Better study resources for understanding ELISA chemically?
September 30, 2012 4:11 PM   Subscribe

Better study resources for understanding ELISA technology?

I'm taking a class in ELISA (Enzyme-linked immunosorbent assay) and I've got a final coming up. I definitely understand how to run an ELISA at this point but I'm having trouble putting together all the details of what's happening chemically at each step. We don't have a textbook for the course as my professor says there's tons of information available on the internet.

But I'm stuck because while I've mastered the wikipedia page on ELISAs, I don't have the mid-level understanding that will enable me to get good information out of the actual technical literature that I'm finding on say, pubmed.

I've downloaded an entire immunology textbook and I've been going through that, but it only has one page on ELISA.

Any suggestions for resources?
posted by sunnichka to Science & Nature (13 answers total) 3 users marked this as a favorite
No specific suggestions, but perhaps ask your professor? If (s)he thinks there's "tons of information available on the internet," (s)he should be able to point you to a good intermediate-level source of info.
posted by Betelgeuse at 4:35 PM on September 30, 2012

My professor is not amenable to that kind of question.
posted by sunnichka at 4:44 PM on September 30, 2012

ELISA is one of my dream Jeopardy categories, so any sticklers you have, feel free to send me a MeFi mail.

Before I send you a million different weblinks that go places you really don't care to go, are you expecting any math on the test? Or is the test just going to be about running assays and interpreting data? What different types of ELISA have you discussed?

Typically ELISA is all about antibody-antigen binding, so a good understanding about that will probably be helpful to you. I can imagine a whole raft of evil questions regarding monoclonal and polyclonal antibodies. And about choice of secondary antibodies.
posted by Kid Charlemagne at 5:20 PM on September 30, 2012 [1 favorite]

What is the focus of the course? The physical chemistry of Ab-Ag interactions (e.g., calculating affinities and on/off rates) or the practicalities of setting up and troubleshooting assays? If it's the former, you need some P-chem background which Kid C can refer you to. If it's the latter, a lot of reagent suppliers have pretty good information in their online catalogs. I always find Pierce (part of Thermo Scientific) a good place to start: try here.
posted by Quietgal at 6:08 PM on September 30, 2012

It will help us to know what kind of literature you are trying to understand,

A public health or diagnostic kind of focus with ELISAs done on human serum?
Basic research kinds of focus where you'll need to have a solid grasp of the crazy shit you can do with sandwich, competitive, and multiple ELISAs?
A P-chem/bio-chem kind of focus that Kid Charlemagne would be really perfect for helping you with?

There is tons of information on the web so with a better understanding of what you need we can find stuff that is better suited to you. What is happening chemically at each step is all about the 3Dimentional structure of the business end of the antibody and the antigen which causes the specificity in binding.
posted by Blasdelb at 11:46 PM on September 30, 2012

Thanks for your answers!

The link that QuietGal gave me is a perfect first next step. The first thing I need to know is troubleshooting, troubleshooting, troubleshooting.

But I get frustrated with all the little diagrams where the substrate is a circle, the antigen is a triangle, the antibody is a little stick antibody. I want to actually understand how the 3D structure of the antibody and the antigen causes specificity in binding, like you said. It is just the most frustrating thing in the world to me when my professor says to memorize that "the substrate binds to the HRP conjugate," "a full-length IgG can be digested with pepsin to form the FAb'2" and "BSA is a permanent blocker." I know that on the test I'll just be expected to spout off those exact phrases and maybe draw some little BSA circles around a little triangle that says Ag, inside of a little rectangle that represents my well. BUT F*CK THAT MESS, I WANT TO ACTUALLY KNOW WHAT IS GOING ON. And I am having trouble finding that kind of information in a way I can take in.

I have had organic chemistry but not physical chemistry, but I've been through CalcII and I'm not afraid of working my way through some math even if it's slightly over my head.

Kid Charlemagne, I would LOVE to have your weblink dump. It might take me a while to get through it but I am definitely interested in going those places.
posted by sunnichka at 12:11 AM on October 1, 2012

My field is slightly different, so I could be mistaken. However, I think that the some of the phrases you quote here fall into the category of science speak for this is what people have observed, but no one has any idea why (e.g. why BSA is a permanent blocker) or even this is a huge area of research that people are making progress on, but still has open, outstanding questions (what causes specificity in protein protein interactions).

