Evaluating the stability of host-reference gene expression and simultaneously quantifying parasite burden and host immune responses in murine malaria

The efficacy of pre-erythrocytic stage malaria antigens or vaccine platforms is routinely assessed in murine models challenged with Plasmodium sporozoites. Relative liver-stage parasite burden is quantified using reverse transcription quantitative PCR (RTqPCR), which relies on constitutively expressed endogenous control reference genes. However, the stability of host-reference gene expression for RTqPCR analysis following Plasmodium challenge and immunization has not been systematically evaluated. Herein, we evaluated the stability of expression of twelve common RTqPCR reference genes in a murine model of Plasmodium yoelii sporozoite challenge and DNA-adenovirus IV 'Prime-Target' immunization. Significant changes in expression for six of twelve reference genes were shown by one-way ANOVA, when comparing gene expression levels among challenge, immunized, and naïve mice groups. These changes were attributed to parasite challenge or immunization when comparing group means using post-hoc Bonferroni corrected multiple comparison testing. Succinate dehydrogenase (SDHA) and TATA-binding protein (TBP) were identified as stable host-reference genes suitable for relative RTqPCR data normalisation, using the RefFinder package. We defined a robust threshold of 'partial-protection’ with these genes and developed a strategy to simultaneously quantify matched host parasite burden and cytokine responses following immunisation or challenge. This is the first report systematically identifying reliable host reference genes for RTqPCR analysis following Plasmodium sporozoite challenge. A robust RTqPCR protocol incorporating reliable reference genes which enables simultaneous analysis of host whole-liver cytokine responses and parasite burden will significantly standardise and enhance results between international malaria vaccine efficacy studies.


Mouse model and sample generation
Immunogens Full-length Plasmodium yoelii circumsporozoite protein (CSP) was synthesised commercially (Genscript, USA) and cloned into a pVR1020 plasmid DNA vector (Vical Inc, USA) downstream from a human cytomegalovirus immediate-early promoter and in-frame with the tissue plasminogen activator signal peptide.Plasmids were purified using an EndoFree plasmid gigaprep kit (Qiagen).A human adenovirus serotype 5 (AdHu5) vector was constructed with a PyCSP antigen using pAd/PL-DEST™ Gateway vector system and Gateway LR clonase enzyme (Invitrogen) following the manufacturer's protocol.Linearised plasmids were transfected into Microbix HEK293 cells (Microbix Biosystems Inc., Canada) using a FuGENE HD transfection reagent (Promega, Australia).The virus was then cultured and purified by ultracentrifugation over a caesium chloride gradient, as previously described 42 .

Immunisations and parasite challenge
Female BALB/c H-2Dd mice aged 5-7 weeks obtained from the Animal Resource Centre (ARC, Australia) were immunised by intramuscular injection (IM) into the anterior tibialis muscle (50 μl/leg) with 100 μg plasmid DNA (Prime) followed 12 days later with an intravenous injection (IV) into the lateral tail vein (200 μl) of 1 × 10 8 infectious units (IFU) of respective AdHu5 virus (Target).At 5 weeks post-boost, as tissue resident memory T cells are present 41 , mice were challenged by IV injection of 1,000 cryopreserved Plasmodium yoelii 17XNL sporozoites (Sanaria Inc., USA) diluted in 200 µl PBS with 2% naïve mouse serum.Unchallenged and unimmunised (Naïve), sporozoite-challenged infection control (IC), and Prime-Target immunised and challenged (PyCSP) mice were studied.All experiments were approved by the Animal Ethics Committee of James Cook University (#A2549), and all procedures were conducted following the 2007 Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, which adheres to the ARRIVE guidelines.

Liver harvesting and RNA extraction
All livers were processed identically as previously described 2,6 ; however, MagMAX™ mirVana Total RNA Isolation Kit (Applied Biosystems) was used for RNA extraction to increase RNA yield 18 .Briefly, whole livers were harvested at 42 h post-challenge in 5 mL MagMAX™ lysis buffer (Applied Biosystems) containing 1% β-2mercaptoethanol (Sigma-Aldrich, Australia) and homogenised with a TissueRuptor II (Qiagen) homogeniser for 1 min.The lysate was stored at -80 °C.RNA was extracted from 50 µL of liver lysate diluted 1:1 in MagMAX™ lysis buffer, following the manufacturer's recommendations with DNase treatment and elution in 50 μL elution buffer.
www.nature.com/scientificreports/cDNA synthesis Extracted mRNA was quantified using a NanoPhotometer® N60 (Implen, München, Germany).RNA (0.4 μg) was then converted to cDNA using the SuperScript™ IV First-Strand Synthesis System (Invitrogen) in 10 μL total volume reactions with random hexamers only with the following modifications from the manufacturer's protocol: cDNA synthesis was conducted with the SuperScript™ reverse transcriptase at half the manufacturers recommended concentration (10U/µL RNA ), as previously described 43 .

