Although many different effects of AuNPs in PCR have been published, it has not been applied to the field of Toxicology or nano-toxicology studies, i.e. where data can be misinterpreted as being non/toxic simply due to assay interference . However, this current report is the first, step-for-step, detailed description indicating how the intracellular residual presence of AuNPs in a sample can alter the RT-qPCR results and, thus, the interpretation of the toxicity of AuNPs based on the assay. This work, therefore, promotes awareness regarding the effects of residual intracellular ENMs (i.e. those remaining in biological samples after cell lysis followed by traditional isolation/purification procedures). The work explains why scientists who conduct molecular biology work related to nanomaterials must be aware that the traditional reasoning for assessing the toxicity of larger chemicals does not apply since those molecules may not be metabolised, whereas metal or metal oxide ENMs may be bio-durable and remain within the cell for long time periods.
The in silico analyses assessed the design of the primers. This analysis predicted that the primers for 18S, GUSB, HSP90, SDH and YWHAZ genes would not form secondary structures, i.e. these primers would amplify and produce acceptable PCR amplicons (see Additional file 1). Thereafter, a preliminary test was performed using SYBR Green in order to screen these primers. This part of the study was completed by analysing the RNA obtained from BEAS-2B cells that had been treated with AuNPs (see Additional file 2).
Furthermore, the manual assessment determined whether or not AuNPs could interfere with the assay. Assessment of the amplification of the reference genes, as influenced by AuNPs, included use of the CFX Manager software to obtain the PCR efficiency, the linearity of the PCR assay as well as the gradient or slope obtained from the standard curve. The melt curves of the final PCR amplicon were also analysed. The preliminary study did confirm that the primers could amplify the required amplicons in accordance with MIQE guidelines . However, possible assay interference caused by AuNPs may have occurred and should not be ignored.
Admittedly, ENMs may have different coating and capping agents, in addition to forming a protein corona when incubated in different culture media, which may alter their intracellular uptake and toxicity . It is only those that internalize and are bio-durable that should be considered when assessing their interference upon cell lysis during assay procedures. The proposed interactions that may have caused this interference are discussed in detail below. Hence, it is generally understood that surface properties of ENMs will affect intracellular uptake and subsequent gene expression within intact cells. In contrast, what is not generally understood is the consequence of internalised ENMs that are already present within the cells before and during lysis. This aspect has not been widely investigated. Therefore, it was decided to test the primers and validate the RT-qPCR assay. Appropriate control samples were included to represent samples with “non-functionalised ENMs” and “no ENMs” so as to focus on the interference specific to the RT-qPCR assay, instead of the biological interaction at the cell’s intact surface membrane.
Assessment of reverse transcription
The universal RNA standard, obtained from ten cell lines, was spiked with AuNPs, before reverse transcription, in order to generate cDNA. This method assesses the effect that AuNPs could have on RT-qPCR if AuNPs were internalised in cells, i.e. it mimics the situation where AuNPs may interact with cytosolic cellular content of mRNA, which is usually exposed during translation. Thus, the treatment has biological significance because it mimics any possible interference of residual ENMs present in biological samples, which would co-precipitate with isolated RNA intended for analyses during toxicity exposure assessment studies.
Initial observations of only the amplification plots obtained from the CFX Manager software indicated a change in the profiles for some of the reference genes, i.e. before and after deliberate addition of AuNPs (see Fig. 2; Additional file 3). The fluorescence reading was quenched and, thus, the Cq increased. Finally, the qPCR software and statistical programs assessed the accuracy and stability of the gene expression. Both traditional analyses, as well as, manual assessments were performed on the results. In addition, the results were analysed based on the point at which the AuNPs were added to the reaction, i.e. either spiked at the reverse transcription step (part 1), or, spiked at the amplification step only (part 2).
