Jeffrey,

My answer to "what does the shape of the attached standard curve mean?" I break down into the following observations:

1. I assume that the yellow(solid) line is the fit curve (Logistic fit?), the dashed lines are the confidence intervals for the fit and the circle are the raw data. Better labeling of the graph would be helpful.

2. The fit line follows an appropriate shape, that is it is sigmoidal for a semi-log plot ; upward indicates that you are using an immunometric, that is a "sandwich" assay.

3.The saturation of the curve is occuring around 1 ng/ml. Reasons for this will follow.

4. There is a high-dose hook occuring around 2 ng/ml. This is due to over-saturation of the binding sites with the target (Ag). Reasons for this will follow.

5. Just to be nitpicky a little, there is no "linear" portion for this type of curve...what you plotted is a logistic curve which is non-linear and is representative of a binding isotherm (ref. David Rodbard et. al. and math of logistic regression analysis). In fact the mathmatical shape of the curve is referred to as a rectangular hyperbola.

What this all means regarding the immunoassay components for your sandwhich assay is:

1. too few binding sites on either your primary, secondary or both antibodies for Ag concentrations above 1 ng/ml (assuming a two antibody system with direct label)

2. reason 1 plus too few attachment sites for the label if a carrier system (strep-aviden/biotin labeling is being used).

3. Low Antibody affinity looks like all of the above.

4. Mismatched Ab affinity looks like all of the above, that is, a secondary Ab with relatively high affinity to the primary will "solublize" the Ag back into solution.

5. The instrument response is "maxing" out (you already caught this possibility). You should look at the performance qualification documentation for your reader and determine if you are within the validated operating ranges for the measurement. Points 1-4 assume that the reader is capable of the measurements beyond the saturation level that you are seeing.

Moving forward:

1. Make all measurements at the high end of your curve within the capability of your instrument.

2. characterize your assay, create a statistical design (partial factorial) to optimize antibody concentrations, incubation times, sample volumes, etc. for maximum dose response range. Analzye especially for first degree interactions.

3. Perform enough replicates at zero dose to confirm that changes you make to extend the upper end of the curve do not impact the lower end metric for analytical sensitivity (LOD).

4. If you are fitting to 4 or 5 PLC, note that you want to optimize your results to increase the dose at 50% signal; this is an excellent metric to monitor curve shift as you alter assay conditions.

I hope that this information is useful. Good luck and happy New Year!

I don't think the change in volume is going to help. The log range of 0.5 to 1.0 seems to be the linear range. It's advisable to use the antibody dilution in this range. This can be called as narrow range in contrary to your claims. The narrow range can be few logs or few fractions of a log, depending on the binding affinity.

Hope I am trying to clarify some concepts.

I am preferring the above range as any study needs to be performed in log phase than lag or stationary phase of the curve.

Best Wishes

Thanks for your input. We will begin sample dilutions to maintain their being within this range

I'm unable to open the attachment, but I have experienced similar issues in the past, when detecting growth products as a means of monitoring growth curves, resulting in the requirement of dilutions. With any closed system, there is a finite amount of the antibodies and once all are used up, the curve reaches a plateau, despite the presence of further antigen (or anitbody, depending on the assay). For certain ELISAs, there may also be inter-batch variability, so if direct comparisons are made, it is important to use the same batch for each point. One example would be if monitoring treatment efficacy, for us it was the points on the growth curve.

Would like to comment but I also could not open to the file

The increase in volume of coating will not help. It seems that your ELISA is too sensitive for the range you are looking. 2 suggestions:

1. if possible with your reader, make a second reading at another wavelength (e.g. 490 nm) which is further from the optimum absorbance of your colorimetric reagent. You can then read the low concentrated samples (1 ng/ml) at 490.

2. you can reduce the sensitivity of you ELISA by adding unlabeled detecting antibody in the assay. But in that case you will loose sensitivity.

It's a nice standard curve and reflects what you can expect in any antigen-antibody binding experiment. For a better mathematical unterstanding make you more familiar with the Hill-equation and the phenomenon of cooperativity. Half maximal effect reflects the apparent Kd value. The higher the cooperativity, the steeper the linear part of the efffect vs conc (log scale) plot. Assume the molecular weight of your antibody when doing the calculation or perhaps you know it approximately from SDS PAGE or literature.

Sorry about the attachment. I converted it to microsoft ppt for 97-2003 and also made it a jpeg. Not sure why the pptx will not work. Thanks for your input.

Here is pdf version of plot

Try to decrease the incubation period of Ag-Ab reaction.

