- Email: [email protected]
Characteristics of binding assays – specificity
CHAPTER 9
Characteristics of binding assays - specificity
9.1. Definition of‘specijicity The specificity of an assay can be defined as ‘thg degree to which an assay responds to substances other than that for which the assay was designed’. ‘Degree’ in this case is a relative term since it will vary considerably with the conditions of the assay - concentration of tracer and binder, separation procedure, etc. For example, in a radioimmunoassay it is possible to design conditions in which at one concentration of antibody there is no significant interference by a cross-reacting material, while at another concentration the interference is such as to render the assay valueless for all practical purposes. The subject of specificity will be discussed under two separate headings, which for convenience and familiarity will be termed ‘specific non-specificity’ and ‘non-specific non-specificity’.The former refers to interference by identifiable materials which are physicochemically similar to the ligand and may thus react directly with the binder. The latter refers to interference by materials which do not directly react with the binder, but can nevertheless affect the primary binder-ligand reaction - for example, acid conditions which produce partial or total inhibition of any antigen-antibody reaction. The two types of non-specificity may be difficult to distinguish, yet in operative terms it is highly important to do so because the means for their elimination may be quite different.
463
464
RADIOIMMUNOASSAY AND
K E L A I F D TFCHNIQUES
9.2. Speci/i’c non-specificity There are many groups of biological materials which are physicochemically very similar - for example the various steroid hormones, or the glycoprotein hormones of the anterior pituitary gland (LH, FSH, TSH). The specificity of their physiological action depends on target organ receptors capable of distinguishing between them on the basis of small differences in the molecule such as the presence or absence of qn hydroxyl group. The use of these receptors in binding assays confers similar specificity on in vitro measurement. However, the use of receptor assays is limited to a relatively small number of substances -clearly they cannot be applied to non-hormonal materials with no target organ specificity, and equally they are of little value in determining important metabolites of biological compounds. For most purposes other types of binder have to be used - chiefly the naturally occurring circulating binding proteins and antibodies. Much of the discussion which follows will be devoted to the latter. 9.2.1. The basis of specific non-specificily
An antiserum to a given material will usually react, albeit less effectively, with closely related materials. There are 3 basic reasons why this may occur: first, a single homogeneous antibody population may react with a range of related ligands, each reaction having a different affinity constant; second, the antiserum may contain populations of antibody molecules one or more of which is directed to a site on the primary material which also occurs on a related material ; finally, the antiserum may contain antibodies directed to contaminants present in the ligand and which are also present in the related material (Fig. 9.1). Examples of these situations will be found in $9.2.2 and 9.2.3. These concepts are important to an understanding of the experimental finding of ‘parallelism’ or ‘non-parallelism’. If the difference between two related materials is solely in their K value with respect to a single homogeneous binder, then standard curves for these materials using a homologous tracer will be parallel. However,
Ch. 9
465
SPECIFICllY
A
B
C
Fig. 9.1. Different types of specific non-specificity. (A) A single population of antibody molecules reacts with two related materials. hut with a different affinity constant (i.e. one molecule is a good .fit’ to the combining site, the other is a less good fit). (B) A single population of antibody molecules reacts with an antigenic site which is common to two different molcculcs (e.g. the subunit of the glycoprotein hormones). (C) The antiserum contains populations of antihodies to different ligands; the non-specific ligand may then cross-react in the assay.
deviation from parallelism will be seen if the tracer is not homologous and if the mass of tracer represents a significant fraction of the total ligand in a given tube. This type of non-parallelism is found only at the lowest concentrations of unlabelled ligand and would be barely apparent from simple inspection of a standard curve. The second and third types of non-specificity - due to heterogeneity of binder or binding sites - can yield more dramatic non-parallelism. The principle by which this can influence the shape of a standard curve is best illustrated by an extreme case (Fig. 9.2). Clearly, the precise result obtained will vary with the set of reagents used and, in the case of radioimmunoassay, the distribution of antibody populations will vary considerably between different antisera. One might yield apparent parallelism between two cross-reacting materials, while another would yield non-parallelism. In a real system the 3 types of specific non-specificity may often co-exist. The analysis of the main factor involved - heterogeneity of K value at a single site or heterogeneity of sites - can then become very complex. However, gross non-parallelism, particularly at higher concentrations af unlabelled ligand. suggests site heterogeneity. This is of practical importance because the identification of this effect suggests that specificity may be improved by absorption of the antiserum. S,,/7,‘<1,,,
466
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
40
%TRACER BOUND
M
I
A-6
\
20
A B
Primary material W Tracer ,A.B
10
.cog
,032 .la .5 CONCEWRATION OF STANDARD
2
Cross-reading material Ant lbodles
B, A A
A
B
Fig. 9.2. Non-parallelism due to heterogeneity. The primary material and the tracer have two antigenic sites, A and B. The antiserum contains 2 corresponding populations of antibodies. Material containing site B only will produce the curve shown. It can never produce complete inhibition in the assay because antibodies of population A will always be unoccupied and can still bind the tracer.
9.2.2. Assessment o f spec fic non-specficity To test cross-reaction serial dilutions of the material in question are prepared and assayed: the resulting curve is then compared with that given by the standard material for which the assay was designed. The most common way of presenting the result is to compare the amount of the material under study which yields 50% inhibition of binding with the amount of standard giving the same inhibition, and then to express the potency of the material as a percentage of that of the standard (Fig. 9.3). For example, if the respective concentrations are 1 and 100, the potency is stated as 1%. A number of problems can arise with this approach: Concentration by weight versus molar concentration : very often potency is calculated on the basis of relative concentrations by weight. If the molecular weights of standard and cross-reacting material are identical this yields a satisfactory result. If they are not then the answer
Ch. 9
461
SPECIFICI IY
Cross - reaction 25%
100
80
% tracer bound
1
2
4 8 16 Concentrationof Standard
32
64
Fig. 9.3. Estimation of the percentage cross-reaction from the amounts of material required to produce inhibition in the assay. The curve on the left is for the standard. that on the right for the cross-reacting material. Note that this procedure cannot be applicd if the curves are non-parallel.
