Dear follow the manufacturer's instructions and look at the conversion factor.

Yes... ideally, the relative levels of cytokines should correlate between the assays. There are several reasons why they might not.

1. Did you run this yourself, or was this done by a lab, either core or commerical?

2. What did you use as the diluent? There are differences in optimizing diluent when you are trying to prevent interactions of antibodies in the plasma with your assays between an ELISA and the Luminex platform. The blocking buffer usually isn't too much of an issue, but the diluent used and how much you dilute can make a difference.

3. The other is... did you use a multiplex kit from Luminex, or did you just detect one analyte? The reason being... depending on the kit, optimization of the protocol depends on the panel of analytes that the kit contains.

4. As for the CV between runs... I figure that the manufacturer should have that information for the kit lot to lot differences. Unless they didn't run QA (but that would be weird...). And any CLIA certified lab should have that information for their assays.

5. Also.. your controls... what are you using? If you really want to control the plate to plate, run to run differences, you need aliquots of a bank of plasma samples that are spiked with a known concentration and unspiked with your analyte. These controls have to fall within an acceptable range every single time on the curve for that run to be accepted. (5% difference is usually the standard...).

6. The dilutions... because ELISAs have an extremely narrow curve (usually only 1 log) whereas Luminex is usually 4 logs... the dilutions for the ELISA have to be very well controlled. The data for the diluted samples should not be on the extreme ends of the curve, since that's usually where most of the variance will be. If they are, you will have to redilute the sample and test again.

There might be more... but this is all that comes to mind right now.

Thanks Shen-An Hwang for that detailed input. Very useful.

1. I ran both platforms myself.

2. Both platforms have their own diluents (manufacturer's). The dilutions were different to accommodate values in the curve.

3. Multiplex was Luminex xMAP 20 analytes.

4. They have CV info and for some anlytes the upper and lower detection std values will have % CV> 10

5.Controls are plasma aliquot stimulated with a known conc. I didn't have an unstimulated control run on the Luminex this time ( may or may not fall in the detection limit for most cytokines).

6. This is the major issue I think. The dilution ranges for both are different and I will have to re-dilute and repeat.

Trying to achieve max information from analytes in luminex the dilution I choose works great but that dilution needs to increase for simplex ELISA.

Would have a suggestive best working ELISA kit and dilution preferred for LPS- stimulated plasma to detect classic pro-inflammatory cytokines, since I am repeating the experiment? I would appreciate that information.

Kalpita

Since you mentioned plasma... is the LPS stimulated whole blood? Or is it really just plasma stimulated with LPS? Are you looking at activation of platelets?

For the ELISAs... the kit I use (mabtech) the upper limit is like 1000pg/mL, and I only ran IL-6 and IFNg. And this is from whole blood stimulated with LPS. I usually run at least two dilutions per sample because not everyone respond the same. So for IL-6 it was 1:100 and 1:200. Most of the samples were readable around 1:200. For IFNg, it was 1:50 and 1:100. I haven't optimized this for my experiments yet, but I have a feeling that it'll have to be probably 3 dilutions by the end.

The unstimulated controls are not really neccesary, I usually do it as the negative control (just in case something weird happens.)

Also, I would use the Luminex results as a guide to what your dilutions will be. Even though the concentration calculated won't be the same, it'll give you an idea where to start.

I must sadly and repeatedly say that both ELISA and Luminex are not quantitative and not reliable methods at all.

IgG (monomer) and Avidin (tetramer) are not specific.

Specificity to ligand increases IgM (pentamer or hexamer) than IgG (monomer) (please see file; The Fascio Effect).

Further, Avidin is recognizing both Lipoic acid and Biotin, and human body fluids usually has c.a. 10-fold higher concentration of Lipoic acid as compared to D-Biotin (please see files; Netherlands Biotin and Lipoic acid Avidin).

Therefore, I must recommend our quantitative and reliable PDMD (Protein-Direct-Microsequencing-Deciphering) method (please see files; JMBT Alopecia and HepG2 Fucoidin).

By the way, human blood has following Cytokine-related proteins as assessed by reliable PDMD method; i.e.,

Interleukin-9 receptor/CD129/Cluster of Differentiation 129 at 4.5,

Epidermal growth factor receptor/EGFR/Proto-oncogene c-ErbB-1/Receptor tyrosine-protein kinase erbB-1 at 3.6,

Antisense basic fibroblast growth factor B/Nucleoside diphosphate-linked moiety X motif 6/Nudix motif 6 at 6.4,

Tumor necrosis factor receptor superfamily member 1A/p55/TNF-R1/Tumor necrosis factor receptor 1 at 1.6,

Metallothionein-1F at 11.3,

Growth differentiation factor 8/GDF8/Myostatin at 11.9,

C-C motif chemokine 17/CCL17/CC chemokine TARC/Small-inducible cytokine A17/Thymus and activation-regulated chemokine at 6.2,

Fibroblast growth factor 13/Fibroblast growth factor homologous factor 2 at 4.0,

Intestine and liver tetraspan membrane protein/Transmembrane 4 L6 family member 4 at 2.3,

Unnamed protein product at 29.5,

Putative testis-specific prion protein/Protein M8 at 5.0,

Steroidogenic acute regulator protein, mitochondrial at 3.1,

GTPase IMAP family member 7/Immunity-associated nucleotide 7 protein at 11.3,

Interleukin-31 receptor subunit alpha at 3.5, and

Protein Wnt5a at 9.1 μg/mg of serum protein, respectively.