The analyses relied on traditional software packages that are specific for qPCR assays in order to perform relevant statistical comparisons (see Additional file 4). To summarise, when comparing results obtained from the universal RNA that had been spiked with AuNPs at the reverse transcription step (part 1), all the statistical analysis programs found that the same reference genes exhibited the highest stability. These included HSP90, SDH and YWHAZ. In addition, stable combinations of reference genes were also identified, i.e. GUSB with HSP90. When working with ENMs, it is recommended to not solely rely on traditional qPCR software analysis programs, but to also do manual assessments, e.g. determine PCR efficiency variations between treatments, as well as changes in the dissociation assay (melts) of the different products formed (see Additional file 5). Due to the changes in the Cq, GAPDH, GUSB, HPRT1 or SDH could possibly be suitable reference genes. However, the melt peak analysis of all the products formed did not identify any significant differences (data not shown).
Assessment of DNA amplification
The second part of this assessment analysed the effect of AuNPs on only the DNA amplification step (part 2; see Additional file 6). This method assesses the effect that AuNPs could have on RT-qPCR when AuNPs are internalised in cells, i.e. it mimics the situation where AuNPs may interact with single-strand DNA, which is usually exposed during cell replication. These experiments were therefore required for continuity in order to compare the interference of AuNPs throughout each point in the experiment, i.e. continuation of the detailed step-for-step assessment.
The cDNA, obtained from the universal RNA standard, was spiked with AuNPs after reverse transcription was completed, in order to amplify the cDNA under PCR conditions with deliberate AuNP interference. This treatment would mimic those samples where DNA was isolated from AuNP-treated samples and amplified by qPCR, e.g. assessment of methylation studies, SNP and mutation analyses or genomic DNA etc. This also implies that testing only the amplification step (but nothing prior), is not sufficient to verify qPCR assays for genotoxic studies related to ENM exposure assessments, where the starting material might be contaminated with residual ENMs. In general, when comparing results obtained from the universal RNA that had been spiked with AuNPs at the amplification step (part 2), the combined results showed that only three reference genes exhibited the highest stability, i.e. GUSB, HSP90 and YWHAZ.
Further analyses deemed GAPDH and GUSB to be inappropriate as reference genes. Separation of the PCR amplicons via electrophoresis identified the formation of multiple products for some of the replicates of GUSB, which was subsequently disqualified as a suitable reference gene. In addition, although GAPDH is a popular reference gene in many studies, it does have limitations [35,36,37]. Firstly, GAPDH plays a role in glycolysis and as such, may result in variable expression in different tissues or disease states. Secondly, some GAPDH pseudogenes are expressed, where primers will detect the presence of both the pseudogenes and the cDNA of the active transcript. Lastly, considering that the human genome may contain up to 60 pseudogenes for GAPDH, DNase treatment may not always degrade the entire genomic DNA in which these sequences reside. It was, thus, proposed that only YWHAZ and/or HSP90 be used as reference genes. In addition, future studies should focus on the clustering feature available in the Precision Melt Analysis™ software, i.e. HRM. In fact, a few HRM shifts were identified by an in-depth analysis of the melt profiles, where readings were taken every 0.2 °C (data not shown). Thus, these genes have been identified as targets for developing a “diagnostic tool” and are currently being investigated further.
Proposed interference based on comparisons to previous studies
There is a continuing concern with assay interference caused by ENMs that has not been addressed, which has a serious implication for nano-toxicity testing in assays using PCR-based techniques. When ENMs enter cells, the residual amount of intracellular ENM remaining in a sample could alter the PCR assay and, thus, generate false readings for gene expression based exposure-studies. The results reported herein indicate that BEAS-2B cells could internalise the AuNPs, where the AuNPs would have access to cellular content during cell lysis. In addition, the deliberate addition of AuNPs into a sample, at either the reverse transcription or PCR amplification stage of the assay would interact with these macromolecules and/or assay reagents to cause assay interference.
The section below serves to highlight the main points to consider for cytotoxic and genotoxic studies, by highlighting common observations within all these different reports.