may be misleading. For example, if the molecular weight of the crossreacting material is 10 times that of the standard, and the potency by weight is given as lo%, then in reality the two materials are equal on a molar basis. To avoid this pitfall calculation as molar concentrations is always to be recommended. Variation with conditions of’ assay: the apparent potency of a cross-reacting material may vary if the conditions of the assay are changed. For example, if an assay is de-sensitised by increase of binder concentration, then a population of non-specific binding sites may become apparent which were of no significance at lower concentrations. More important, specificity can show great variation according to the nature of the tracer employed: an example of this is Sublrrr rndrx p 531
468
KADIOIMMUNOASSAY A N D K E L A l f D 1 ECHNIQUtS
given in 9: 2.2 and Fig. 2.1A, B. Similar though less obvious effects may result from the substitution of highly purified tracer ligand with less purified material. Calculation if' curves are non-parallel: if the curves yielded by standard and cross-reacting material are non-parallel, then clearly a calculation of potency made at 50% inhibition will be very different from that made at 10% (see Fig. 9.2). There is no simple answer to this problem, and non-parallel cross-reaction must be judged in relation to the intended biological or clinical use of the assay. Ariing from this is the question of how to judge parallelism. Though usually a subjective impression, this can be highly misleading: substantial differences can easily be missed, particularly at the extremes of the curve. There are two approaches to this problem. The first is to perform a logit transformation of the results (Q 7.2) and then to make a statistical comparison of the 2 linear regressions. The second is to calculate the apparent potency of each dilution, and then to examine the figures for any systematic deviation between standard and crossreacting material. In judging non-parallelism it is also important to exclude experimental artefact, such as that which can arise as a result of serial dilutions of the standard; indeed comparison of serial dilutions of a material with independently prepared dilutions of the same material is almost certain to yield some degree of nonparallelism (see Q 2.5). Relevance to physiological and clinical situations : the eventual significance of a cross-reacting material lies in the extent to which it is likely to interfere with the practical operation of an assay. This is best illustrated by taking 2 extreme examples. Radioimmunoassays for thyroxine (T,) often show a substantial cross-reaction (10% or more) with triiodothyronine (T,) ; however, as the circulating levels of T, are at least 2 orders of magnitude lower than those of T, this cross-reaction is unimportant in practice. At the other extreme, a radioimmunoassay for T, may show a cross-reaction of 1 with T,. For exactly the reason given above this could be highly critical; with such an assay, circulating T, would make a substantial contribution to the result obtained.
Ch. 9
SPECIFICITY
469
Choice of' materials to examine jor cross-reaction: it is clearly impossible to examine every conceivable biological material for crossreaction in an assay. In practice, a judicious selection has to be made from those substances whose physicochemical nature suggests that cross-reaction is likely. In the case of steroid hormones, the choice is relatively easy since the different structures and their relationship are well understood. Furthermore, a wide range of highly purified materials are available for study. A similar situation exists with respect to small peptides. However, with larger proteins the choice of materials and the interpretation of results may be considerably more difficult. Highly purified preparations are generally scarce, and relatively large amounts may be needed to exclude cross-reaction at the 1 and 0.1 'i:levels. The use of impure material can be misleading since this may contain related proteins which react in the assay, but do not feature in the definition of 'purity' derived from studies with another type of assay. A good example of this is seen with placental and pituitary glycoprotein hormones : striking discrepancies may occur between the results of biological assays and radioimmunoassays (see 8 11.4). Finally, with some substances which are biologically unique (or apparently so) there is no logical basis for selecting other materials for specificity studies: specificity must then depend on the physicochemical characterisation of the substance, together with its apparent behaviour under physiological conditions. 9.2.3. Methods for improving specificity With the naturally occurring binders (circulating proteins or cell receptors) there is little which can be done to alter the specificity of the assay itself, other than the use of the preliminary extraction procedures which have already been described ($6.2).With antibodies, by contrast, the variability and heterogeneity of the primary reagent offers many opportunities for the direct improvement of specificity. Potential approaches include the type of immunisation schedule used, manipulation of the conditions of the assay, and absorption to remove unwanted populations of antibody mofecules. Zrnmunisation scliedule: clearly the most important factor here is S u h p i l in&\ p 531
470
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
the nature of the immunogen. Non-specificity of type C (Fig. 9.1) (due to the presence of irrelevant materials in the ligand) can be obviated if it is possible to use pure ligand. This approach is limited by the availability of highly purified material and by the fact that low-level contaminants may nevertheless be highly immunogenic. Non-specificity of type B (Fig. 9. I ) (due to the presence of common antigenic sites on the ligand and related material) can be avoided if the immunogen used is a fragment of the ligand which does not contain the common antigenic site. A good example of this is seen with the glycoprotein hormones (LH, FSH, TSH and hCG) each of which consists of 2 subunits: the so-called a-subunit which is identical in all four; and the /%subunitwhich is different in all four and confers biological specificity. Because of the common a-subunit, assays based on intact hormone often show striking cross-reactions among all types. However, the use of P-subunit as immunogen (and often as tracer too) yields an assay which is specific for the individual hormone. The drawback to this approach is that the isolated subunit may have a tertiary structure different from that in the intact molecule, and therefore a different antigenic structure. This may explain why assays based on this principle are often less efficient, in terms of antibody affinity and sensitivity, than the less specific assays directed to the intact molecule. In contrast to types B and C, non-specificity of type A (Fig. 9.1) (due to reaction of a single homogeneous antibody population with a range of related antigenic sites) cannot be influenced by the choice of immunogen. Manipulation o f conditions ofassay :as already noted the specificity of an assay may vary with the conditions chosen. For example, nonspecificitydue to a population of low affinity antibody molecules may become insignificant if the antiserum is used at very high dilutions. Non-specificity may also result from the use of short incubation disequilibrium assays ; binding of a ligand and cross-reacting material may be virtually equivalent under these circumstances, and only as equilibrium is approached does the relatively greater dissociation rate of the cross-reacting material yield a relative excess of bound
