Hi,

I am perfectly agree with Thomas Mohr. You should use always the same serum for negative control or check your procedure. Moreover, the quality of water which you are used to dissolve antibodies, tracer or buffers is very important.

Good luck!

Hanna

Mohamed

If the work is being done in a lab at Ohio state you should be able to control variables such as water and temperature easily. As Thomas Mohr states there should be a constancy in the control you use. A possibility is to use a pooled control i.e. a mix of a number of negative samples from birds to get a larger control volume for repeated tests. We have run ELISA for Salmonella, Pastuerella, Gumboro and Marek's in our field lab in Ethiopia. We used this approach for our negative and positive controls. We did not have the luxury of controlling other variables well, so set a stringent range for our positive and negative controls being within 10% of the mean OD across all runs. It meant extra work (about a quarter of samples had to re-analysed), but this should be achievable under lab conditions in the USA if we could do it in a very basic lab

I agree with the suggestions above, I would further suggest that you include another negative control using the extraction buffer alongside the negative control from the sample serum, just to be able to eliminate the influence of other factors, like water, pH and temperature.. a range between 5-10% of the mean OD in all the runs is acceptable

Thanks all for this suggestions.I usually use 4 different negative controls serum in each plate and i use them in different plates but when i compare the OD of one them across plates,i found OD ranging from 0.09 till 0.2 using the same sample.So,what is the advice to control the procedure or this can considered as biological variation?

In addition, i was thinking aboyt using a pool of the negative serum and run them in duplicate or triplicate.

Definitely run triplicates where you can. Actually a variation between 0.09 and 0.2 is not that large-how does this compare to a buffer blank and what OD are your positives?  If they are over 2.0 then that degree of variation is perhaps not surprising.  I would be covered if they ranged from 0.05 to 0.5, but depending on how a blank and positive runs this may just be the inherent variation in your assay.

i use one positive control per plate starting from 1:50 and serially diluted across the plate.The OD at 1:50 is 2.6 and starts to go down with dilution.Howevere, the degree it decreases is very low especially at low dilution wells and it can possibly explained by the plateau effect.For the blank ranges from 0.039 to 0.06.I use 4 wells for blank per each plate.

Kindly note the attached file for plate layout.

I totally agree with Paul Wigley.. The only thing that can bail you out, is check if there is  variation also for the buffer blanks, if not then there could have been a biological variation in your serum samples for the negative control

Other things that can affect your negative control: plate type (high-binding vs. medium binding plates), plate coating (well-to-well variation in the amount of target protein coated on the plate), blocking solution, and detergent concentration. I have seen cases where high-binding plates have high background signal that goes away when using medium-binding plates without losing signal in the positive samples. I have also seen cases where increasing the detergent concentration or adding an extra blocking step before addition of sample reduces the background. We use 0.2% tween in our ELISA sample buffer, then drop to 0.1% for subsequent steps. Some assays work well with BSA as a blocking agent, some better with milk powder. You may need to play with the concentration of the blocking agent as well.

Hello! I think you should use two negative controls: the first is the culture supernatant without the stimulus you are evaluating; the second is that you must use only the culture medium that was used in the experiment as another negative control. Another thing is that you should avoid using the side plate wells because these wells may have increased readings "edge effect".

Thanks Jarad.I am using high binding plates and also i use 5% skim milk dissolved in PBS as blocking agent.Do yo recommend increasing the concentrationof block buffer?

We also use 5% skim milk. I would look first at adding some tween-20 (we use up to 0.5% for the initial binding step, and 0.1% for subsequent steps), and I would also try some medium binding plates. The plate type made a big difference for us.

Jarad,I will try addition of Tween to the blocking buffer and will see the results.Do you any recommendation about methods for calculating the cut off value to determine the endpoint titer?

If possible, I prefer interpolating to a reference standard. Ideally it would be one where you know the actual concentration, then you can report concentrations instead of titers. If you don't know the concentration, you can just define your reference standard as 100 units, and report in EU, and if you measure the concentration of the standard later you can do a simple conversion. If you really want to do titers, if you have a strong positive sample that gives you a full 4PL curve, you can define the titer threshold as a point low down on the curve, and interpolate to that point.