Interference with the dye used for detection
When a fluorophore is in close proximity to a metal nanoparticle displaying plasmon resonance, its fluorescence emission may change . These authors identified factors that affect the fluorescence of a fluorophore when it is near AuNPs, i.e. the particle size, coatings, as well as, the wavelengths of the incident light and emitted light. Hence, they concluded that fluorescence may be enhanced or quenched by changing the distance between the fluorophore and the AuNP. In fact, Rosa and colleagues developed a tool for dealing with this modulation between fluorescence enhancement or quenching when close to AuNPs . The authors concluded that this was a more accurate method for determining fluorescence emission near AuNPs. In other words, one can correct for the spectral response when fluorophores are conjugated to AuNPs, which is relevant to nano-diagnostics that rely on quantification assays.
The fluorescence detected in qPCR is generated by a DNA-binding dye. This implies that the conformational structure of DNA requires and ensures a specific spatial interaction with the dye. A study was found to be similar in design to that reported herein, i.e. a presentation by Prado and colleagues at the qPCR-NGS 2013 Symposium held in Germany . The authors noted that the amplification plot of SYBR Green was affected by the addition of increasing concentrations of Fe3O4 NPs. In contrast, in our study reported herein, AuNPs that were 14 nm in size and citrate stabilised, were used as the treatment. It should also be emphasised that the dye used by Prado and colleagues was SYBR Green. This is a “relocating” dye, where even though the SYBR dye melts off at a dissociated part of the DNA strands, it has the potential to re-attach at another point in the same DNA strand that has not melted. Therefore, SYBR Green has the potential to generate false readings, where a study by Yang and colleagues reported partial quenching of the fluorescence of SYBR Green in qPCR caused by AuNPs . The authors explained that since SYBR Green only becomes fluorescent after binding to the minor groove of dsDNA, it implies that AuNPs must first bind to dsDNA before it can quench the fluorescence in a qPCR assay. This is, therefore, a non-specific interaction between the ENM and the DNA, which could explain conflicting genotoxicity reports published to date. Irrespective of the mechanism, by using amplification profiles and melt analyses, changes in a PCR have been attributed to AuNPs that cause fluorescence quenching and DNA duplex destabilisation . These authors concluded that a thorough evaluation and validation of the impact of AuNPs on any qPCR assay should be undertaken and further serves to emphasise the importance of our study presented herein. The possibility that SYBR Green could further be compromised by interacting with ENMs was again proposed recently . Shaat and colleagues investigated AuNPs as carriers for delivering siRNA when they detected possible assay interference. However, they concluded that their results did not change when they repeated the experiment using an end-point RT-PCR assay in the absence of the SYBR Green dye, i.e. the interference was not specifically due to the interaction with the dye. Hence, they referred to our own previous study where we have reported assay interference during an RNA isolation procedure . Again, the importance of the effect of residual intracellular ENMs in the elucidation of genotoxicity studies has become a recurring theme in the latest publications.
EVA Green was used as the dye in our study reported herein. This is a “saturated” dye, which binds at each position in the DNA strands and reduces false readings, i.e. it is a better option compared to SYBR Green. Even so, as mentioned above, we observed that the addition of AuNPs reacted with the PCR enzyme cocktail and changed the colour to a clear solution. It should be noted that the super-mix contains the Sso7d-fusion polymerase, which employs an antibody-mediated hot-start feature in order to sequester the enzymatic activity prior to the initial PCR denaturation step. Once the antibodies denature irreversibly during the heat activation step, the DNA polymerase is released and is fully active. Therefore, it is proposed that in our study the AuNPs have the potential to interfere with the PCR assay by interacting with the buffer components, the hot-start antibody or the polymerase enzyme, but not necessarily the dye.
Influence of thermal conductivity of ENMs
Gold colloids can form a highly ordered liquid layer that can lead to higher thermal conductivity . Therefore, when DNA is in close proximity to gold colloids it can induce a fast heat transfer and, thus, enhance PCR locally around that DNA . Li and colleagues proved that most of the primer and DNA did not bind the gold colloid, even though some reports propose that ssDNA replaces the citrate ions to bind to the gold colloids. Li and colleagues went further and analysed PCR reactions using both traditional DNA polymerase (Taq), as well as, SYBR Green enzyme cocktails, with citrate-capped AuNPs 12.7 ± 0.8 nm in order to investigate the influence on PCR efficiency. They determined the effective concentration of gold colloid to be 0.7 nM. Our study reported herein used a FC 0.72 nM for the 25% AuNP-spiked samples. Therefore, the lowest concentration used in our study was sufficient to investigate possible assay interference. Li and colleagues proved that the thermal conductivity of the AuNPs played a significant role in shortening the time required for heat dispersion or equilibrium and, thus, increased the efficiency of the traditional end-point PCR reaction tested. When Li and colleagues tested a qPCR reaction, similar increased PCR efficiencies were obtained, but the Cq values shifted by 11 cycles compared to the positive control. The authors interpreted this finding as increased PCR efficiency. However, it may actually be a sign of ENM-induced assay interference in a PCR.
