Plasma samples with lower OD than blank - (Aug/13/2012 )

Hi all,

I could really use some help for problems with my in-house sandwich ELISA development. I have put together a protocol as detailed at the end of the post to detect a 130kDa protein using polyclonal antibodies. This works very well and generates a linear standard curve with an R of 0.98-99, 1ng/ml of recombinant protein giving an OD around 0.7, with a background (PBST blank) around 0.06-0.07. When I came to run patient plasma samples, I found that, although I could detect my protein of interest, patient controls were consistently generating an OD substantially lower than the PBST-only wells (~0.03). Sequential dilutions (20 fold) of the plasma sample in PBST returned the OD back to 0.06, suggesting something in the plasma is inhibiting detection antibody binding (could it be heterophile antibodies?). My protein concentrations in my patient samples are not high enough to support a 20 fold dilution, so I suspect I need a new standard and/or sample diluent. There are several commercially available products and some that aim neutralise heterophile antibodies. Has anybody encountered this problem and found a strategy or product to overcome it?

Any help would be greatly appreciated!

Tom


Method:
- Coat carbonate/bicarb with mouse antibody overnight (4 deg)
- Block 2% milk PBST (0.05%)
- sample (neat) and standard (diluted in PBST) incubated 2 hours at 37 degrees.
- rabbit detect (milk PBST)
- anti-rabbit HRP conjugated secondary in milk PBST
- TMB substrate

washes performed in PBST (with final washes in water prior to TMB)

The composition of your samples is entirely different from the composition of your curve and it would not be unexpected to have results as you describe due to matrix bulk effect differences.

If you don't have endogenous protein in normal human plasma, then prepare your curve in normal human plasma and you'll be comparing like with like. You may have an anti-rodent antibody effect, but I would expect this to elevate your response in matrix rather than to decrease it. Perhaps consider a 2 fold minimum required dilution (MRD) of your standards and samples (both in human plasma) in something like 1% casein-PBS (Pierce) containing 5% mouse plasma to be safe.

If you don't have a source of human plasma without endogenous analyte, then consider mouse plasma to build the curve (and it shouldn't contain immunoreactive analyte because your capture antibody was made in mouse). Alternatively, try a range of commercial diluents. ICT has a sample pack. It is important to demonstrate parallelism between your curve, and analyte spiked into the study sample matrix.

When you have figured out a suitable curve diluent (human or animal plasma, or commercial diluent) and the need for an MRD or not, perform selectivity experiments in multiple lots of plasma from donors matching your study samples and the blanks should be blank, and the spikes should recover consistently against the curve. If there is endogenous, then the unspiked measured concentrations + spiked measured concentrations should add up.

Is your milk grocery store stuff? If it is, then perhaps move to a commercial vendor casein which has a lot #/expiry data assocated with it.

Also, drop the 37 degree incubations, it usually results in edge effects better to keep everything at lab ambient temperature, go for longer if you need to.

Also, don't know the rationale for the final water wash, especially as your HRP has been happily sitting in physiological buffer since preparation and the brief exposure to zero salt concentration may result in your immuno sandwich to fall apart and wash off.

good luck!

Hi Ben, Thanks for all your advice. Sorry for the delayed reply, I wanted to try out some of your really helpful suggestions.

I quite agree on the 37 degree incubation - on checking, it was giving a significant edge effect.

The wash step in water was a recommendation from a colleague as I was trying to lower my background signal. I'll swap it for a soak step in PBST instead.

I have just trialled diluting my recombinant protein standards in control human plasma, starting with a high concentration and then diluting down serially in plasma. The plasma 'blank' is lower than the spikes, but increasing the concentration does not increase the OD values (as compared to standards in PBST which produces a nice standard curve). Presumably that is matrix effects interfering with antibody binding. I've ordered in casein, heat inactivated FCS and 3 commercial diluents. I was planning to trial 5% FCS, 5%FCS 1%casein and the 3 commercial products as diluents.

