (k) and Serratia marcescens in Figure 2. The best correlation is between E.coli O18(K1+) LPS and the LPS of E.coli R1 (k) (r=0.81), where a group of MAbs react strongly with both LPS (possibly R1 core-type specific, indicating E.coli O18 expresses this core type). There is much less correlation between E.coli O18(K1+) LPS and the LPS of Serratia marcescens (r=0.41). Many more MAbs react strongly with E.coli O18(K1+) LPS than with the LPS of Serratia marcescens, while the other reactive MAbs vary in their degrees of reactivity with these LPS.

Correlation cluster segregation and ELISA rearrangement.

The maps of the rearranged MAb (C) and LPS (A) correlations and their relationship to the rearranged ELISA results (B) are shown.

The identities of the LPS are listed in rearranged order in Table 1, and the identities of the MAbs are listed in rearranged order in Table 2 (a and b). These may be related to the larger scale map of the ELISA results shown in Figure 4, which includes scales which relate to the sort numbers in these tables. There is clear congregation of LPS and MAbs into clusters along the identity diagonal which relate to ELISA reactivity patterns (Fig 3).

Relationships are also evident between families off the diagonal. These relationships may be positive (high correlation) where families are similar, or negative (high negative correlation) where families oppose in reactivity patterns, e.g. MAbs reacting predominantly with E.coli R2 oppose MAbs reacting predominantly with E.coli R1 when several examples of each LPS (rough and smooth) are present in the ELISA. The resolution of either LPS or MAbs into families is dependent on the variety of either in the ELISA: if a particular reactivity is not present the characteristics which determine it will not be resolved.

MAb WN1 222-5 is MAb 107

(Table: 1)

(Table: 2a)

(Table: 2b)

MAb family specificities for LPS core.

The major MAb families numerically and/or in ELISA reactivity pattern are listed in Table 2 (a and b) and their positions are marked on the sorted MAb correlation map (Figure 5). The principal characteristics of these families are as follows.

Family a:

Predominantly from fusion ES-184, these MAbs are E.coli O18 LPS serotype specific. All react with E.coli O18 K1+ (Bort) LPS, but some fail to react with its unencapsulated mutant: many also react with Serratia marcescens LPS. Some MAbs react with E.coli O18 (K1+) LPS, its rough mutant LPS (E.coli Ra exO18), and LPS from E.coli R1 (F470) and S.typhimurium Ra (strain ST1542) LPS. In general, these MAbs appear to have restricted specificity for this series of LPS.

Family b:

We believe these MAbs to be specific for the E.coli R1 core and cross-reactive on LPS expressing this core type: the reaction of some of them with S.typhimurium Ra (strain ST1542) LPS is unexplained. The smooth E.coli LPS with which they react are serotypes O16, O75, O2, K235, O4, O18 (K1+), and O128:B12.

Family c:

These antibodies appear to recognise Re chemotype LPS and show wide cross-reactivity with larger LPS, especially the "R1" type LPS seen by Family b MAbs. As well as E.coli and Salmonella LPS they bind to P.aeruginosa, Klebsiella, Vibrio cholera, Shigella and Serratia LPS. They bind weakly to monophosphoryl lipid A (ex S.minnesota Re) but not to diphosphoryl lipid A (ex E.coli Re D31m4)

Family d:

These are specific for the LPS from the rough mutant of E.coli O18(K1+).

Family e:

These MAbs recognise the Rc LPS chemotype, but not Rd2 or smaller LPS. Some, such as WN1 222.5, are widely cross-reactive with smooth E.coli, Salmonella and Shigella: others, such as SZ-43 27.11.2, are less cross-reactive and show strong reactivity predominantly with smooth LPS of E.coli "R1" type and some others.

Family f:

We have no interpretation of the binding pattern of these MAbs. They react with some Ra or Rb LPS from E.coli and Salmonella, and are not very cross-reactive. They are not strongly reactive in ELISA with any LPS.

Family g:

These MAbs are from a large family which shows a narrow range of binding activity for E.coli R2, R4 and C62 rough LPS and in some cases also for E.coli O127:B8 and O12 smooth LPS. Some also react with one strain of S.typhimurium Ra and smooth S.typhosa LPS.

Family h:

These are Re chemotype reactive MAbs which do not in general react with larger LPS, with the exception of E.coli C62 and S.typhimurium Ra strain ST1542 LPS.

Family i:

This group of MAbs appear to bind to E.coli R2 and R4 cores, with weak binding to R3 and K-12 cores. They also bind strongly to S.typhimurium Ra and moderately to S.minnesota Ra and S.typhimurium Rb. They bind to a range of smooth E.coli and all smooth Salmonella, which must share core serology.

Family j:

These MAbs are like Family i, but lack reactivity with E.coli O128, O127, O12, 55 and O26. Some MAbs react with Serratia marcescens and the unencapsulated E.coli O18(K1-) smooth LPS. While retaining Salmonella and E.coli R2 and R4 rough LPS reactivity, they predominantly react only with smooth LPS from Salmonella.

Family k:

These are very similar to Family b MAbs (E.coli R1 core type specific with smooth LPS cross-reactivity), but also have a weak reactivity with most LPS, including lipid A.

Family l:

These are restricted to E.coli R1 rough LPS and a small number of E.coli smooth LPS, especially E.coli O75.

Family m:

These very weak broadly cross-reactive MAbs may have some lipid A or Re LPS specificity.

Family n:

These show an unusual, restricted specificity for E.coli O127, O55, O26, O111 smooth LPS and S.minnesota O smooth LPS, with other minor weak binding activities.

Family o:

These are like Family n but show more extensive cross-reactivity, including reactivity with S.enteritidis and S.typhosa, and E.coli O15, O128, O127, O12 and O18(K-). They are not strongly reactive with rough LPS.

Family p:

These are rough Pseudomonas aeruginosa PAC605 specific.

Family q:

These MAbs are lipid A specific and broadly cross-reactive with all LPS (the missing reactivities in Figure 8 are LPS to which they were not tested in ELISA).

Family r:

These MAbs react with E.coli R2 and R3 rough LPS but may be E.coli R3 core type specific in smooth LPS, reacting with E.coli O86, O15, O26, O55, O111 and Shigella flexneri O(1A) smooth LPS.

Family s:

These overlap Family r specificity, but do not react with one of two E.coli O111 LPS. They also react with E.coli O128 LPS, and some react moderately with some Salmonella smooth LPS.

Family t:

These are Rc-reactive, broadly cross-reactive LPS like Family e, and may have segregated differently from Family e because they were not tested on the LPS added later to the ELISA series. They may also show some R3 core predominance, whereas Family e may show some R1 core predominance.

Family u:

These are weakly reactive MAbs with some Ra and S LPS reactivity for E.coli LPS which has no clear pattern.

Family v:

These are lipid A reactive MAbs with broad LPS cross-reactivity. They differ from Family q in showing better binding to rough LPS than to smooth LPS.

Family w: