Hi Ekaterina,

It does sound extremely unlikely that all four primer pairs are incapable of producing a passable result. So, I think you're right that something else is going wrong. Here are some ideas for troubleshooting to work out what it could be:

• As you mentioned, inhibitors preventing the PCR reaction. One way to get an idea is to run your cDNA samples on Nanodrop, the A230, A260 and A280 ratios can give you and indication of purity and hence contaminants.
• It could be that something is wrong with your qPCR machine. Is there a positive you could run? Doesn't even need to be the same samples or primers, just anything you know has worked historically. Though as you're getting strange amplification patterns with the gradient PCR too, this seems unlikely - unless there is a technical problem there too. Did you run these both with positive controls?
• Contamination in your reagents. Did you run negative controls?
• Are you certain your procedures worked up till cDNA production - ie. is there enough/functional template for the polymerase to work with? An easy way to check is by just running a gel and checking the integrity of the mRNA (or cDNA) and looking for the characteristic  rRNA bands at 28S, 18S and 5.8S. If they are present and prominent, that's an indicator that the RNA survived the procedure. Though I would highly recommend sending your samples to a bioanalyzer to see what the quantity and quality of your RNA is. To my knowledge this is the best way to check whether your is good to go for downstream processing.

Some of these troubleshooting strategies are very basic and I realize that you may have already done some or all of them. But it's a good place to start, whether for you or a discussion on how to address the problem. Best of luck!

" Also, if primer design was bad - could it be bad for 4 primer pairs simultaneously?"

Yes of course, it's like the dough in the pizza, if the dough is bad so will the end product (pizza)! 

What is the annealing Tm of your primers? Generally you should be using annealing temperatures of about 5°C below the Tm of your primers.

In my experiences, quantification beyond 30 Cts is problematic/unreliable due to high variances in Cts (stochastic variations). What kind of ladder are you using. I find it hard to believe that you can use 1-fold change ladder down to 1 copy accurately. This requires perfect pipetting! I noticed that in your standard curve you have 2-3 Ct differences, this translates to 2^2=4 and 2^3=8-fold differences whichc cannot be ignored. What is your standard curve template? Is each bar graph different target? I also noticed that in some curves you have 3, then 4 and in others 5 points. What happend there? 

Have a google search image for "pcr standard curve template" and see what people use. Also have a read of this

https://www.thermofisher.com/tr/en/home/life-science/pcr/real-time-pcr/qpcr-education/real-time-pcr-troubleshooting-tool/gene-expression-quantitation-troubleshooting/poor-pcr-efficiency.html

"R^2 were good (>0.98 for every primer pair), and all the curves were almost parallel, but efficiency was merely 48%."

Could you attach a picture of the amplification curves (ideally the data) of these dilution series?

Anna, thank you!

About NanoDrop: as I said, I didn't purify the cDNA before cycling, so there's no point in analyzing it - there would be remaining random primers, RNA, DNAse I, M-MLV and DNAse stop solution (EGTA), and also some salts and other buffer components. Apparently these adulterants don't normally hamper PCR because its common practice to use non-purified cDNA. But now I'm considering to actually purify it using magnetic beads and check if there is a difference. I checked the extracted RNA on NanoDrop, and 260/280 was about 2, but 260/230 ratios were not so good and different in different samples: from 1.68 to 2.27. In all samples there was some very high peak about 220 nm, I don't know what is this (definitely not protein and not phenol).

About PCR mashine: people use it every day and no one complains. I myself performed qPCR with some other primers and cDNA from rat's cultured neurons a month ago (I extracted and reverted RNA from these samples myself using the same protocols I used for the snail samples), and one primer pair had efficiency 95%, the other - 80% (and latter was  perfectly explainable, because this second primer pair had predicted Ta 4ºC lower than was used in cycling). I also asked my colleagues if they had problems with calibration curves recently. They said that they don't calibrate in every run if they don't change the conditions, and last time they calibrated was a few weeks ago, and efficiencies were high.

I did run two negative controls for every primer pair in every run using MQ water instead of cDNA template, and there was no amplification (well, technically there was in a few cases, especially when low Ta and many (50) cycles were used, but never in both replicas and with Ct>35, even >40; I analyzed these samples on agarose gel and there were very short products, about 50 bp, and I think that these were primer dimers, not actual contamination).