Ch. 9
SPECIFICITY
47 1
primary ligand. It is essential for this reason to re-check the specificity of an antibody under the exact conditions used in the assay. The specificity can also be influenced by the nature of the tracer; if this consists of highly purified ligand then some of the deficiencies of the antiserum may be obviated. An example of the dramatic effect of using different tracers with the same antiserum is shown in Fig. 2.1A and B. Absorption to remove unwanted popidations of antibody: nonspecificity of type B and C can be eliminated by prior absorption of the antiserum - in the case of type B with a purified preparation of a fragment of the molecule containing the common antigenic sequence; in the case of type C with a preparation of the contaminant material. On occasion the 2 approaches are combined. For example, in the earlier days of gonadotrophin assays it was customary to absorb antisera to hFSH with a preparation of hCG. This had 2 effects: first, to remove populations of antibodies directed towards the common a-subunit ; second, because of the close similarity between LH and hCG, to remove populations of antibodies directed towards LH which arose as a result of the presence of LH as a contaminant in the preparations of FSH used as immunogen. Elimination of type B non-specificity by absorption can also be used to render an assay specific for a ‘neo-antigen’, that is to say, an antigen which is expressed in a fragment of the molecule but not, because of the differences in tertiary structure, in the intact molecule. This is well illustrated by the assay of fibrinogen degradation products (Fig. 9.4). An antiserum to the terminal fragment D contains antibodies specific to the fragment itself, and in addition antibodies which will react with intact fibrinogen. Absorption of the antiserum with fibrinogen removes the latter population and thus yields an assay specific for fragment D with no interference by the parent molecule. In practical terms there are 2 possible approaches to the absorption of an antiserum: (1) simple addition of the cross-reacting material, and removal of the resulting immuno-precipitate by centrifugation (this has the disadvantage that soluble complexes may remain due to the ‘prozone’ phenomenon) ; (2) addition of the cross-reacting Suhpir rnrlcr p 53/
412
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
n
45
z
3
s ap
t
Fibrinogen
1
30 15
1.2
20
320
5m
80,m
nglml
Fig. 9.4. Improving specificity by absorption of an antiserum. In this case the antiserum was raised to pure fragment D (FgD), one of the terminal degradation products of fibrinogen, and then absorbed with fibrinogen. Used with a tracer of '2sI-fragment D this antiserum shows no cross-reaction with intact fibrinogen at the concentrations tested and relatively little cross-reaction with fragment E (FgE). Since fragment D is part of the fibrinogen molecule, this assay is specific to a 'neo-antigen', i.e., one that is revealed during the course of proteolytic destruction of fibrinogen. (From data kindly supplied by Dr. Y . 9. Gordon.)
material in insoluble form, for example, in the case of a protein, after cross-linking with glutaraldehyde. A more general technique with wider applications is the use of antigen coupled to a solid phase such as cellulose or Sephadex. The solid phase can then be added batchwise to the antiserum or used in the form of a column to which the antiserum is applied and eluted. Separation technique ,for bound and ,free ligand: the nature of the separation procedure can affect the specificityof an assay. For example, it has been shown that the cross-reaction of 5a-dihydrotestosterone can be substantially reduced if ammonium sulphate is used for separation rather than dextran-coated charcoal (Tyler et al. 1973). The mechanism of this phenomenon is uncertain.
9.3. Non-spec f i c non-specficity This term refers to interference in an assay by factors other than those which can be clearly identified by their physicochemical
Ch. 9
413
SPECIFICITY
similarity to the ligand. In its most familiar form this is manifest in an assay of an unknown sample by the measurement of an apparent concentration of ligand which is elevated by comparison with the concentration measured by other techniques. On occasion, however, non-specific effects may lead to a reduction in the apparent concentration when compared with the results of other methods. There are 4 basic mechanisms which may lead to non-specific nonspecificity : the presence in the sample of material which interferes with the binder-ligand reaction ; variations of 'blank' values in samples ; destruction or sequestration of binder or tracer; and destruction or sequestration of the unlabelled ligand. 9.3.1. Presence qf'materia1.Y which interfere with the binder-ligand
reactioii There are a variety of materials which will interfere non-specifically with binder-ligand reactions. Examples include the presence of large amounts of highly charged molecules such as heparin; high concentrations of low molecular weight materials such as salt and urea; and conditions of acid or alkaline pH. An example of the effect of acidity is shown in Fig. 6.2. 9.3.2. Variations of'bluiik values in samples
The nature of the fluid used in the assay can have a striking effect on the blank values. For example, with separation by organic precipitation the assay blank may be considerably higher in tubes containing whole serum than in those containing diluent with addition of carrier protein (standards) ; under these circumstances the values given by unknowns will be lower than the true value (Fig. 9.5). Furthermore, the assay blank value may vary between different samples of the same fluid. This could be avoided by carrying out an assay blank determination on every sample (i.e. replicate tubes with sample and tracer but no antibody); such a determination would be very tedious to perform and in practice is rarely necessary. Subject Inder p 531
474
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
100
80
7
actual reading of sample (2) true reading of sample (4)
60
%tracer bound 40
20
1
2
4
8
16
32
Concentration of Standard Fig. 9.5. Model curves showing the effect of a variable blank value on the specificity of a binding assay. If the blank value for the sample is higher than that of the standard, then the result obtained will be artefactually low. This problem is largely avoided if the composition of the standards and unknowns is identical.