Further, LC and HCC liver tissues have following Interferon and interleukin; i.e.,

LC tissue (with leprosy) has Interferon alpha-14/Interferon lambda-2-H at 1.3, Interleukin-32/Natural killer cells protein 4 at 1.5, and Interleukin-4 /B-cell stimulatory factor 1 at 0.97 μg/mg of tissue protein, respectively.

HCC tissue (with PBC) has Interferon beta/Fibroblast interferon at 1.4, and Interferon alpha/beta receptor 1/IFN-alpha/beta receptor 1 at 9.5 μg/mg of tissue protein, respectively.

LC tissue (named as No.6) has Interferon-induced protein with tetratricopeptide repeats 1/Interferon-induced 56 KD protein/IFI-56K at 2.0, and Interleukin-1 alpha/IL-1 alpha at 2.1 μg/mg of tissue protein, respectively.

HCC, tissue (No.6) has Interferon-alpha receptor/IFN-alpha-Rec at 0.41, and Interferon delta-1 at 0.89, and Interleukin-7 receptor alpha chain/IL-7RA at 0.29 μg/mg of tissue protein, respectively.

On the other hand, control normal liver (with pseudo liver cancer) has no IF and IL related proteins.

Therefore, the liver cancer is induced by virus (HIV-1 and HCV) (please see file again; HepG2 Fucoidan).

I am very grateful to Dr. Kalpita R Karan (Department of Psychiatry, Columbia University, New York City, USA) for leading me to this important result.

Shen-An Hwang Thanks for your input!

Mine is LPS-stimulated whole blood. I am not looking at activation platelets in this assay but we have thought about it. The spin of plasma gets rid of most of the platelets if not all (they are too many). I think LPS does not effectively stimulate platelets as they lack nucleus but they may activate them based on LPS conc. one uses?

PS: Update-saw 1 or 2 papers n TLR4 in platelets. I donot have a purified population but in whole blood they may be their during LPS-stim.

I will use R&D systems Duo Set to look at dilutions and matrix effect this time. Will see 1:10 through 1:100 and so on. Hopefully that will lead to a better correlation with Luminex. But I am aware that 1:4 of LPS stimulated plasma was on the highest end of Std curve for IL-6 but I could capture >15 cytokines with that dilution alone on Luminex.

Kalpita

Kou Hayakawa That's a really important detail! Unfortunately with the current sample amount and data I have to correlate I will not be able to utilize any other technique but I will look into it for future work.

Is this method commercially utilized or has people utilizing here in the US? I could explore further in that case?

Thank You.

Kalpita

Since quantitative Edman degradation analysis is only achievable with the Shimadzu PPSQ protein sequencer/microsequencer (Shimadzu Co., Kyoto, Japan). Proteins were bound covalently by the EDC/1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride according to method of Salnikow et al. (Salnikow J, Lehmann A, Wittmann-Liebold B (1981) Improved automated solidphase microsequencing of peptides using DABITC. Anal Biochem 117: 433-442). Labile cysteine was determined after the pyridylethylation and determined as the pyridylethyl-cysteine (PEC/S-[2-(4-pyridyl)ethyl]-L-cysteine). Coefficient of variation/CV/RSD values of PTH-amino acids of Shimadzu have been similar low values to my another independent result (please see file; PTH Roma RP-HPLC).

Therefore, I am now waiting that this reliable and powerful PDMD method is to be utilized among the biochemical researchers in the world.

Kalpita R Karan

Hi, were you able to identify the correlation issue? I wanted to also mention if you knew whether your ELISA antibodies were the same used in the xMAP kit. If not, the binding affinities may be different. In addition, if the sample dilutions are not the same you may need to adjust them in the kit so you are not losing signal. Generally, the expected CVs between runs are within acceptable range. If using a kit, this information should be provided in the manual. Feel free to reach out to [email protected] if you need further assistance.

Jacqueline Surls Hi,

Yes, my guess is the dynamic range and the dilution differences make the best or worst correlations. Due to my experimental design and some technical limitations, I could not di teh same dilution but in Luminex, the dilution worked for most targets on my plate. The challenge is that some Luminex compatible kits have batch variations more than others and having an internal sample as a batch std helps. The only issue is that batch variations not be encountered for the same lot of the Luminex kits (a manufacturer-based issue).