I think changing the coating volume from 200 to 300 uL will not change the curve. We used the concentration at which extinction is half-maximal as an optimal concentration.

Infected culture can effect elisa results about diagnosis.

The shape of your curve and the max binding is a function of your antibodies used in the ELISA. If you have multiple antibodies to try in your assay, I would try all combinations. That would be my first trys. If you have the best antibodies possible, there are a few things you can do to increase the dynamic range of your assay: 1) Increasing the amout of coating antibody can help but as mentioned earlier you will eventually reach a max. I would increase the concentration of the coating solution to test this parameter. A Max volume of 100uL is usually best. 2) What detection enzyme reagent are you using, HRP or AlkPhos? If possible, I would try both. These do not always give the same curve shape 3) along the same line, substrate can sometimes increase the dynamic range. I was always partial to TMB. 4) Incubation times and temperatures can effect the dynamic range; both for your primary incubations and for your detection antibody incubations.

Dilution of samples is not an issue, just be sure to have all of the specifics qritten in your developmental report and finally written in your SOPs.

If the OD is on the Y axis and the concentration(s) of known protein concentration is on the X axis.

As the concentration of the protein to be measured increases , the OD will increase and the line of the curve will also increase until it reaches maximum or the optimum where it will then plateau. At a certain concentration , the curve will then go down (in line with the principle of region of antibody excess and antigen excess).

When you read your assay, you will then have an OD reading. Then put the OD reading against the standard curve and you will be able to obtain the concentration of your protein. If you increase the surface area of wells coated with antibody capture, of course your curve may change

Jeffrey,

My answer to "what does the shape of the attached standard curve mean?" I break down into the following observations:

1. I assume that the yellow(solid) line is the fit curve (Logistic fit?), the dashed lines are the confidence intervals for the fit and the circle are the raw data. Better labeling of the graph would be helpful.

2. The fit line follows an appropriate shape, that is it is sigmoidal for a semi-log plot ; upward indicates that you are using an immunometric, that is a "sandwich" assay.

3.The saturation of the curve is occuring around 1 ng/ml. Reasons for this will follow.

4. There is a high-dose hook occuring around 2 ng/ml. This is due to over-saturation of the binding sites with the target (Ag). Reasons for this will follow.

5. Just to be nitpicky a little, there is no "linear" portion for this type of curve...what you plotted is a logistic curve which is non-linear and is representative of a binding isotherm (ref. David Rodbard et. al. and math of logistic regression analysis). In fact the mathmatical shape of the curve is referred to as a rectangular hyperbola.

What this all means regarding the immunoassay components for your sandwhich assay is:

1. too few binding sites on either your primary, secondary or both antibodies for Ag concentrations above 1 ng/ml (assuming a two antibody system with direct label)

2. reason 1 plus too few attachment sites for the label if a carrier system (strep-aviden/biotin labeling is being used).

3. Low Antibody affinity looks like all of the above.

4. Mismatched Ab affinity looks like all of the above, that is, a secondary Ab with relatively high affinity to the primary will "solublize" the Ag back into solution.

5. The instrument response is "maxing" out (you already caught this possibility). You should look at the performance qualification documentation for your reader and determine if you are within the validated operating ranges for the measurement. Points 1-4 assume that the reader is capable of the measurements beyond the saturation level that you are seeing.

Moving forward:

1. Make all measurements at the high end of your curve within the capability of your instrument.

2. characterize your assay, create a statistical design (partial factorial) to optimize antibody concentrations, incubation times, sample volumes, etc. for maximum dose response range. Analzye especially for first degree interactions.

3. Perform enough replicates at zero dose to confirm that changes you make to extend the upper end of the curve do not impact the lower end metric for analytical sensitivity (LOD).

4. If you are fitting to 4 or 5 PLC, note that you want to optimize your results to increase the dose at 50% signal; this is an excellent metric to monitor curve shift as you alter assay conditions.

I hope that this information is useful. Good luck and happy New Year!

My short answer is: yes, I believe you are interpreting the curve correctly and no, adding more capture antibody to the wells will not help (it will only increase your absorbance at baseline, which in your case is the opposite of what is desired). The shape of the standard curve you attached looks completely typical to me. Keep in mind that ELISA is less of a quantitative assay than a qualitative (diagnostic) assay with limited quantitative abilities. Some ELISAs will give you the standard curve you attached here, with a small dynamic range, and others might give something you are more used to, with a larger range.