I didn't check my RNAs on Agilent Bioanalyzer (now I'm considering to actually do this, too). But I use Bioanalyzer to check RNAs in other experiments (with RNA-Seq), and I know that RNA extracted with Trizol is ofted degraded (fragmented, with smeared rRNA peaks and additional peak about 100 nt). I strongly suspect that my snail RNAs were degraded because they were obtained with Trizol and I'm not very good at Trizol extraction. But my designed amplicons are quite short, 90-140 bp, so I doubt that fragmentation would do a lot of harm in this case. I also perform gel-electrophoresis after every qPCR amplification, and the lenghts of the products are as expected (again, when I used 50 amplification cycles, bars were more smeared indicating unspecific products, but for 40-45 cycles there were no smears).

(By the way, undegraded total RNA obtained from snail should have only one 18S peak, unlike mammalian RNA, because snails normally have a cut in the middle of 28S rRNA, and every half of 28S has almost the same lenght as 18S).

Can, thank you!

"Yes of course, it's like the dough in the pizza, if the dough is bad so will the end product (pizza)!" Well, you know, you may be right because my primers are specific for the snail FASTA sequences I used, but these sequences themselves may be incorrect. Helix lucorum snail had no annotated genome and no annotated transcriptome last time I checked, and I used unannotated transcriptome assembly we made in our lab, and I aligned genes of other molluscs to this homemade assembly to find homological fragments. My primers could have some mismatches I don't know about because there's no RefSeq for these templates. My colleague has told me that she had similar problems with guinea pig genes a few years ago. But to be sure that primers are bad I must sweep aside other possible problems first. Every primer pair has passed initial checks in silico (Primer-BLAST, Multiple Primer Analyzer) and in vitro (they produced bars of expected lenghts on agarose gel).

I designed primers using Clone Manager, and this program predicted Ta=60 for three primer pairs, and Ta=66 for the fourth. (Don't tell me, I know that it's really bad to run such primers in one plate. But there were very harsh restrictions of primer placement - we need to distinguish two splice isoforms, so one pair had to be inside a very short intron, the other had to have a primer with left half immediately before the intron, the right half immediately after it. When I tried to change the lenghts/positions of primers to equal their Tms, primers became either not placed right or very prone to 3'-dimerization. In the end I just decided to use compromising 63°C or so and put up with 80% efficiency). I also checked four other Ta calculators, and they gave me different figures from 54 to 62°C for three pairs and from 59 to 68°C for the fourth. I tried to use 63, 59 and 61°C in my runs so far (I don't count a single gradient experiment here), and 63°C was the least crappy. 59°C was the worst: I had not only poor efficiency, but also very bad calibration curve (two the most diluted standards weren't amplifyed at all, others had huge dispersion of replicas), and more smeared bars on gel (but I also used 50 cycles in this run, this encouraged unspecific amplification).

For relative calibration curves attached here, I used as a "ladder" serial dilutions of an actual experimental sample of my snail ganglion cDNA. It's called "relative" because I don't know the exact number of DNA copies in my standards, and I don't have to, if I only need to measure expression change between samples in folds. Usually for relative calibration curves people use dilution factors 2, 4 (because it's easier to calculate log2) or 10 (for broad range). The dilution factor I used here was 2.5, it's not very handy and easy for mental arithmetic, but I calculated that the factor of 2 would give me very closely placed points, and the factor of 3 would give too diluted standards that won't amplify at all (I found that my cDNA shouldn't be diluted more than 1:100, there's not much template). So, here are dilutions 1:2.5, 1:6.25: 1:15.63, 1:39 and 1:97.7 of the same (problem?) cDNA as standards. With hypothetical 100% efficiency dCt between the points should be log2(2.5)=lg(2.5)/lg(2)=1.32 cycles. Each of 4 graphs here represents different target. Did you mean differences between different graphs when you said "2-3 Ct differences"? In this case they can be ignored because it's normal to expect different quantities of different cDNA templates in the sample. It was the whole point of experiment to compare quantities of these templates. There isn't 2-3 Ct differences between replicas on the graphs (sometimes 1.5-2 Ct differences between replicas were observed, for the most diluted standards, but I had to exclude these replicas from analysis).

I used 5 dilutions, but in some cases 1 or 2 most diluted standards didn't amplify (or there were big dispersions and I had to omit these points), hence different number of points in different curves.