Presumably, given the failure of the standard curve in neat plasma, my first step should be seeing if I can reproduce the standard curves in any of the other diluents (it didn't work in 2% milk-PBS). After that, I should spike a fixed amount of recombinant protein into multiple aliquots of control plasma, serially diluted in the chosen diluent, to see which dilution factor gives the best recovery. Would you say that is the correct way to go?

I use TWEEN at 0.05%. Would increasing the concentration potentially improve sensitivity/specificity?

Thanks again for your help!

So are you saying that you get NO signal in 100% human plasma? This is concerning. The performance of the assay in PBS-Tween may be nice, but it is irrelevant unless you are going to replace the blood of the patients with PBS-Tween.

It is essential to demonstrate that you can spike analyte into multiple individual donor lots of 100% human plasma (blank, low and high concentrations) and get consistent blank responses and get acceptable spike recoveries against your curve. This is an assessment of selectivity.

The only reason not to use a curve prepared in 100% pooled human plasma is if human plasma contains endogenous analyte at concentrations above the limit of detection.

What you do after you have your samples/controls/curve in 100% human plasma is the art. Investigate the diluents, using sample (cuve and controls) dilutions of none, 2, 5 and 10 fold (or more) and demonstrate that your individual donor lots of plasma (diluted at 0, 2, 5, 10 etc) have consistent blank ODs, and the spiked individual donor lots of matrix (diluted 0, 2, 5 etc) back calculate acceptably (usually +/- 20%) against the pooled human plasma curve treated in the same way (i.e diluted 0, 2, 5 and 10 x etc)

If you have endogenous analyte in plasma, a surrogate matrix should be used for the curve, and the selectivity experiments described above should match up against the curve factoring in the endogenous + added concentrations. This is also where parallelism comes in, as a high plasma spike, should dilute with surrogate curve matrix in a parallel manner with a high surrogate curve matrix spike diluted in surrogate curve matrix.

I don't think additional tween will change much.

If you have something in the plasma that is specifically interfering with your assay (binding protein etc), we can talk about ways to identify and deal with it.

Good luck

Hi Ben, thanks for the suggestions again. Sorry for the delayed response, I wanted to be sure of my data. I suspect by protein is present at detectable levels in nearly all plasma, though higher in patient populations. I've run 3 plasma samples with spikes of recombinant protein in decreasing concentrations. The curves are all parallel, and when you take into account the different 'blanks' (signal from plasma without spike), they overlap so it suggests the ELISA is detecting the protein reliably in plasma.

As I mentioned before, the signal of recombinant in PBST is much stronger. My problem is selecting a suitable diluent for the standard curve, and a suitable dilution factor for the samples. I've tried a standard curve in 10% FCS PBST (instead of PBST), which does more closely recapitulate the signal in plasma spikes, but it is still pretty far off (OD values in FCS are still higher than in neat and diluted plasma). After trialing 5 different diluents and a range of dilution factors of plasma in diluent (up to 1:20) I cannot find a condition that brings plasma spikes in line with a standard curve (though the spikes are reliable between different donor plasma).

The only option I could think of is purchasing pooled normal plasma to use in my standard curve, and using the same batch throughout all analysis (with the proviso there may be low levels of my protein in there). What do you think? Any other suggestions?

Thanks for your help!



Tom

Hi,

It is encouraging that your plasma lots are parallel and adjusting for blank makes them superimposable. That suggests you have an assay.

Vendors who provide serum/plasma should provide you with 5 mL or so from a larger individual lot to try, and then release the full draw to you if it meets your requirements. If you can find a lot that is relatively blank then this would be the preferable route to go. Don't work with pools in the first instance, especially if you are right in that true blanks are relatively rare in the population and a pool might have the good lot you want mixed with some bad lots. If you want to pool a few blanks after testing them to make sure they match whatever blank criteria you apply then that would be a better approach. Retain some of your good pool so that you will always be able to requalify a new lot when you begin to run out.

If you assess 20 individual plasma lots (for example), run them blank, low and high spike, exclusion criteira might include elevated blank value, or repsonse at either spike not representative of the bunch. Then release the remaining volume of your acceptable lots and use them individually or make a pool.

Good luck!