So, what should you do if you want to increase your predictive range (the length of the straight part) of your standard curve? My interpretation of your curve is that your lower limit of detection is fine (about 4 ng/ml) but your upper limit is, well, limited (about 1 ng/ml). I suggest you check your antibodies, specifically, their valence, species of origin and antigen (i.e. bovine serum) affinity. If you use a multivalent antibody in any step of the assay, particularly as your detection antibody, that may be your problem. If this is the case, try a different, monovalent antibody. See if any of the other antibody parameters conflict as well. You may find it appropriate to change the type of ELISA you are using (e.g. switch from indirect to competitive ELISA, or vice versa). Otherwise, diluting your sample may be your best choice.

I hope this helps. Good luck!

I think sir the shape of your curve and the max binding is a function of your antibodies used in the ELISA. I assume that the yellow(solid) line is the fit curve , the dashed lines are the confidence intervals for the fit and the circle are the raw data. Incubation times and temperatures can effect the dynamic range; both for your primary incubations and for your detection antibody incubations.

Finally Try to decrease the incubation period of Ag-Ab reaction.

Hi Jeffrey, I guess that you are trying to set up the same conditions for all your ELISAs and this particular case doesn't fit with your protocol, so you will have to do this separate. From the graph (by the way, I could open the pptx) the linear portion of your curve is between ~ 0.6 to 1.0 ng/ml. So, unless you make only minimal dilutions in that range you will have the same problem. People mentioned that to increase the amount of coated antibody will not change this. I agree, but instead I would reduce (or do titration) the amount of coated/capture antibody and I would try also different dilutions of your anti secondary HRP (or whatever you use). Sometimes the recommended dilution for the secondary is far for the one you find that works better (i.e 1/4,000 and I use 1/20,000). Also, is it possible that your detection antibody binds unspecifically to your capture Ab? For instance human serum binds unspecifically to recombinant proteins due to the presence of the alpha GAL epitope in the recombinant protein and the presence of anti-alpha GAL antibodies in 1 % of human IgGs.

I think you should follow the manufacturer instructions and, after done, you can do the analysis without using the program reader for calculations. I also do not like to dilute the samples.

From the attached chart, there is only limited amount of information you can draw. Try to answer one question at the time. Your curve suggest that optimal concentration of antibody (formation of ag-ab complexes) is 1 ng/ml. You should use such concentration in (for example) competition ELISA.

Wow! truly valuable information provided. I am grateful for the nit-pickiness. You are correct, prism 5 was used to generate this curve as it appears to me to be precisely as you describe: goes up linearly to a point, levels off, and as Ag increases, hooks downward. My apologies for the limited information up front.

Yes, the graph provided (that some of you can download; by the way, I could not download image once I uploaded it) demonstrates the data (orange-ish line with circles), fitted line (yellow line), 95% confidence intervals (black dashed lines). We are developing this ELISA in lab so we are responsible for preparing instructions. We do use other ELISA assay, however this is one we have been playing with for some time.

I need to think about your answers and formulate my reply (basically go back to notebooks to get answers to some of your ?). We have titrated Ab (1o and 2o), but some other modifications suggested may be of use. Thanks very much. Depending upon where you are in the world, some of you it is already 2013. For the rest of us, happy new year.

First, you are interpreting the curve correctly with respect to the first-degree linear portion.

Second, one wants to saturate the wells with capture Ab to maximize sensitivity. I agree with Ezio Petrella above:

(1) Go back and try different concentrations of capture Ab (which is anti-what?) to see where the binding of Ag levels off.

(2) Use 100 µL of reagent per well, except for blocking agent, where I use 150 µL.

However, I am not totally sure how to interpret your graph because of the way you labeled the x-axis. Are you measuring the titers of bovine serum against a known amount of Ag per well? Are you graphing [total serum protein] or [Ab]?

Also, without error bars (and I know they can be large in ELISA) and N per point on the graph, one can not really determine if the drop off the plateau at the high-concentration end is real. If the drop at the high end is real, and I read the graph right, then you have excess primary Ab, which causes an inhibitory effect of reduced sensitivity, as others have noted. Please give us more information.

Warning: There's a semantic problem, at least in English, with terms for Ab-Ag interactions. Many people call the effect of excess Ag "prozone", while I learned that effect as "postzone" and the effect of excess (primary) Ab as prozone. It is imnportant to note that these terms were developed to describe observations of titer plates, not of ELISAs. Also, while I learned the use of the term "hook effect" to describe excess Ag _only_, others seem to use the term to describe the effect of excess Ab, or of excess of either. From experience, I doubt seriously that your [secondary Ab] is causing any problems. I usually follow manufacturer's directions on conjugated 2° Ab, but I have doubled the concentration when dealing with biotin-streptavidin assays.