Interactions with components of the PCR enzyme cocktail mix
A recent report assessed the effects of specifically AuNPs on PCR . Briefly, the authors reported that an excess of ENMs would inhibit PCR by either adsorbing to the polymerase, to Mg2+, to oligo-nucleotide primers or to the cDNA/DNA templates. The study by Bai and colleagues differs to that reported herein since lamda (λ) DNA and amino-modified silica-coated magnetic NPs were used. Instead in our study we used AuNPs and RNA from ten cell lines so that the study would have biological significance, i.e. the results obtained mimic those of genotoxic analyses after AuNP exposure. Furthermore, the authors disproved the theory that ENMs inhibit non-specific amplification by false priming. Rather, they showed that it was a concentration-dependent phenomenon, i.e. low concentrations of NPs inhibit amplification of long amplicons, and, increased amounts of NPs inhibit amplification of short amplicons. This means that for our study, where non-cytotoxic amounts or < 50% vol/vol of AuNPs were used, the low concentration of NPs might only interfere with the longer amplicons, e.g. 18S, ACTB, GUSB, PPI and TBP (see Table 1). This is currently being investigated further with the development of the proposed diagnostic tool.
Furthermore, Li and colleagues reported that enhancement of the PCR efficiency depended on the DNA polymerase used . These enzymes may have different proof-reading abilities, which would influence the binding and amplification of the DNA strand, where altered or damaged templates would have different amplification efficiencies. Hence, the findings by Li and colleagues support our hypothesis that AuNPs have the potential to interfere with the hot-start antibody or the polymerase enzyme, as mentioned above. The complex interaction between AuNPs and the DNA polymerase (Taq) was also reported elsewhere .
Binding of nucleic acids relevant to DNA amplification
The composition of the DNA sequence may also play a role in the degree of assay interference observed. For example, one study found that the dA mononucleotide had a stronger affinity for 5 nm AuNPs, most probably due to the freely accessible amine group . Therefore, mRNA that contains a characteristic poly-A tail at the 3′-end of the molecule might be more susceptible to interference caused by gold-based ENMs. This means that for some genotoxicity studies, which may use mRNA as a starting material, the results could be influenced by this affinity for small AuNPs. However, the poly-nucleotides tested by Yang and colleagues were more rigid due to the phosphate backbone of DNA, which resulted in altered affinities to the AuNPs (see further discussion below regarding the effect of charge). In addition, they pointed out that the dsDNA dissociation caused by AuNPs was strongly dependent on the particle size (see further discussion below regarding the effect of size). These findings imply that primers that are designed to amplify A-rich regions in the DNA sequence may be problematic. Also, amplification of the 3′-end of gene may be challenging, especially when using mRNA as the starting material.
Another study went further and explored the effect of charge on DNA binding . Initially, they found that the small uncharged AuNPs were able to bind the minor groove in DNA, but did not damage the structural integrity of the helix or disrupt Watson–Crick pairing. Thereafter, they used charged AuNPs that were 1.4 nm in diameter and found that at high concentrations it could bind DNA via electrostatic interactions from the cationic ligands, which then lead to bending and strand separation. In contrast, our study used larger citrate-stabilised AuNPs (14 nm in size), with a negative net charge. Hence, interaction between our large negatively charged AuNPs and the negatively charged phosphate backbone of nucleic acids found in template RNA/cDNA is not very probable. In fact, a thiol group is usually required in modified dsDNA terminals in order to enable covalent bonding with the metal surface . Rather, the proposed interaction between the AuNPs and the polymerase enzyme or the Mg2+ in the buffer, as well as, the effect caused by a higher thermal conductivity around the DNA are more plausible reasons for any variations observed in the results, i.e. assay interference.