Have you tried looking at a basic biochemistry textbook? I think that's where you'll find the best descriptions of what leads to specificity, probably some more information on what pepsin does, and a description of how secondary antibodies work (you might look for information on westerns to see this idea in a different context).
posted by lab.beetle at 1:49 AM on October 1, 2012

"I want to actually understand how the 3D structure of the antibody and the antigen causes specificity in binding, like you said."

Simplistically, antibodies have all kinds of functional groups sticking out of them with all kinds of charges that recognize opposite charges on their respective antigens - that is in addition to other kinds of intermolecular forces. I tend to think of them metaphorically as intricately shaped magnet keys with positive and negative charges all over them that will only bind to the similarly intricate corresponding magnetic lock.

If you are thinking about accurately modeling the structure of specific antibodies, oh damn are you not alone.

This is generally done by x-ray crystallography and it is a pain in the ass. Here is an example of a very pretty picture published in PNAS

If you have a specific antibody you'd like to model on the cheap and have a sequence for it here are some tools you can use, WAM - Web Antibody Modeling or Prediction of Immunoglobulin Structure (PIGS) to get a rough idea.
posted by Blasdelb at 3:57 AM on October 1, 2012 [1 favorite]

I want to actually understand how the 3D structure of the antibody and the antigen causes specificity in binding, like you said.

Are you and I related? Because deep down inside I feel like this all the time. That said, you're not going to be able to pull this off unless we get about four times as close to a Vingian singularity as we are now between now and your final. In the field you are pretty much at the mercy of the whims of some critter's immune system and all the technical knowledge in the world isn't going to allow you to change things.

Basically the paratope on your antibody and the epitope on your antigen will bind more tightly the closer they are to mirror images of one another both topologically and electronically. Every place there is a bump on one, there is a divot on the other and every place one is positively charged, the other is negatively charged. The hydrogen bonding down the center of a DNA molecule is like this is a linear spiral rather than over a 2-D surface, but the principal is the same. Binding sites that have the wrong charges and physical topology won't stick as tightly as those that do. (An exception is that you can have water bridging, say two divots up against one another, both slightly negative, both interacting with the two slightly positive sides of a water molecule. (Figure 4 of this paper shows this idea in 2D for smaller molecules.) Oh, and there is also hydrophobicity which helps drive things so that even areas which aren't charged at all help drive things.

The next level of thinking about this brushes up against the immune system. Developing the electronic topologies on antibody binding sites is not a directed thing. Affinity maturation is a trial and error process. And your immune system has a vested interest in killing an infection fast so it doesn't have time to screw around with making ideal antibodies that perfectly bind the antigen. Just "good enough." To make an analogy here, you wouldn't want to write your class notes on a block of ice because they'd be gone by time you needed them. Nor would you want to inscribe them on a piece of hardened steel plated over with gold. Baring disaster, your paper notes will last for about as long as you're likely to need them and then some. For antibodies, somewhere around 1 nM-1 is the immune system sweet spot. (And here's the very best two page paper on this subject in the whole wide world.)

Why do antigens and antibodies stick to the plate? These same forces are at work here thanks to the way the plates are treated creating charged zones pretty much randomly over the surface of the plate.

The last thing you need to remember is that if you have an antibody and an antigen in a well, the range of their interaction is VERY VERY SHORT! Most of the assay is diffusion mediated. If you want to really model what's going on in your plate (and you usually don't) you need to run Receptor-ligand kinetics into the Stokes-Einstein equation describing diffusion and sprinkle on some probability theory and mechanics to adjust for the orientation of the plate bound ligands. Don't forget to factor in the fact that your reaction isn't taking place free in solution but is being done on a two dimensional surface.

All of this said, this is all probably a much lower level of understanding than you are likely to need unless you're playing very close to your theoretical limit of detection in a overly complex assay. For purposes of your test, this is going to be a lot like brushing up on the ideal gas law for a class on automobile repair. At some level it matters, but you can pretty much count on it not being on the test.

The first thing I need to know is troubleshooting, troubleshooting, troubleshooting.