9.3.3.Destruction or sequestration of binder or tracer In most systems the binder is relatively stable and is not subject to irreversible damage. The same, however, does not apply to the tracer. At least 3 mechanisms exist which can grossly affect the performance of the latter - enzymic destruction, absorption to surfaces, and binding by endogenous antibodies. Enzymic destruction is an important factor in assays using iodinated protein hormones and in particular small peptides such as vasopressin and angiotensin. These are highly susceptible to hydrolysis by enzymes present in normal plasma or serum, a process which leads to a reduction in binding which may be difficult to distinguish from that produced by the presence of unlabelled ligand (Fig. 9.6). Absorption to surfaces, such as those of glass tubes, is an effect which is also commonly seen with small
Ch. 9
415
SPECIFICITY
80 70
8 60
4
40
-
I
$ 30
20 10
I
’8
0I ’
50
I
I
I
25 12.5 6.3 CONCENTRATION
I
1
I
3.1 1.5 0.8 X PLASMA (sb)
I
0.4
Fig. 9.6. The effect of enzyme destruction on a tracer. [‘251]oxytocin is incubated with an antiserum to oxytocin and varying concentrations of oxytocin-free human late pregnancy plasma (horizontal axis). After 24 hr there is virtually no bound hormone in those tubes containing the highest concentrations of plasma. The loss of immunoreactivity is due to destruction of the tracer by the placental enzyme, oxytocinase.
peptides and which can occur with steroid hormones. Under these circumstances the surface may ‘compete’ with the binder for tracer or unlabelled ligand and thus produces effects which vary widely with thenatureoftheother constituents in the assay tube. Finally, biological samplesmay contain antibodies to the ligand - for example, the serum from a diabetic patient treated with insulin. Depending on the system used for separation, this can lead to grossly discrepant results because of competition for the tracer between the endogenous and added antibody. 9.3.4. Destruction or sequestration of unlabelled ligand The effects described above may also apply to the unlabelled ligand. This applies particularly during the process of collection, preparation Suhjecl indexp. 5 J I
476
RADIOIMMUNOASSAY A N D HELATED TECHNIQUES
and storage of the sample. Small peptides, for example, may be rapidly destroyed by endogenous enzymes and are very liable to irreversible absorption to the surfaces of syringes and tubes, most notably when the latter are made of glass.
9.3.5. The detection and elimination of non-speciJi’crton-specficity The traditional method for detecting non-specificity of any type is the examination ofparallelism between dilutions of sample (unknown) and standard. But, as with specific non-specificity, this can be misleading and should not be relied upon to guarantee identity. Other approaches should also be examined. The most satisfactory approach to the elimination of non-specific non-specificity is to ensure that the overall composition of the standard is as near as possible identical to that of the unknown. For example, if the fluid examined is serum and standards are prepared in ligand-free serum then problems are unlikely to occur. However, difficulties arise with fluids of highly variable composition, such as urine, or, more important, when a source of ligand-free fluid is not readily available. There are two potential sources of ligdnd-free fluid. The first and best is from a subject in whom none of the ligand is present: for instance, in the case of a drug, from an untreated patient; in the case of a hormone, from a subject in whom the gland of origin has been removed (however, see 9: 6.1), or in whom production has been inhibited by an appropriate drug (e.g. dexamethasone suppression of ACTH); in the case of a placental product, from a non-pregnant subject. The second approach is to use fluid from which all endogenous ligand has been removed by absorption; a typical example is the removal of insulin or thyroxine from serum by pretreatment with charcoal. The problem with this approach is the possible absorption of other factors which can interfere non-specifically in the system. As a result, the so-called ‘hormone-free’ serum is not strictly comparable with fresh normal serum and can yield misleading results. If total comparability between sample and standard cannot be guaranteed, and in particular if a reliable source of ligand-free fluid
Ch. 9
477
SPECIFICITY
50
r
40
30 % BOUND
20 10
1
3 4 5 6 7 8 9 FRACTION NUMBER iThin-Layer Chromatography )
2
10
Fig. 9.7. Thin-layer chromatography of an extract of urine after adsorption to and elution from glass beads (see 6 6.1.2 and Table 6.1). Each segment from the plate was extracted with 1 ml60':; (w/v) aqueous acetone. The extract was dried, dissolved in an aqueous buffer solution and incubated with 0.05 ng [12sI]oxytocin and a rabbit O . urine with no antiserum to oxytocin at a final concentration of 1 in 4000. L added hormone; C - H , urine withadded syntheticoxytocin; - 0 , synthetic oxytocin alone run on the same plate. Note that the elution patterns of endogenous material and recovered exogenous material are virtually identical. Neither, however, is identical to synthetic oxytocin alone, indicating some alteration of the material after addition to and extraction from urine.
is not available, there are a number of additional approaches to the elimination or at least the identification of non-specificity. At the level of the assay itself, careful attention to technique can eliminate many potential sources of error. Factors such as absorption of tracer or unlabelled ligand to the surface of tubes can be identified by adding and removing tracer ligand from the tube, and estimating the proportion of counts which remain. It is usually possible, by trial and error, to select a tube which does not show this phenomenon. Enzymic destruction can be a t least partly eliminated by the incorporation in the incubation mixture of enzyme inhibitors such as Trasylol. Extraction and concentration of ligand (Ch. 6 ) can serve
478
RADlOlM M UNOASSAY A N D RELATED TECH NlQUES
to transfer material from an ill-defined environment such as urine or a tissue extract into a well-defined aqueous buffer system. As noted, however, extraction itself may yield non-specific effects. Physicochemical studies may also be of value to determine the precise nature of material measured in an assay. For example, the demonstration that the material inhibiting the binder-ligand reaction has the same properties as the authentic material in several systems of chromatography can greatly enhance the confidence in the results (Fig. 9.7). Finally, and perhaps most important, it is critical with any system to examine the biological specificity of the results. In other words, that the answers obtained with the assay are comparable to those which would be expected from other information on the ligand in question. For example, it is possible with many hormones to obtain an indirect estimate of their circulating levels from a knowledge of their half-life in the circulation and of the amount of administered exogenous material which will produce maximal endorgan response. If the apparent levels measured in an assay are greatly in excess of indirect estimates, then it is likely that non-specific factors are operative. Similarly, the estimated levels should show appropriate variation under physiological or pathological conditions; for instance, material recorded as ‘basal levels’ of ACTH should be reduced or disappear following administration of corticosteroids.