" know that RNA extracted with Trizol is ofted degraded (fragmented, with smeared rRNA peaks and additional peak about 100 nt)."

I've worked with Trizol extraction in tissue, it shouldn't be causing this problem because "TRIzol works by maintaining RNA integrity during tissue homogenization, while at the same time disrupting and breaking down cells and cell components."

Is it possible that your samples were degraded prior to extraction? Get your RNA integrity checked, then we can ponder about cDNA and PCR reaction if problems persist. 

Dear Jochen, I attach the amplification plots (sorry I couldn't get larger pictures, so there's only bottom fragments of sigmoid curves for you to see how the curves cross the threshold; they all did reach plato phase) and Ct data here. I omitted some replicas for which no amplification was observed or which were definitely outliers. I didn't omit the last (most diluted) point for TBP, but I should have done this - today I checked one of 3 amplified replicas on gel, and there was unspecific product; the fourth replica didn't amplify.

In this table "quantity" 2.5 means dilution 1:97.7, "quantity" 6.25 means 1:39, and so on. Only the ratio between adjacent points must be exact, in other respects I could set any numbers I like.

To Can: Trizol definitely can fragment your RNA during extraction. I attach here some manual of RNA integrity assessment, and I should add that I myself frequently observed graphs like the third picture here after Trizol RNA extraction. I suppose my samples weren't degraded before Trizol because ganglions were alive all the time right to the point of  homogenization. I didn't prepare ganglions myself, but I totally trust the man who did this for me, he works with these preparations for years. They still can be degraded BECAUSE of Trizol. I plan to check this, but as I said, I have short amplicons, so fragmentation possibly isn't so bad in this case.

Also to Can: "It is generally thought that high-quality, undegraded RNA is required for all types of analysis. Though this may be true for microarrays or for gel-based protocols such as northern blot analysis, it is not necessarily true for qRT-PCR. Many of the primer/probe sequences used for qRT-PCR are quite short, ranging from ~50 bases to ≥200 bases, allowing even significantly degraded RNA to be used for analysis, without detectable effects."

https://www.thermofisher.com/ru/ru/home/references/ambion-tech-support/rtpcr-analysis/general-articles/rna-isolation-for-qrt-pcr.html

First link does not suggest that trizol degrades RNA, it shows what  Trizol looks like on the electrophoretogram for degraded RNA. So if you were extracting samples that were degraded prior to trizol extraction this is what it looks like (or some left sample with trizol on the bench and came to extract it next day-heard someone do that). If Trizol did cause degredation it wouldn't be added to the solution now would it? It's like saying you added proteases to your protein extractions when in fact you add protease inhibitors. The second link clearly shows that RNA integrity does influence PCR results, it's just less sensitive than microarray/NGS which typically require RIN >8. 

I can not judge from the amplification cuve plots what the efficiencies of the infividual amplifications are (i.e. the slopes of the straigt line segments in a half-log plot. It will be too inexact to estimate these slopes from the pictures.

Another typical cause of strange efficiencies can be the (incorrect) background correction. Your curves are corrected and it seems at least that this is not your problem. However, it might be worth to check the uncorrected curves if there is a problem with the background and if the "amplification-inedpendent change in flurescence" (if present) is corrected in a sensible way.

I think your problem is in part due to the low input amounts. The PCR does not seem to be very stable when the Ct is about 30 or higher. This in turn may be caused by a possible co-amplicication of unspecific products. Have you checked melting curves and agarose (or better: PA-) gels? If there are lots of unspecific products amplified, you will have to design new primers.

If this is not the problem, then I suggest that you generate a standard of a higher concentration (run a standard PCR, purify the PCR product, dilute it to get a Ct of about 14-16 for the starting dilution and then use this in a dilution series)*. If the efficiencies are ok then, you know that neither the primers nor the PCR mix is a problem. You can determin the LOQ and organise your sample preparation to have sufficient concentration to be above the LOQ.

---

* do this in a different lab (best in a different floor or building) to avoide cross-contaminations. Take only the working dilution back to the place where you prepare your qPCRs.

If you are concerned about inhibitors, the best test is a "poison control", which no one seems to know about any longer.  Take a good positive polymerase chain reaction and spike in some of your questionable template.  Does it still amplify or not?