If you don't have any strong positive sera yet, then I have seen people use either a constant cutoff of 0.2, or 2x the negative control value, or the mean negative control value +2SD or +3SD. Any of them will work, but all are much more sensitive to plate-to-plate variability than interpolating to a standard on the plate. A reference standard helps normalize out the variability.

I have always avoided a constant cutoff due to the variability across the plates, and I use negative control value +3SD. However, the challenge I face with this is when there is no standard deviation on the readings of the negative wells, making the cutoff to be only the value of the negative control.. I have not come a cross any other better method..I believe both of these methods have their flaws.. Any suggestion or idea will be a great help.

You shouldn't calculate the SD based on the wells on a single plate - run at least 10 plates and use the aggregate to calculate your standard deviation. More runs is better, within reason. You can use the mean from the single plate, but an SD from only 2 or 3 wells will not be reliable as you have noticed.

This is also why I really prefer to have a reference standard on the plate - that will do a much better job of standardizing. Even a simple low positive control that you can use as your cutoff threshold will be a more stable reference point than the negative OD plus 2 or 3SD if you are only using 2 or 3 wells to get the SD.

There are lot of good suggestions here, triplicates, buffers, blockers - but isn't the important issue how the negative control, and its reaction affects the cut-off threshold - how do you calculate that? would one approach be to express the results not as ODs but to either correct each sample OD on a plate by subtracting the OD of the negative(s), or to express them as sample/positive or negative ratios, as the case may be?

This gets to the question of what endpoint you are trying to calculate. If you just want a positive/negative answer, you check to see if the sample if above some threshold. If you don't have a reference sample to use to set the threshold, you have to pick something. As I said above, many people used a fixed OD of 0.2. Another common method is taking the negative, measuring it many times to calculate the standard deviation, and going 2 SD above it (which is approximately the 95% confidence interval) or 3 SD above it. But as I noted above, that SD should be based on a decent statistical sample size, not just 2 or 3 wells. These methods are what you do in the absence of a reference. They are not ideal, but many projects have to start with this until they can generate and characterize some good reference sera.

The next step up is to calculate discrete titers - what dilution of a sample is the lowest dilution that crosses the threshold. This gives a crude measure of concentration. You can refine the titer method by using curve-fitting and interpolating between dilutions to calculate a continuous titer, which will improve the resolution of the measurement, although it is still fairly crude.

In order to get good precision out of an ELISA, you really want to have a reference standard. Even for positive/negative results, setting the threshold based on a reference sample is much less variable than using fixed OD thresholds or negative + 3SD. Using a reference standard to set and normalize the titer threshold, or better yet to interpolate to the reference and calculate relative concentration gives better precision and accuracy than thresholds based off the negative control.

What Jarad says above is absolutely spot on.  For us the field ELISAs were more reliable once we had a better definition from around 20 repeats. Dried skim milk works very well in chicken systems and is cheap and easy. We tended not to use tween in blocking steps, but it is certainly worth trying

Just forget all the blocking. Milk proteins can be recognized by different individual differently. The detergent in the blocking buffer will decrease the test stability during storage. 

1. to avoid temperature gradients over the plate, make sure that all buffers are at room temperature. Don't incubate at 37 °C

2. To decrease unspecific reaction you can increase the salt concentration in your incubation buffers (normally in PBS NaCl is at 0.15 mol/L, just go up to 0.3 mol/L). The incubation buffer should contain some inert protein. This can be BSA, HSA ... or working in most cases a hydrolysed gelatine which is available as a drug for infusion as plasma expander in patients. That's a standardized controlled product and gives only problems in some ELISA for the detection of antibodies against collagen.

3. Use a modular buffer system: washing buffer: BSS + 0.1% tween 20 (the increase higher that this is not that successful), incubation buffer (for sample dilution, sample in conjugate incubation): washing buffer + 1 ... 3% hydrolysed gelatin + 0,1 mg/mL phenol red (to see much better where you are currently pipetting)

4. search for an optimal sample dilution. The incubation buffer is also to keep the incubation conditions constant despite of the individual samples. The best sample dilution is the one with the highest P/N-Ratio (positive/negative). Try this in a checkerboard titration. Commercial assays using often a sample/cut-off-Ratio. It's comparable but for this, you need to get your cut off calculated.

5. Calculated the cut off value with the help of the Youden index. Y = (Specificity + Sensitivity)-1. Plot the Y on the OD ... and you will see an curve with an optimum. See also the attached paper.