Characteristics of binding assays - specificity
9.1. Definition of‘specijicity The specificity of an assay can be defined as ‘thg degree to which an assay responds to substances other than that for which the assay was designed’. ‘Degree’ in this case is a relative term since it will vary considerably with the conditions of the assay - concentration of tracer and binder, separation procedure, etc. For example, in a radioimmunoassay it is possible to design conditions in which at one concentration of antibody there is no significant interference by a cross-reacting material, while at another concentration the interference is such as to render the assay valueless for all practical purposes. The subject of specificity will be discussed under two separate headings, which for convenience and familiarity will be termed ‘specific non-specificity’ and ‘non-specific non-specificity’.The former refers to interference by identifiable materials which are physicochemically similar to the ligand and may thus react directly with the binder. The latter refers to interference by materials which do not directly react with the binder, but can nevertheless affect the primary binder-ligand reaction - for example, acid conditions which produce partial or total inhibition of any antigen-antibody reaction. The two types of non-specificity may be difficult to distinguish, yet in operative terms it is highly important to do so because the means for their elimination may be quite different.
463
464
RADIOIMMUNOASSAY AND
K E L A I F D TFCHNIQUES
9.2. Speci/i’c non-specificity There are many groups of biological materials which are physicochemically very similar - for example the various steroid hormones, or the glycoprotein hormones of the anterior pituitary gland (LH, FSH, TSH). The specificity of their physiological action depends on target organ receptors capable of distinguishing between them on the basis of small differences in the molecule such as the presence or absence of qn hydroxyl group. The use of these receptors in binding assays confers similar specificity on in vitro measurement. However, the use of receptor assays is limited to a relatively small number of substances -clearly they cannot be applied to non-hormonal materials with no target organ specificity, and equally they are of little value in determining important metabolites of biological compounds. For most purposes other types of binder have to be used - chiefly the naturally occurring circulating binding proteins and antibodies. Much of the discussion which follows will be devoted to the latter. 9.2.1. The basis of specific non-specificily
An antiserum to a given material will usually react, albeit less effectively, with closely related materials. There are 3 basic reasons why this may occur: first, a single homogeneous antibody population may react with a range of related ligands, each reaction having a different affinity constant; second, the antiserum may contain populations of antibody molecules one or more of which is directed to a site on the primary material which also occurs on a related material ; finally, the antiserum may contain antibodies directed to contaminants present in the ligand and which are also present in the related material (Fig. 9.1). Examples of these situations will be found in $9.2.2 and 9.2.3. These concepts are important to an understanding of the experimental finding of ‘parallelism’ or ‘non-parallelism’. If the difference between two related materials is solely in their K value with respect to a single homogeneous binder, then standard curves for these materials using a homologous tracer will be parallel. However,
Ch. 9
465
SPECIFICllY
A
B
C
Fig. 9.1. Different types of specific non-specificity. (A) A single population of antibody molecules reacts with two related materials. hut with a different affinity constant (i.e. one molecule is a good .fit’ to the combining site, the other is a less good fit). (B) A single population of antibody molecules reacts with an antigenic site which is common to two different molcculcs (e.g. the subunit of the glycoprotein hormones). (C) The antiserum contains populations of antihodies to different ligands; the non-specific ligand may then cross-react in the assay.
deviation from parallelism will be seen if the tracer is not homologous and if the mass of tracer represents a significant fraction of the total ligand in a given tube. This type of non-parallelism is found only at the lowest concentrations of unlabelled ligand and would be barely apparent from simple inspection of a standard curve. The second and third types of non-specificity - due to heterogeneity of binder or binding sites - can yield more dramatic non-parallelism. The principle by which this can influence the shape of a standard curve is best illustrated by an extreme case (Fig. 9.2). Clearly, the precise result obtained will vary with the set of reagents used and, in the case of radioimmunoassay, the distribution of antibody populations will vary considerably between different antisera. One might yield apparent parallelism between two cross-reacting materials, while another would yield non-parallelism. In a real system the 3 types of specific non-specificity may often co-exist. The analysis of the main factor involved - heterogeneity of K value at a single site or heterogeneity of sites - can then become very complex. However, gross non-parallelism, particularly at higher concentrations af unlabelled ligand. suggests site heterogeneity. This is of practical importance because the identification of this effect suggests that specificity may be improved by absorption of the antiserum. S,,/7,‘<1,,,
466
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
40
%TRACER BOUND
M
I
A-6
\
20
A B
Primary material W Tracer ,A.B
10
.cog
,032 .la .5 CONCEWRATION OF STANDARD
2
Cross-reading material Ant lbodles
B, A A
A
B
Fig. 9.2. Non-parallelism due to heterogeneity. The primary material and the tracer have two antigenic sites, A and B. The antiserum contains 2 corresponding populations of antibodies. Material containing site B only will produce the curve shown. It can never produce complete inhibition in the assay because antibodies of population A will always be unoccupied and can still bind the tracer.
9.2.2. Assessment o f spec fic non-specficity To test cross-reaction serial dilutions of the material in question are prepared and assayed: the resulting curve is then compared with that given by the standard material for which the assay was designed. The most common way of presenting the result is to compare the amount of the material under study which yields 50% inhibition of binding with the amount of standard giving the same inhibition, and then to express the potency of the material as a percentage of that of the standard (Fig. 9.3). For example, if the respective concentrations are 1 and 100, the potency is stated as 1%. A number of problems can arise with this approach: Concentration by weight versus molar concentration : very often potency is calculated on the basis of relative concentrations by weight. If the molecular weights of standard and cross-reacting material are identical this yields a satisfactory result. If they are not then the answer
Ch. 9
461
SPECIFICI IY
Cross - reaction 25%
100
80
% tracer bound
1
2
4 8 16 Concentrationof Standard
32
64
Fig. 9.3. Estimation of the percentage cross-reaction from the amounts of material required to produce inhibition in the assay. The curve on the left is for the standard. that on the right for the cross-reacting material. Note that this procedure cannot be applicd if the curves are non-parallel.