OK, let's talk troubleshooting. The first thing to remember is that when you're doing an ELISA, you don't get to find out that you screwed up until the end of the day when you put your substrate on. So, depending on what you're plate looks like, there are a number of things that could have gone wrong to cause that kind of error. Also, there are issues of execution, where you screw up a known functional assay through contamination, forgetting a step, reagent error or bad technique and issues of design, where you're trying to develop a new assay and you've squeezed everything up against your background or upper asymptote by choosing too high or low of concentrations. The troubleshooting tree is going to be different for these two types of situations.

I'll go digging for some links for you.
posted by Kid Charlemagne at 11:41 AM on October 1, 2012 [1 favorite]

WOW! Are you a professor? I wish you were my professor. I would love to have more links.
posted by sunnichka at 4:03 PM on October 1, 2012

No, just a big geek who likes to take things apart to see how they work and spent a number of years doing ELISA for host cell proteins in Bio-pharmaceuticals. That said, looking at the links in my library, most of them are pretty focused on some oddball topics that were of interest to me and probably won't help you much at all (thought I did just ping some total stranger to see if I can get a link to her PhD thesis for you!)

Here's a good trouble shooting list from Abcam. Here's another from R&D systems. "ELISA troubleshooting" gets you a lot of links on

In the perfect world you run a standard curve which you dilute from a stock standard and then have a low and high control sample which is well characterized, known to be stable and which you can just load onto the plate with minimal handling. That way, you have a degree of compartmentalization and when you get to the end, if everything is a mess except your controls you know it was something you did when you were making up your standards and samples (the part of the process with the most opportunity for screwing up, particularly since you'll have your concentrated standard out) vs. some other phase of the operation.

In my experience the most common failure modes for assays that were working before are:
  • Forgetting to do a step entirely.
  • Doing a step with the wrong reagent - In my group we used color coded caps - red for standard, green for primary antibody and blue for our labeled secondary antibody. If you ever screw up and coat your plate with your biotinylated antibody it'll ruin your whole day.
  • A plate washing problem - You have to wash them thoroughly, but, well, BSA is only a permanent blocker until you really crank your plate washer or forget to replace the storage buffer with your wash buffer. Or if you're measuring a bacterial impurity and you have something growing in your lines. Or if you have a clogged pin so that some wells aren't being washed properly.
  • Contamination - I once got a free trip (for me, not for my corporate masters) to Austria to troubleshoot an assay where the difference between 1 ng/mL of analyte and 4 ng/mL of analyte was pretty vast. We shipped it there and it didn't work. See also stuff growing in your plate washer lines.
  • Bad reagents - storing dilute solutions of antigens and antibodies for any length of time is a bad idea. See my comment here. If your substrate is starting to change colors before you apply it to the plate, maybe it's time to open a new bottle. Never pour anything you aren't 100% sure of back into your buffer bottle.
  • Samples in weird matrices - high salt, high or low pH or a viscosity that wildly differs from that of your standards. Running a spiked replicate really helps here.
  • Denatured samples - if you're using polyclonal antibodies it doesn't much matter since most of them will be against primary structure and bond more tightly if the sample is no longer neatly folded, but with monoclonals you can have cases where your epitope gets buried or is broken up by the denaturation of your analyte and your antibody no longer recognizes it or recognizes it with a different affinity. This sometimes presents itself as non-linearity.
Have you guys talked much about spiking? It's useful as a troubleshooting technique and as a go-no go gauge for precision.

For that matter, how much have you talked about precision?
posted by Kid Charlemagne at 6:28 PM on October 1, 2012 [1 favorite]

We've talked about precision a little bit. I've done some of my own math on my pipetting precision between wells, and my coefficient of variation is generally under 5%. We've talked about sensitivity, how to calculate to insure that the lowest standard is distinct from the zero standard. I don't know anything about spiking; what is that?

I'm reading your comment about your Austria troubleshooting trip: First, ng/mL is equal to parts-per-TRILLION, correct? I didn't realize that ELISAs could be that sensitive, that's neat. (That makes me wander off topic and start thinking my hobby, wine chemistry, where many of the volatile compounds that we can detect by smell are only present in the wine in parts-per-trillion concentrations. ) BUT to stay on topic, I can see how easy it would be to contaminate your samples when you're talking about a 3 ng/mL difference, but how do you think the contamination was actually happening in that Austrian lab?

I've read about denatured samples on the Molecular Station ELISA forum, and so I know that detergent salts from the wash steps can dehydrate and thus denature your proteins unless a tiny bit of the wash buffer remains in each well.