Currently I run a lot of ELISAs. I can provide you references if you wish.

A few things are important in standard curve ELISA,Dilution and blocking,You should use the dilution solution that ,Does not reduce the power Intrinsic blocking of sample.

I think it is easier to dilute the samples with the buffer 1 to 2 sistematically till you fit at the middle of the curve. It is very important to maintain the sensitivity of the assay as well

It is simple - you have reached the saturation point. in ELISA is a good rule to have the OD in the range of 0 to 1. Dilutions should be used as was previously menthioned.

I just managed to have a look at the graph that you have attached. My previous answer still remain. My additional points are as follows:

1. Please provide a lit bit more information on your graph so that we do not have to

assume.

a) Is the red line referring to the bovine serum with dilution 1:64

b) Is the yellow line referring to the bovine serum with dilution 1:2

c) Then what are the black dashed lines referring to?

2. If I assume that the curve in red is the reading of BSA (1:64) and the curve in

yellow is the reading of BSA 1:2, the logic is:

As the antigen concentration increases from 0.25 to 1.0 ng/ml, both curves

(red and yellow) increase and reach their maximum at OD 3.0, but with the

yellow curve slightly lower than the red curve. Both curves then plateau off

only at certain concentrations of the antigen.

The red curve reaches its maximum OD of 3.0 at antigen concentrations

between 1.0 to 1.75 ng/ml. As the antigen concentration increases to 2.0ng/ml

the curve then dips down indicating that the reactions have arrive at the area of

antigen excess. Using antibody dilution of 1:64, the reaction reaches its

maximum at antigen concentrations between 1.0 to 1.75 ng/ml. Therefore the

optimum antigen concentration is 1.0 ng/ml.

The yellow curve plateau off further at antigen concentration >2.0 ng/ml. Using

dilution of 1:2, the amount of antibody is still proportional to the amount of

antigen. If the reaction is allowed to proceed with increasing amount of antigen,

(>2.5 ng/ml) the yellow curve will dip down at a point.

Final point is for this assay the opitmum antibody dilution is 1:64 and the optimum

antigen concentration is 1.0ng/ml.

Apart from above expert suggestions u've to follow some important steps like ELISA is a powerful and well-characterized application, attempting to develop and optimize a specific assay can be difficult. The method involves the assembly of a large immune complex with multiple components, therefore failure to capture signal can potentially be caused by any of the factors and optimization is essential.

When setting up an ELISA it is advisable to first generate and optimize a standard curve for the analyte of interest before analyzing multiple samples of unknown composition. If the standard curve displays the correct sensitivity, range and linearity, the researcher can proceed with confidence to process the samples.

Hi Ramelah,

I appreciate your time to request additional information. Here are my comments/clarifications:

1. Please provide a lit bit more information on your graph so that we do not have to

assume.

a) Is the red line referring to the bovine serum with dilution 1:64

b) Is the yellow line referring to the bovine serum with dilution 1:2

c) Then what are the black dashed lines referring to?

Answer: The red line with dots is the actual absorbance data from the plate reader, the yellow line is the line from fitted or predicted data and the black dashed lines are the 95% confidence intervals. I used a program called graph pad prism which among other things provides fitted curves of data which are not linear over the range of dilutions prepared. In this case, the program fitted the line from absorbance values associated with log10 transformed concentrations. I started with cattle serum having the concentration of a biomarker of approximately 1000 ng/mL (actual concentration was 980 ng/mL). I made serial dilutions from 1:2 to 1:64. The absorbance values obtained from microtiter plate wells were plotted against log10 transformed concentration of each dilution.

2. If I assume that the curve in red is the reading of BSA (1:64) and the curve in

yellow is the reading of BSA 1:2, the logic is:

As the antigen concentration increases from 0.25 to 1.0 ng/ml, both curves

(red and yellow) increase and reach their maximum at OD 3.0, but with the

yellow curve slightly lower than the red curve. Both curves then plateau off

only at certain concentrations of the antigen.

The red curve reaches its maximum OD of 3.0 at antigen concentrations

between 1.0 to 1.75 ng/ml. As the antigen concentration increases to 2.0ng/ml

the curve then dips down indicating that the reactions have arrive at the area of

antigen excess. Using antibody dilution of 1:64, the reaction reaches its

maximum at antigen concentrations between 1.0 to 1.75 ng/ml. Therefore the

optimum antigen concentration is 1.0 ng/ml.

The yellow curve plateau off further at antigen concentration >2.0 ng/ml. Using

dilution of 1:2, the amount of antibody is still proportional to the amount of

antigen. If the reaction is allowed to proceed with increasing amount of antigen,

(>2.5 ng/ml) the yellow curve will dip down at a point.