may be misleading. For example, if the molecular weight of the crossreacting material is 10 times that of the standard, and the potency by weight is given as lo%, then in reality the two materials are equal on a molar basis. To avoid this pitfall calculation as molar concentrations is always to be recommended. Variation with conditions of’ assay: the apparent potency of a cross-reacting material may vary if the conditions of the assay are changed. For example, if an assay is de-sensitised by increase of binder concentration, then a population of non-specific binding sites may become apparent which were of no significance at lower concentrations. More important, specificity can show great variation according to the nature of the tracer employed: an example of this is Sublrrr rndrx p 531
468
KADIOIMMUNOASSAY A N D K E L A l f D 1 ECHNIQUtS
given in 9: 2.2 and Fig. 2.1A, B. Similar though less obvious effects may result from the substitution of highly purified tracer ligand with less purified material. Calculation if' curves are non-parallel: if the curves yielded by standard and cross-reacting material are non-parallel, then clearly a calculation of potency made at 50% inhibition will be very different from that made at 10% (see Fig. 9.2). There is no simple answer to this problem, and non-parallel cross-reaction must be judged in relation to the intended biological or clinical use of the assay. Ariing from this is the question of how to judge parallelism. Though usually a subjective impression, this can be highly misleading: substantial differences can easily be missed, particularly at the extremes of the curve. There are two approaches to this problem. The first is to perform a logit transformation of the results (Q 7.2) and then to make a statistical comparison of the 2 linear regressions. The second is to calculate the apparent potency of each dilution, and then to examine the figures for any systematic deviation between standard and crossreacting material. In judging non-parallelism it is also important to exclude experimental artefact, such as that which can arise as a result of serial dilutions of the standard; indeed comparison of serial dilutions of a material with independently prepared dilutions of the same material is almost certain to yield some degree of nonparallelism (see Q 2.5). Relevance to physiological and clinical situations : the eventual significance of a cross-reacting material lies in the extent to which it is likely to interfere with the practical operation of an assay. This is best illustrated by taking 2 extreme examples. Radioimmunoassays for thyroxine (T,) often show a substantial cross-reaction (10% or more) with triiodothyronine (T,) ; however, as the circulating levels of T, are at least 2 orders of magnitude lower than those of T, this cross-reaction is unimportant in practice. At the other extreme, a radioimmunoassay for T, may show a cross-reaction of 1 with T,. For exactly the reason given above this could be highly critical; with such an assay, circulating T, would make a substantial contribution to the result obtained.
Ch. 9
SPECIFICITY
469
Choice of' materials to examine jor cross-reaction: it is clearly impossible to examine every conceivable biological material for crossreaction in an assay. In practice, a judicious selection has to be made from those substances whose physicochemical nature suggests that cross-reaction is likely. In the case of steroid hormones, the choice is relatively easy since the different structures and their relationship are well understood. Furthermore, a wide range of highly purified materials are available for study. A similar situation exists with respect to small peptides. However, with larger proteins the choice of materials and the interpretation of results may be considerably more difficult. Highly purified preparations are generally scarce, and relatively large amounts may be needed to exclude cross-reaction at the 1 and 0.1 'i:levels. The use of impure material can be misleading since this may contain related proteins which react in the assay, but do not feature in the definition of 'purity' derived from studies with another type of assay. A good example of this is seen with placental and pituitary glycoprotein hormones : striking discrepancies may occur between the results of biological assays and radioimmunoassays (see 8 11.4). Finally, with some substances which are biologically unique (or apparently so) there is no logical basis for selecting other materials for specificity studies: specificity must then depend on the physicochemical characterisation of the substance, together with its apparent behaviour under physiological conditions. 9.2.3. Methods for improving specificity With the naturally occurring binders (circulating proteins or cell receptors) there is little which can be done to alter the specificity of the assay itself, other than the use of the preliminary extraction procedures which have already been described ($6.2).With antibodies, by contrast, the variability and heterogeneity of the primary reagent offers many opportunities for the direct improvement of specificity. Potential approaches include the type of immunisation schedule used, manipulation of the conditions of the assay, and absorption to remove unwanted populations of antibody mofecules. Zrnmunisation scliedule: clearly the most important factor here is S u h p i l in&\ p 531
470
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
the nature of the immunogen. Non-specificity of type C (Fig. 9.1) (due to the presence of irrelevant materials in the ligand) can be obviated if it is possible to use pure ligand. This approach is limited by the availability of highly purified material and by the fact that low-level contaminants may nevertheless be highly immunogenic. Non-specificity of type B (Fig. 9. I ) (due to the presence of common antigenic sites on the ligand and related material) can be avoided if the immunogen used is a fragment of the ligand which does not contain the common antigenic site. A good example of this is seen with the glycoprotein hormones (LH, FSH, TSH and hCG) each of which consists of 2 subunits: the so-called a-subunit which is identical in all four; and the /%subunitwhich is different in all four and confers biological specificity. Because of the common a-subunit, assays based on intact hormone often show striking cross-reactions among all types. However, the use of P-subunit as immunogen (and often as tracer too) yields an assay which is specific for the individual hormone. The drawback to this approach is that the isolated subunit may have a tertiary structure different from that in the intact molecule, and therefore a different antigenic structure. This may explain why assays based on this principle are often less efficient, in terms of antibody affinity and sensitivity, than the less specific assays directed to the intact molecule. In contrast to types B and C, non-specificity of type A (Fig. 9.1) (due to reaction of a single homogeneous antibody population with a range of related antigenic sites) cannot be influenced by the choice of immunogen. Manipulation o f conditions ofassay :as already noted the specificity of an assay may vary with the conditions chosen. For example, nonspecificitydue to a population of low affinity antibody molecules may become insignificant if the antiserum is used at very high dilutions. Non-specificity may also result from the use of short incubation disequilibrium assays ; binding of a ligand and cross-reacting material may be virtually equivalent under these circumstances, and only as equilibrium is approached does the relatively greater dissociation rate of the cross-reacting material yield a relative excess of bound