The effect of ENM size has also been investigated . Although 5, 10 and 20 nm sized particles were tested, the authors reported that AuNPs could inhibit a PCR reaction. However, that this was due to the concentration of the AuNPs used. Specifically, the larger sizes caused PCR inhibition at lower concentrations, in comparison to smaller sizes. The “threshold” concentrations were determined to be 5.5, 1.1 and 0.24 nM for the 5, 10 and 20 nm sized particles, respectively. They proposed that their observations could be due to AuNPs that bind to the polymerase. They also went further and explained that all their results showed that the total surface area of the AuNPs altered the PCR product yield, i.e. the size influences the surface area, which then prompts interactions with the PCR components. Specifically, a higher concentration of the smaller sized AuNPs was required to cause inhibition of the PCR. In addition, when the concentration was adjusted to make sure that total surface area in the PCR reaction was the same for the 5, 10 and 20 nm sized particles, it resulted in equal inhibition of the PCR. This could explain why different studies show different degrees of PCR inhibition, i.e. the different studies do not take into account the entire surface area of the ENM that is able to interact with the PCR components. The only problem with their findings is that these statements were based on observations made on amplification occurring between 35 and 50 cycles. It is well documented that primer-dimers will form after 35 cycles of amplification. Hence, most qPCR assays do not exceed 40 cycles to avoid quantification of a non-target amplicon. However, the authors declared that a 119 bp invA amplicon was observed after electrophoresis .
Selective adsorption of single-stranded oligo-nucleotides onto the surface of AuNPs, via electrostatic interactions, has been reported . However, dsDNA did not adsorb and this difference between the two states of DNA was used to detect changes in oligo-nucleotide lengths in the presence of a dye, i.e. parts of hybridised chains would have electrostatic properties of ssDNA and other parts would have properties of dsDNA (due to contact between the dye and gold). This observation was further developed where the authors reported on probes, consisting of small oligo-nucleotides, which are able to bind to RNA. The fluorescence of unbound probes was quenched by the presence of AuNPs. This difference in (quenched) fluorescence was used to indicate that probes (which were designed for a particular target RNA), did actually bind in the sample, i.e. identification and confirmation of the target sequence . The structure of the RNA was determined based on which of these probes hybridised, i.e. mismatched sequences would result in weaker binding. Furthermore, in a different study performed by Yang et al. of a qPCR assay with SYBR Green, they reported that smaller AuNPs had a higher binding affinity for single-stranded (ss) DNA, i.e. the nitrogen atoms in the amines of DNA underwent covalent interaction with the gold atoms and disrupted the hydrogen bonds formed between complimentary oligonucleotides . Yang and colleagues went further to propose a non-specific interaction between small AuNPs and the phosphate backbone, which may disrupt geometric hybridisation and, therefore, cause partial dissociation of dsDNA. These findings correlate to the Railsback study mentioned above, as well as, our own previous findings from the study of the interaction of AuNPs with single-stranded RNA . It should be noted here that different kinds of interactions are being referred to, i.e. the non-specific binding refers to uncharged ENMs that are able to fit in the minor groove of a dsDNA complex, whereas the charged ENMs are able to cause electrostatic interactions that can bend and separate the DNA stand and, lastly, the covalent interactions rely on sharing of electrons, e.g. thiol group with a metal group. In addition, the reports discussed above imply that when AuNPs dissociate dsDNA, it could lead to lower melting point of the final PCR products because the structure of the DNA-duplex is distorted, similar to the mechanism discussed above, where increased heat transfer caused by a gold colloid could lead to easier DNA duplex strand separation for PCR amplification, i.e. lower Cq values . In other words, the qPCR melt peak would shift to a different/lower temperature and the HRM profile would also change for qPCR-based results. Hence, an analysis of HRM profiles is being developed further by us as diagnostic tool.