Final point is for this assay the opitmum antibody dilution is 1:64 and the optimum

antigen concentration is 1.0ng/ml.

Even if struggling to understand it (and actually not finally understanding everything), I like Neal Siegel's answer.

If we consider that the binding sites of Ag/Abs and affinities are the ones we have to live with, then the sensitivity range of the ELISA could be widened by increasing the concentrations of the 1ary or the 2ary Ab, depending on which is the limiting one.

I don’t understand why some people rules out that increasing the 1ary Ab concentration wouldn’t help. In an imaginary situation, if there was only one 1ary Ab molecule bound to plastic, this would immediately get saturated, giving a nothing or all signal, similar to the rectangular hyperbola you have. I can also imagine that the limiting Ab is the 2ary (I might be discovering the garlic soup now, but anyway, my contribution…)

Oh well, Jeffrey, but you say you have already titrated the Abs… It’s not that then.., how annoying…

I don’t think there is a problem of unspecific binding of the anti-2ary-HRP, because this would increase the general background, also at low dilutions.

CRAZY IDEA: maybe you are trying to detect something that above certain concentration in the serum is complexed to something else and that’s why you can’t detect above a certain point? (I don’t know if this usually affects in ELISA, but I know that things bind to BSA and stuff. …although these should be high affinity complexes as to compete with a 1ary Ab… Maybe other Abs in serum of an immune cow?) So you would be narrowed from the downside by the sensitivity of the ELISA and from the upperside by that competing thing in the serum.

About the “hooking down” effect at the highest Ag concentration, inspired by what Siegel said, I can imagine that it can be due to a high concentration of Ag taking the 1ary Ab to swim, rather than the 2ary taking the Ag to swim, because the 2ary is not increasing in concentration.

And happy new year!

You should use results only of the linear part of the curve .

try plotting both axes on non-log scales. If there is a linear portion of the curve that passes thru the origin use that. Where the curve plateaus at higher values this indicates saturation.

Please remember that absorption measurements (i.e. ODs) are NOT linear above about 2. Therefore your data MUST fall into the linear reading below 2 units to make any sense.

Dear sir where the curve go to highly value this indicate saturation

Thanks to everyone for your answers, the time you spent preparing them. To give you an update, we have gone back and looked at concentration of capture antibody, concentration of sample and concentration of secondary antidoby in this assay. Basically following a checkerboard system pairing different concentrations of capture in well, secondary applied after analyte binding. We have been able to identify the concentration of capture which provides most reliable results and also secondary which provides linear standard curve. Finally, we run into very high absorbances in sick animals and have started analyzing these with serial dilutions to ensure they are within the range of the standard curve. We run approximately 500-1000 samples per year and the assay works fairly well. When analyzing samples on multiple plates, the CV is

Assuming you are using a normal UV-Vis instrument for your optical density readings you cannot rely on any values above approx 2 - very small levels of parasitic light (depending on which instrument you use) will cause the OD readings to level off and eventually head decrease.

Here is an brief discussion from the book "Introduction to Fluorescence"

The ability of a molecule to absorb light at particular wavelengths can be considered in the context of the Beer–Lambert law, sometimes known as the Beer– Lambert–Bouguer law to recognize the contributions of August Beer (a German physicist), Johann Heinrich Lambert (a Swiss astronomer), and Pierre Bouguer (a French astronomer). Interestingly, Pierre Bouguer calculated the attenuation of light passing through the atmosphere in 1729, which anticipated the work by Lambert in 1760 and by Beer in 1854. The Beer–Lambert law may be stated as A I = = I cl       OD log10 = 0  (2.7) where A is the absorbance of light through a material of pathlength l and concentration (of the absorbing substance) c. The term ε is the molar absorptivity or, more commonly, the extinction coefficient of the absorbing material. The terms I0 and I correspond to the intensity of the light entering and exiting the material, respectively. The term OD corresponds to the optical density. Since the log of intensity ratio is unitless, it follows that the units of e will be M−1 cm−1. As a rule of thumb, the useful range over which the absorption of solutions can be accurately determined is generally about 0.01–2.0 optical density (OD) units, depending on the precise instrumentation utilized. The upper limit is usually due to parasitic light in normal spectrophotometers, which results in deviations from linearity as the OD exceeds about 2 (although some instruments have a higher linear range), while the lower limit is generally due to the inherent noise in the instrument. I should note that the most accurate absorbance measurements are usually obtained near an OD of 1.