Ch. 9
SPECIFICITY
47 1
primary ligand. It is essential for this reason to re-check the specificity of an antibody under the exact conditions used in the assay. The specificity can also be influenced by the nature of the tracer; if this consists of highly purified ligand then some of the deficiencies of the antiserum may be obviated. An example of the dramatic effect of using different tracers with the same antiserum is shown in Fig. 2.1A and B. Absorption to remove unwanted popidations of antibody: nonspecificity of type B and C can be eliminated by prior absorption of the antiserum - in the case of type B with a purified preparation of a fragment of the molecule containing the common antigenic sequence; in the case of type C with a preparation of the contaminant material. On occasion the 2 approaches are combined. For example, in the earlier days of gonadotrophin assays it was customary to absorb antisera to hFSH with a preparation of hCG. This had 2 effects: first, to remove populations of antibodies directed towards the common a-subunit ; second, because of the close similarity between LH and hCG, to remove populations of antibodies directed towards LH which arose as a result of the presence of LH as a contaminant in the preparations of FSH used as immunogen. Elimination of type B non-specificity by absorption can also be used to render an assay specific for a ‘neo-antigen’, that is to say, an antigen which is expressed in a fragment of the molecule but not, because of the differences in tertiary structure, in the intact molecule. This is well illustrated by the assay of fibrinogen degradation products (Fig. 9.4). An antiserum to the terminal fragment D contains antibodies specific to the fragment itself, and in addition antibodies which will react with intact fibrinogen. Absorption of the antiserum with fibrinogen removes the latter population and thus yields an assay specific for fragment D with no interference by the parent molecule. In practical terms there are 2 possible approaches to the absorption of an antiserum: (1) simple addition of the cross-reacting material, and removal of the resulting immuno-precipitate by centrifugation (this has the disadvantage that soluble complexes may remain due to the ‘prozone’ phenomenon) ; (2) addition of the cross-reacting Suhpir rnrlcr p 53/
412
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
n
45
z
3
s ap
t
Fibrinogen
1
30 15
1.2
20
320
5m
80,m
nglml
Fig. 9.4. Improving specificity by absorption of an antiserum. In this case the antiserum was raised to pure fragment D (FgD), one of the terminal degradation products of fibrinogen, and then absorbed with fibrinogen. Used with a tracer of '2sI-fragment D this antiserum shows no cross-reaction with intact fibrinogen at the concentrations tested and relatively little cross-reaction with fragment E (FgE). Since fragment D is part of the fibrinogen molecule, this assay is specific to a 'neo-antigen', i.e., one that is revealed during the course of proteolytic destruction of fibrinogen. (From data kindly supplied by Dr. Y . 9. Gordon.)
material in insoluble form, for example, in the case of a protein, after cross-linking with glutaraldehyde. A more general technique with wider applications is the use of antigen coupled to a solid phase such as cellulose or Sephadex. The solid phase can then be added batchwise to the antiserum or used in the form of a column to which the antiserum is applied and eluted. Separation technique ,for bound and ,free ligand: the nature of the separation procedure can affect the specificityof an assay. For example, it has been shown that the cross-reaction of 5a-dihydrotestosterone can be substantially reduced if ammonium sulphate is used for separation rather than dextran-coated charcoal (Tyler et al. 1973). The mechanism of this phenomenon is uncertain.
9.3. Non-spec f i c non-specficity This term refers to interference in an assay by factors other than those which can be clearly identified by their physicochemical
Ch. 9
413
SPECIFICITY
similarity to the ligand. In its most familiar form this is manifest in an assay of an unknown sample by the measurement of an apparent concentration of ligand which is elevated by comparison with the concentration measured by other techniques. On occasion, however, non-specific effects may lead to a reduction in the apparent concentration when compared with the results of other methods. There are 4 basic mechanisms which may lead to non-specific nonspecificity : the presence in the sample of material which interferes with the binder-ligand reaction ; variations of 'blank' values in samples ; destruction or sequestration of binder or tracer; and destruction or sequestration of the unlabelled ligand. 9.3.1. Presence qf'materia1.Y which interfere with the binder-ligand
reactioii There are a variety of materials which will interfere non-specifically with binder-ligand reactions. Examples include the presence of large amounts of highly charged molecules such as heparin; high concentrations of low molecular weight materials such as salt and urea; and conditions of acid or alkaline pH. An example of the effect of acidity is shown in Fig. 6.2. 9.3.2. Variations of'bluiik values in samples
The nature of the fluid used in the assay can have a striking effect on the blank values. For example, with separation by organic precipitation the assay blank may be considerably higher in tubes containing whole serum than in those containing diluent with addition of carrier protein (standards) ; under these circumstances the values given by unknowns will be lower than the true value (Fig. 9.5). Furthermore, the assay blank value may vary between different samples of the same fluid. This could be avoided by carrying out an assay blank determination on every sample (i.e. replicate tubes with sample and tracer but no antibody); such a determination would be very tedious to perform and in practice is rarely necessary. Subject Inder p 531
474
RADIOIMMUNOASSAY A N D RELATED TECHNIQUES
100
80
7
actual reading of sample (2) true reading of sample (4)
60
%tracer bound 40
20
1
2
4
8
16
32
Concentration of Standard Fig. 9.5. Model curves showing the effect of a variable blank value on the specificity of a binding assay. If the blank value for the sample is higher than that of the standard, then the result obtained will be artefactually low. This problem is largely avoided if the composition of the standards and unknowns is identical.