Limitations of assay
Another factor to consider is the limit of detection of the qPCR assay itself. According to the MIQE guidelines , a lower Cq is correlated with higher target expression in a sample. In other words, the Cq values are inversely proportional to the amount of target nucleic acid present in the sample, where the lower the Cq, the greater the amount of target present. In addition, the Cq value must be compared to the non-template control (NTC) as well as to either, (a) another Cq value in another sample to calculate the relative expression, or, (b) a standard curve where known amounts of target are analysed leading to absolute quantification. Without this normalisation step, the Cq values should not be used to draw any conclusions. For example, a recent study noted that high concentrations of AuNPs increased Cq values and concluded that it was an inhibitory effect . Although the deliberate addition of AuNPs did create a false estimation of the initial DNA, the authors interpreted this change as increased sensitivity. It is a misconception to link lower Cq values with increased sensitivity . A true indicator of sensitivity is actually the limit of detection. The results reported by Gurjar and colleagues should rather be viewed as an example of how 33.2 and 51.51 nm sized AuNPs interfered with the SYBR Green based qPCR assay. Upon closer inspection, it appears that the assay interference caused by the AuNPs altered the melt profiles of the PCR amplicons, where the melt peak shape and peak amplitude was changed, i.e. their analyses of the final PCR products generated indicated that the products were not the same.
In another recent study, upconversion nanoparticles (UCNPs) were reported to improve the specificity of the PCR . This was an end-point assay, not a “real-time” qPCR assay and the results were interpreted as increased specificity because the number of additional amplicons decreased as the intensity of the main PCR amplicon also decreased, i.e. total PCR inhibition. The gene target in their PCR was a 120 bp 5S rRNA repeat. These 5S RNA repeats are present as multiple copies of target sequence in the genome. The 5S rRNA genes are organised as tandem repeated clusters and the gene copies range from 100 to 300,000. Therefore, depending on the stringency of the primer design, a primer pair may amplify many similar copies, which will appear as many different bands on an agarose gel. Hwang and colleagues observed multiple products in the control sample that did not contain UCNPs, which was most probably due to the primer design. Subsequently, when the PCR was inhibited (including the main amplicon), all these non-specific products were also inhibited. In contrast, to unambiguously show an increase in PCR specificity, one would first need an optimised PCR reaction without non-specific products and then prove that the addition of a specific ENM altered the rate of amplification or the proof-reading ability and resulted in the production of a superior PCR product. Hwang and colleagues did emphasise a decrease in PCR amplification as caused by the UCNPs and, although it may not be an example of increased PCR specificity, it most definitely is another example of assay interference. Again, this recurring theme in the latest publications of how residual intracellular ENMs are able to alter assays needs clarification for the correct interpretation of genotoxicity studies.
Although the reports of previous studies mentioned above are ambiguous, continued analyses of studies with the deliberate addition of AuNPs can be used to predict assay interference, i.e. deliberately change the mechanism of a PCR reaction to mimic conditions where intracellular ENMs may unintentionally change the dynamics of the PCR components. This applies to all situations where ENMs, irrespective of the functionalised surface modification, are internalised into cells. It is important because in such cases where intracellular ENMs remain in PCR-related samples, the results would incorrectly be interpreted as an indicator of gene expression and be used to determine ENM toxicity in the sample. Where one suspects residual intracellular ENM contamination of starting material (especially gold), the following should be avoided:
Design of primers to amplify near A-rich regions in the DNA sequence.
Amplification of the 3′-end (near the poly-A tail) of gene when using mRNA as the starting material.
Other recommendations include the use combinations of genes (e.g. HSP90 or YWHAZ), for improved gene expression normalisation. It is further recommended that RNA standards, which have been spiked with known amounts of the ENM (e.g. AuNPs), should be run in conjunction with the unknown RNA samples. Thus, one could determine the degree of error with the associated compensation required for assays influenced by ENMs. This implies that multiple forms of analyses are required, in order to determine the degree of error in the assay. As described herein, this can be achieved by using both traditional software programs and via manual assessments of the individual parameters.