Try several dilutions of the template in your reaction to dilute out any possible inhibitors if so. 

Have you run the old frozen sample of snail ganglion cDNA together with your freshly made samples? Additionally, have you run any analysis of your RNA quality before reverse transcription, such as looking at A260/280 ratio?

If you run your old samples and get good efficiency, or if you see something off in your fresh RNA samples (such as protein contamination) the culprit could be the quality of your RNA.

Thank you everybody,

I will try to answer more minutely to everyone personally in a few days. I should now say that I do check melting curves and I do perform gel electrophoresis of each sample after PCR (not all 96 wells, but one replica of every point and also every "suspicious" well with different melt curve), and there is usually no byproducts (there were more of them when I did 50 cycles, but at 40-45 cycles they were only in 1-2 wells of 96, usually the most diluted ones). I did check A260/A280 of my RNA, it was 2. I didn't dilute my RNA 100-fold before RT (there was 1 ug of RNA in each RT reaction), I diluted the cDNA sample before amplification. I agree that there's too little amount of cDNA in my samples, and so all my dilutions may just lie in non-linear part of the calibration curve, but I'm still confused because I've seen some other calibration curves that were still linear even after 30th cycle. I definitely will redo reverse transcription once again to try to receive more cDNA, and I will redo RNA extraction and will use DNAse on RNA-extracting columns to prevent DNAse from getting into solution. I already said some of these things above, in the question itself and in comments.

About Trizol: I did some research and agree that I was wrong: Trizol itself isn't RNA-degrading. But it's still quite common to obtain bad RINs using Trizol because in some cases its RNAse inhibitor capacity is not enough, and you need titration to calculate the right amount of Trizol you need for particular tissue type (and usually nobody does such titration): https://www.researchgate.net/post/What_is_the_rationale_about_RNA_integrity_number_RIN_in_transcriptome_sequencing

Good luck Ekaterina!

Dear Joshen,

I attached to my previous answer to you the table with Ct data as you requested, but you said that you can't judge the efficiencies. Did you mean that there's not enough replicas, or that most of the Cts were >30 and it's unreliable, or something else? Do you need any more data from me, or the whole experiment was just non-quantitative because of high dilutions?

For background correction the second fluorescent dye (ROX) was used, it was added in the pre-mixed master mix, and ROX fluorescence wasn't supposed to change during amplification. Fluorescence signal shown on these graphs was calculated as relative fluorescence, SYBR/ROX ratio. I used authomatic background correction, without this correction I can see that some (2-4 of 96) wells had more fluorescence to begin with, but it seems that high background is random and doesn't depend on dilution or primers.

Dear Hilary,

I've never heard about the "poison control" before, thank you very much, it's a good idea and I will remember it!

Unfortunately I've got too little of this dubious template, it's not enough even to prepare not-too-diluted standards, leave alone "spiking". I will definitely prepare some new templates, but I'm still curious what's wrong with the samples I already tested. As I said, the standards here were prepared from one experimental sample diluted up to 1:100. If there were indeed inhibitors, wouldn't they be diluted and stopped working, so the standard curves wouldn't be linear?

Aaron, thank you!

My old frozen samples are not just first-strand cDNA, they are actually amplified cDNA, so I guess there's plenty of material and I should definitely use them as a positive control, thank you for your suggestion!

In my fresh RNA samples A260/A280 ratios were very close to 2, I believe it means no protein contamination. But there was something with very high peak on 220 nm (or even lower, NanoDrop scale just begins with 220, so I can't be sure it was the highest point and not the shoulder of the peak), and I have no idea what substance it could be. Do you know anything about such peaks?

http://www.nanodrop.com/Library/T009-NanoDrop%201000-&-NanoDrop%208000-Nucleic-Acid-Purity-Ratios.pdf

230 is usually an indicator for chemicals, i.e. phenol in extraction chemicals.

Dear Can,

I suppose phenol absorbs at 230 AND 270 nm (Fig.3 in the manual you linked), and at 270 all was clear (RNA peaks at 258-260 were very sharp, no "shoulder" at the right). I always especially worry about phenol peaks when I work with Trizol and check 230 and 270 nm first thing.