9.3.3.Destruction or sequestration of binder or tracer In most systems the binder is relatively stable and is not subject to irreversible damage. The same, however, does not apply to the tracer. At least 3 mechanisms exist which can grossly affect the performance of the latter - enzymic destruction, absorption to surfaces, and binding by endogenous antibodies. Enzymic destruction is an important factor in assays using iodinated protein hormones and in particular small peptides such as vasopressin and angiotensin. These are highly susceptible to hydrolysis by enzymes present in normal plasma or serum, a process which leads to a reduction in binding which may be difficult to distinguish from that produced by the presence of unlabelled ligand (Fig. 9.6). Absorption to surfaces, such as those of glass tubes, is an effect which is also commonly seen with small
Ch. 9
415
SPECIFICITY
80 70
8 60
4
40
-
I
$ 30
20 10
I
’8
0I ’
50
I
I
I
25 12.5 6.3 CONCENTRATION
I
1
I
3.1 1.5 0.8 X PLASMA (sb)
I
0.4
Fig. 9.6. The effect of enzyme destruction on a tracer. [‘251]oxytocin is incubated with an antiserum to oxytocin and varying concentrations of oxytocin-free human late pregnancy plasma (horizontal axis). After 24 hr there is virtually no bound hormone in those tubes containing the highest concentrations of plasma. The loss of immunoreactivity is due to destruction of the tracer by the placental enzyme, oxytocinase.
peptides and which can occur with steroid hormones. Under these circumstances the surface may ‘compete’ with the binder for tracer or unlabelled ligand and thus produces effects which vary widely with thenatureoftheother constituents in the assay tube. Finally, biological samplesmay contain antibodies to the ligand - for example, the serum from a diabetic patient treated with insulin. Depending on the system used for separation, this can lead to grossly discrepant results because of competition for the tracer between the endogenous and added antibody. 9.3.4. Destruction or sequestration of unlabelled ligand The effects described above may also apply to the unlabelled ligand. This applies particularly during the process of collection, preparation Suhjecl indexp. 5 J I
476
RADIOIMMUNOASSAY A N D HELATED TECHNIQUES
and storage of the sample. Small peptides, for example, may be rapidly destroyed by endogenous enzymes and are very liable to irreversible absorption to the surfaces of syringes and tubes, most notably when the latter are made of glass.
9.3.5. The detection and elimination of non-speciJi’crton-specficity The traditional method for detecting non-specificity of any type is the examination ofparallelism between dilutions of sample (unknown) and standard. But, as with specific non-specificity, this can be misleading and should not be relied upon to guarantee identity. Other approaches should also be examined. The most satisfactory approach to the elimination of non-specific non-specificity is to ensure that the overall composition of the standard is as near as possible identical to that of the unknown. For example, if the fluid examined is serum and standards are prepared in ligand-free serum then problems are unlikely to occur. However, difficulties arise with fluids of highly variable composition, such as urine, or, more important, when a source of ligand-free fluid is not readily available. There are two potential sources of ligdnd-free fluid. The first and best is from a subject in whom none of the ligand is present: for instance, in the case of a drug, from an untreated patient; in the case of a hormone, from a subject in whom the gland of origin has been removed (however, see 9: 6.1), or in whom production has been inhibited by an appropriate drug (e.g. dexamethasone suppression of ACTH); in the case of a placental product, from a non-pregnant subject. The second approach is to use fluid from which all endogenous ligand has been removed by absorption; a typical example is the removal of insulin or thyroxine from serum by pretreatment with charcoal. The problem with this approach is the possible absorption of other factors which can interfere non-specifically in the system. As a result, the so-called ‘hormone-free’ serum is not strictly comparable with fresh normal serum and can yield misleading results. If total comparability between sample and standard cannot be guaranteed, and in particular if a reliable source of ligand-free fluid
Ch. 9
477
SPECIFICITY
50
r
40
30 % BOUND
20 10
1
3 4 5 6 7 8 9 FRACTION NUMBER iThin-Layer Chromatography )
2
10
Fig. 9.7. Thin-layer chromatography of an extract of urine after adsorption to and elution from glass beads (see 6 6.1.2 and Table 6.1). Each segment from the plate was extracted with 1 ml60':; (w/v) aqueous acetone. The extract was dried, dissolved in an aqueous buffer solution and incubated with 0.05 ng [12sI]oxytocin and a rabbit O . urine with no antiserum to oxytocin at a final concentration of 1 in 4000. L added hormone; C - H , urine withadded syntheticoxytocin; - 0 , synthetic oxytocin alone run on the same plate. Note that the elution patterns of endogenous material and recovered exogenous material are virtually identical. Neither, however, is identical to synthetic oxytocin alone, indicating some alteration of the material after addition to and extraction from urine.
is not available, there are a number of additional approaches to the elimination or at least the identification of non-specificity. At the level of the assay itself, careful attention to technique can eliminate many potential sources of error. Factors such as absorption of tracer or unlabelled ligand to the surface of tubes can be identified by adding and removing tracer ligand from the tube, and estimating the proportion of counts which remain. It is usually possible, by trial and error, to select a tube which does not show this phenomenon. Enzymic destruction can be a t least partly eliminated by the incorporation in the incubation mixture of enzyme inhibitors such as Trasylol. Extraction and concentration of ligand (Ch. 6 ) can serve
478
RADlOlM M UNOASSAY A N D RELATED TECH NlQUES
to transfer material from an ill-defined environment such as urine or a tissue extract into a well-defined aqueous buffer system. As noted, however, extraction itself may yield non-specific effects. Physicochemical studies may also be of value to determine the precise nature of material measured in an assay. For example, the demonstration that the material inhibiting the binder-ligand reaction has the same properties as the authentic material in several systems of chromatography can greatly enhance the confidence in the results (Fig. 9.7). Finally, and perhaps most important, it is critical with any system to examine the biological specificity of the results. In other words, that the answers obtained with the assay are comparable to those which would be expected from other information on the ligand in question. For example, it is possible with many hormones to obtain an indirect estimate of their circulating levels from a knowledge of their half-life in the circulation and of the amount of administered exogenous material which will produce maximal endorgan response. If the apparent levels measured in an assay are greatly in excess of indirect estimates, then it is likely that non-specific factors are operative. Similarly, the estimated levels should show appropriate variation under physiological or pathological conditions; for instance, material recorded as ‘basal levels’ of ACTH should be reduced or disappear following administration of corticosteroids.