Hi Rob,

Thanks for giving so many details, this helps the community a lot!

Let me start by saying I don't have fibronectin ELISA experience. I do know that different proteins stick to different materials... er... differently. It's actually quite tricky to try to predict what kind of surface will be sticky for a particular protein. My suggestion would be to try different plates (high-bind, low-bind, mid-bind, etc.). Hopefully someone here has material a suggestion for you. You can also look up antibodies for fibronectin and see what kind of protocol/materials they use for them. If that doesn't work, at least you have an antibody for fibronectin you can buy to test if it is sticking to the plate.

One comment, I see you coat with 100 μL and then block with 100 μL. I typically block with a larger volume (~250 μL) to make sure my protein doesn't see unblocked surface during rocking. It might be a way of reducing your background.

Good luck!

@Jeremy Pronchik thanks for your thoughts, I'm repeating the assay with a different plate that colleagues have used for cellular studies using fibronectin, so hopefully trying a new plate will help!

You could try chemically activated plates to covalently bind amino-or carboxy- groups of your protein instead of relying on hydrophobic bonds only.

However, in my experience, dot blots are easier to perform, more sensitive and more reliable than ELISA, with chemiluminescence detection they have a linear range of 5 orders of magnitude. Use a 96-well vacuum manifold to suck your samples through a PVDF (not NC!) membrane, marking the top left and bottom right well. The protein will bind to the membrane near quantitatively (binding to ELISA plates is usually only 3% or so). Then develop the blot like a western. I use a membrane holder made in a craft-shop to detect the signal in a luminescence counter.

Hi Engelbert Buxbaum,

Your comment caught my attention and I'd like to learn more about your dot blots. I did a little journal research and can't find any dot blots with a 5 order of magnitude linear range. Could you share some data/conditions where you have seen this? Is vacuuming critical for achieving this linear range? For the precision and reproducibility?

Sorry, Rob Barringer, to hijack your thread!

The vacuum manifold ensures that all spots have the same geometry, and that the spots coincide later with with the sensitive areas of the reader. Thus, the results are more reproducible than just spotting samples with a pipette. In addition, the possible sample is 100 times that of a spot blot, where you are limited to a few µL. That can be important when you are working at the limit of detection.

The critical aspect is that you need to use a chemoluminescence counter (from experience, I can recommend https://www.berthold-bio.com/microplate-readers.html, but that doesn't mean the others are bad, I just never worked with them). If you use photographic film to record the luminescence, you are limited to the difference between the lightest and darkest gray the film can distinguish from white and black, which is usually 1-2 orders of magnitude. The same is true for ELISA, where you are limited to the absorbance range of several 10 to several 100, where Lambert-Beer's law is valid. You can extend this range a little by using standard curves rather than standard lines, but not by much.

PVDF membranes of course need to be pre-wetted with MeOH and then washed with dd water and finally buffer, before being used. Other than that, the assay is straight forward: I used blotto (PBS with 0.05% Tween 20, 0.1% fat-free milk powder and 0.01% Timerosal as preservative, don't use azide as it inactivates HRP!) for developing and 5% milk powder in blotto for blocking (15 min RT), and HRP-coupled secondary antibodies (Dako (https://www.agilent.com/en/product/immunohistochemistry/antibodies-controls/secondary-antibodies#3) and Accurate (http://www.accuratechemical.com/products) have proven reliable in my experience, with the caveat mentioned). Primary antibody o/n in the cold room on an orbital shaker, secondary 37 °C, 1 h. The primary can be re-used many times, for the secondary I use fresh dilutions every time. After each antibody, wash 3 times 15 min with blotto at RT on an orbital shaker. For the reagent, you can use commercial mixtures or make your own (cheaper!).

Thanks Engelbert Buxbaum, it sounds like you are cleverly making a 96 spot sheet and reading that in your plate reader. I like it. While reading about dot blots, I found this company in China that makes white-walled plates with NC in the bottom. Their website isn't loading for me right now; here's their paper: Article Quantitative dot blot analysis (QDB), a versatile high throu...

I'm thinking about the limits of these assays for the first time, and I think I've convinced myself that no assay will really go beyond ~2 orders of magnitude sensitivity as long as we have ordinary binding.

I have attached a plot I just made of four binding curves. The ones with circles were made with Langmuir's adsorption equation:

fraction bound = K*c/(1 + K*c) ; K = binding constant, c= [ligand]

There are three binding strengths that go as Yellow > Green > Blue, and plotting these log-linear they all span the same number of orders of magnitude (~1-2).

If we use the Hill Equation:

fraction bound = K*c^n/(1 + K*c^n)

and the steepness of the slope can be varied with n. I need to use an n ≈ 0.3 (red triangles) to have the curve span 5 orders of magnitude. This implies a strong degree of non-cooperativity in binding.

How does that sound to you? I trust the instrument can detect 5 orders of magnitude, especially if you can adjust the gain, but have you ever seen 5 orders for a single assay?

Jeremy Pronchik : Yes, such plates are available from several manufacturers, see for example http://www.merckmillipore.com/DE/de/life-science-research/cell-culture-systems/cell-growth/multiscreen/ELISpot-optimization/Plates/qkeb.qB.BrUAAAFPiVVFgZQx,nav, here even with PVDF membrane. PVDF has a higher protein binding affinity and capacity than NC, and is thus much more suitable for this application. In fact, I never got the described assay to work with NC membranes, the standard deviation was too high. Thus, the extra money for the PVDF-membrane is well spent.

The ready-made plates are of course convenient, but IMHO suffer from two disadvantages:

One thing I forgot to mention in my description: The filter manifold needs to ensure complete separation of the wells, you need one with a silicon mat seal. Those that simply use filter paper are much less reliable. Mine was from Bio-Rad.

For data spanning several orders of magnitude in both dependent and independent variable, it is advisable to use a log/log plot. Then you can just about get away with 5 orders of magnitude. Of course, the reproducibility will be better in the middle range.

Engelbert Buxbaum, thank you for all of the time you put into my questions, I appreciate it; I understand your usage of linear range now and agree the ends will suffer the effects of things like pipettor/dilution errors and photon noise making them more difficult to work with.

Thanks for all of the helpful links and getting me thinking about these membranes more deeply!