How do you end up with samples in 'weird matrices'? Do you mean that you've diluted your samples in something too acidic, too salty, or too viscous? I understand that acidity is a problem because the plate well material, polystyrene, is itself acidic, so the coating buffer needs to be strongly basic to counteract that. I understand that too much salt is a problem because it denatures proteins. But how does the viscosity cause problems?

We've had a lot of trouble in our class with the substrate starting to change color even in the trough before we add it to the wells. We'd discussed pipetting and wash technique--that if we drag the pipette across the top of the well on the way out, a little bit of conjugate can remain there at the top of the well and not get washed out thoroughly in the plate washer. Then the substrate will bind with it there at the top of the well and fall into the well, causing high background or a false positive. But we hadn't discussed the possibility that the substrate solution itself might be going bad. I think our solutions are donated from a drug development company so it would make sense that they might not be in the prime of their youths.

And thank you for the troubleshooting links too! I um, don't know why it hadn't occurred to me to specifically google "troubleshooting ELISA."

I'd love to look at the PhD thesis if you hear from the total stranger.

(comment to be continued)
posted by sunnichka at 7:54 PM on October 1, 2012

Ppm is a confusing thing because, well, per million parts of what? The way we would report, we'd test our drug at 1 mg/mL and so 1 ng/mL would be one ppm by our accounting - one part impurity to 999,999 parts drug and we could care less about how much buffer was in there.

I'm still not sure of the exact vector in Austria, I just know that everything I saw the lady doing the analytical work do looked good. On the last day we were there in the lab, doing our thing, and a guy came in from behind the giant blast doors of the purification room (I exaggerate a skosh, but just a skosh) to check a pH. I found myself sitting there wondering, "Whats in that beaker? If someone spilled 50 mL of bulk drug product at 5-7 mg/mL all over the bench, and I'm over here thinking the difference between 1 and 1.5 ng is significant.... Hmmmm." Truthfully, it could have been anything though, like somebody sucking material up into a pipette.

For us, weird matrices were usually elution buffers from ion exchange chromatography columns and the issue was usually high salt. Extremes of pH will make antibodies let go. In fact, when you do affinity chromatography with your antigen bound to a resin to purify antibodies from serum, you typically used something like a pH 3 acetate / glycine buffer to elute the column and then do a buffer exchange to get the antibodies into something more friendly.

We typically used Dulbeccos PBS with 0.1% Tween 20 in it as a wash buffer and as a diluent. Sometimes we'd use this with 10% blocking buffer (I always used Pierce Superblock in the assays I developed).

The problem with viscous samples, well, remember when I said ELISA is diffusion mediated? I'd have to look up the exact equation to be sure but I seem to recall that if you double the viscosity (like say you have a drug sample that is 1 mg/mL and a standard diluted in just PBS/Tween) then the signal in your more viscous sample is going to be reduced by a factor of 1.414. This is why it sometimes makes sense to dilute in a diluted blocking buffer so that everything is in a 1 mg/mL protein solution.

Spiking is adding a known amount of your analyte to the material your testing to see if you can recover the spike. So, let's say you're diluting a sample of serum to 1000 fold prior to testing and putting 100 µL of diluted sample in each of your replicate wells. If you wanted to run with spiking, you'd diluted it 500 fold, then put 50 µL into two sets of replicate wells, then into one set you'd add 50 µL of your diluent and into the other you'd add 50 µL of a standard from the lower middle of your curve. (For what I mostly did, I typically used my 16 ng/mL std.) So now your assay is done and you want to see how well you did. You check your sample and it's 12.7 ng/mL. Ideally your spiked replicates will be 20.7 ng/mL because you added 8 ng/mL of standard (16 ng/mL*50µL/(50µL + 50µL)). Of course there would be some error, but if you chose reasonable concentrations in the strong part (precision wise) of your curve, you should usually be able to get within 25% of your target.

Oh, and here's a pipetting trick. Always approach your target number from above. The reason why is backlash and it's an issue with everything involving screws. (Like metal lathes and micrometers in pipettes.)
posted by Kid Charlemagne at 10:08 PM on October 1, 2012 [1 favorite]

« Older Image link with some examples....   |  Can you tell me (or make an ed... Newer »
This thread is closed to new comments.