Dear Aaron (if you still here): I've performed another try with different matrix (my old frozen amplified cDNA) and different PCR kit with Sso polymerase. Only the primers were the same. Ct were from 18 to 35. And again, all four primer pairs demonstrated 50% efficiency with R^2>0.98 (there was "compression" though, intervals between curves that are supposed to be equal were increased with each dilution, and if I consider only the first 2-3 dilutions efficiencies were 65-75%; troubleshooting guide in Sso protocol says that "compression" means that there are indeed inhibitors).

I settled on redesign all the primers and get new RNA samples with column method.

The parallelity is striking. Are you sure you did the dilution steps right? Using a wrong factor could cause the problem. Do you have any positive controls where you known that the efficiency would be almost ideal?

Agree with Jochen, check your calculations and can you remind me the concentration of your initial template and how you made the dilutions. 

For the first graph (posted in the question itself) the dilution factor was 2.5. dCt between two adjacent dilutions was supposed to be log(2)2.5=1.32 cycles assuming 100% efficiency. For the first dilution, I took 14 ul of the sample and added 21 ul of MQ water, so I got 35 ul (=14*2.5). From these 35, 16 were used as templates (4 primer pairs*4 replicas=16 wells, 1 ul per well) and 14 were used for next dilution. All next dilutions were prepared using the same calculations (14 ul of previous dilution+21 ul MQ=35, 16 in wells, 14 for next dilution and so on...) 14 ul of undiluted sample corresponded to ~0.5 ug of total RNA before reverse transcription, I didn't measure the cDNA concentration because cDNA wasn't purified.

For the second graph that I posted today the dilution factor was 3. dCt between two adjacent dilutions was supposed to be log(2)3=1.59 cycles assuming 100% efficiency. I took 9 ul of the sample and added 18 ul of MQ water and got 27 ul (=9*3). From these 27, 16 were used as templates, 9 were used for next iteration like I described above. I don't know cDNA concentration of this sample either, I only can say that I used ~0.15 ug total RNA, then performed reverse transcription and amplified 1/10 volume of this cDNA in 21 cycles of PCR (amplified all types of DNA molecules in the sample using adaptor primers). This sample was stored at -20C for two years, so I guess it degraded a bit; I also diluted it 3-fold BEFORE preparing the first calibration dilution, so the first dilution was actually 1/9, not 1/3.

I didn't use "normal" dilution factors 10, 5 or even 4 in this experiment because 3th-5th iterations would be too diluted and wouldn't amplify. I don't have "ideal" positive controls for this experiment, but I guess that next time I should definitely use more than 0.5 ug of RNA.

If I would ever try anything else with these primers it would be INCREASING of Ta. I tested the same "reference" sample in three experiments with Ta 59, 61 and 63C, and its Ct were the lower the higher was Ta, despite my calculations that predicted optimal Ta about 60 or lower.

I must add that my frozen amplified (unpurified) cDNA turned out to be bad positive control: on agarose gel there were long byproducts (400-700 bp while my target products are 100-150). I suppose these adaptor primers present in the sample still worked despite their dilution. I didn't see these byproducts when I used this template just to check that new primers work, I used 20 cycles then; now I used 40 cycles and they amplified. So the real efficiency is even lower than 50% here because some fraction of SYBR fluorescence was due to byproducts.

Ekaterina, I hope you have more success with the new RNA samples and the new primers. In the mean time, I'll ask one of my colleagues here to see if he can look at this thread and offer some insight. But let us know your results when you run your new samples.

There are some considerations that need to be taken into account when doing pre-amplification PCR. Perhaps this article may hep you.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4673511/

Thank you everybody.

I'm going to begin with just one primer pair, GAPDH, to not overcomplicate the task. I designed another GAPDH primer pair; next time I choose primer positions in the most conservative parts of the sequence to mitigate possible sequencing errors in my target contig. I also will perform Bioanalyzer electrophoresis of my RNA when I accumulate 11 RNA samples for a full chip (we scrimp these chips, so it'll have to wait about a week).

I've checked my RNA samples; they're not degraded. In fact, they look much better than I anticipated.

Please note that in snail samples total RNA shouldn't have 2 rRNA bands, you should see only 18S (~2000 nt) band, but very high. Snails have 28S rRNA, of course, but what makes it peculiar is that it consists of two separate pieces held together by ribosome proteins. When you purify total RNA, you remove these proteins and 28S rRNA falls apart. Each half of 28S rRNA is about 2000 nt, so its peak merges with 18S peak precisely.