Ideally, negative control reactions are clean and negative. However, it depends on the application and the severity of a potential signal in a negative control reaction if it is problematic or not. For diagnostic purposes where sensitivity is important, I would not accept a signal in a negative control. However, for research purposes I can accept a little positivity in a negative control reaction. Our default guidelines are that a positive negative control should be 5 cycles later than a real sample with the lowest concentration (highest Cq values). Please have a look in Biogazelle's Knowledge center (under publications) for articles and book chapter elaborating on this issue.

With respect to the primer sequences being published, did you use them with the same master mix, thermal protocol and primer concentration? These 3 elements all influence primer behavior. 200 nM final concentration of each primer is fine for SYBR Green I (we routinely use 250 nM). Lowering is not recommended.

http://www.biogazelle.com/knowledge-center/publications

For 2 pairs of primers they used same master mix and concentration, however in 25 ul instead of my 20 uL. Their criteria was no amplification in negative control >=35 cycles. For another 2 sets of primers - no data available.

200 nM is recommended by SYBR FAST manual, so I stick to this concentration.

As for Ct values, my working dilution has Ct of 20 for my gene and 37 for negative control. But you mentioned lowest dilution "Our default guidelines are that a positive negative control should be 5 cycles later than a real sample with the lowest concentration (highest Cq values)" .

How low is very low? When I did standard curve, my lowest dilution is 0,001 ng and of course its Ct value close to Negative control (30 to 37). If I dilute further, it would be even closer to my neg. control. Which 5 cycles apart are you talking?

PS. Thank you for the link. Did you mean any specific book? I see a list of publication regarding various PCR issues.

Another feature to be weighed is the idea that primer dimers often do not form in the presence of an ample amount of template. 'When the cat is away the mice will play' type of thing. E.g., in the total absence of template, primers will interact in more and more thermodynamically rare ways to form dimers (if dimers are possible due to sequence). Melt curve analyses show this in the standard curve at the more dilute points (if primer dimers are thermodynamically able to form). A very very small bump (in the primer dimer region) preceding actual amplicon can sometimes be ignored/considered inconsequential, but the more pronounced it is, the less valid that reaction's measurement becomes. If the last point or so in the standard curve is all primer dimer, it is not an actual point on the standard curve and thus should not be used. Given the idea that 1 copy of something (in highly-efficient reactions - which it appears you have) can appear anywhere from ~35 to 37 Cq (and on up to 48 cycles e.g., in ~70%-efficient reactions), you are already dealing with the possibility that much of the signal from 35 on up is largely dimer (unless the melt curve speaks otherwise!). Your melt curves are your compass to determine which reactions are valid or not.

Primer dimer signal (only) is of course the same thing as 'no signal' for whatever specific target you are examining. For gDNA contamination carryover signal, the 5 cycle spread idea [also] applies. I.e., you don't want to count gDNA contamination signal (as seen in reactions using the original RNA [no-RT-controls] when RNA is gDNA-contaminated to a certain extent from the outset) as real transcriptomic signal. But, if that gDNA contamination signal is 5 Cq or more later than the real signal in your final qPCR target reactions, it is considered to be a negligible contribution to signal. Good DNase treatment of RNA beforehand or using primers designed to exon-exon junctions [genomic intron-exon junctions] help to minimize/eliminate this possibility). The dogma oft' runs thick, wherein different mastermixes can also let primers perform slightly different dogmatic/thermodynamic tricks. The increased randomness and increased need to consider Poisson distribution effects in the Cq nether region (e.g. cycles 34 to 50) commands one to carefully weigh whether or not something actually, truly, walked or talked like the duck one was specifically hunting in the first place. The gift of SYBR Green-based qPCR is the ability to view melt curves in order to weed out what was the real duck, and what was not. There may be two faux ducks of concern, however: primer dimer formation, and gDNA contamination.

Thank you Jack,

I don't worry much about gDNA contamination, as not only I rigorously treat my RNA but I also don't have amplification in other set of primers.

As for melting curve, indeed for highest dilution - it is still my product (Tm consistent with higher concentrations), but for negative control Tm differs and contributes a small pick compare to lowest concentration high pick.

I will design new pair of primers and see if problem persists.

Your problem negative control (wherein the 'sample' is just water) is behaving somewhat normally then ... being that primer dimers can often form in those reactions - but won't form when template material is present. If you can't design a better primer pair, the ones you're already working with still may be OK. I understand wanting it perfectly clean though; it's easier to convince others that your results are unequivocal if your negative controls produce no event. Your melt curves are already telling you good news about your work. Best of luck.