Quantitative PCR (qPCR)
Assay setup qPCR was performed using ssoAdvanced SYBR® SuperMix (BioRad) following the manufacturer's recommendations (hot start 2 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 30 s at 60 °C) 18 .Reactions were run at 5 μL total volume amplifying 1 μL sample, as previously described 43 .Reactions were measured by QuantStudio 5 Real-Time PCR Machine running QuantStudio Design and Analysis Software (v1.4.3,Applied Biosystems), using technical triplicates and no template negative controls.Amplification efficiencies were calculated for all qPCR primers by calculating calibration curves from log diluted cDNA of pooled (n = 5) naïve whole mouse liver, or pooled (n = 5) infection control whole mouse liver when testing Py18s primers, as per MIQE guidelines 21 (Table 1).Cycle threshold values (Ct) were determined with the threshold set in exponential phase amplification at ΔRn0.3.All reactions were followed by a melt curve analysis ensuring primer specificity and contained desalt-grade PrimerBank™ 44 primers (Sigma-Aldrich) run at 500 nM.Reference gene expression stability and whole liver protectivity were evaluated by directly comparing Ct as amplified from 25 ng cDNA per reaction.The host-cytokine response to infection and immunisation was assessed by amplifying 50 ng cDNA per reaction, as per the optimised protocol.

Quantification of host-cytokine expression and parasite burden
Host-cytokine expression was calculated with the standard delta-delta cycle threshold (2 −ΔΔCt ) method relative to the geometric mean of the endogenous control reference genes SDHA and TBP, as previously described 24 , using naïve mice as the control group.Parasite burden was analysed using a modified 'Fold-reduction' approach (Supp Protocol.1), wherein several adaptations were made to the standard 2 -ΔΔCt protocol.Briefly, since the www.nature.com/scientificreports/parasite burden of test groups would be expected to be equal or less than IC mice, the fold-change (i.e., 2 -ΔΔCt ) calculation was inverted to fold-reduction (i.e., 2 ΔΔCt ) using IC mice set as the control group.Since a standard deviation (σ) of P. yoelii 18s expression within the IC group was equal to 1 Ct (i.e., 1σ = 0.957 Ct), the threshold of 'partial protection' was set 2σ (i.e., 2 Ct) from the mean.Since the inclusion of qPCR data with Ct > 35 may increase false positive pathogen detection 24,45 , the limit of detection (LOD) of the qPCR assay was set to Ct = 35.Therefore, 'partial protection' is between 2 σ from the mean of the infection control, to the LOD of the assay.'Sterile protection' is defined as a greater than the LOD of the assay.Both the 'partial protection' and LOD were normalised relative to the experimental geometric mean of endogenous control reference genes SDHA and TBP.

Flow-cytometric assessment of parasitemia
Parasitemia was assessed using the flow cytometric assessment of blood (FCAB) assay 16 from day five postchallenge until the infection had resolved.Briefly, blood from the tail vein was stained with anti-CD71-PE (BioLegend, USA), fixed with PBS containing 4% w/v paraformaldehyde and 0.0067% w/v saponin and then resuspended in buffer containing 0.5 µg/ml bisbenzimide Hoechst 33,342 (Sigma-Aldrich, USA).Flow cytometric analysis was performed on an LSR Fortessa (BD Biosciences, NSW, Australia) using a high-throughput sampler.

Statistical analysis
Reference gene expression stability of Ct values was analysed using an Ordinary One-way ANOVA with a Bonferroni-corrected multiple comparisons test against naïve mice.All data were tested for Gaussian distributions with a Shapiro-Wilk normality test.Reference gene expression stability was analysed with RefFinder software as previously described 22 .Briefly, three packages (BestKeeper, geNorm and NormFinder) employed individual statistical approaches to assess reference gene expression stability, which was ranked and tabulated by RefFinder.Host-cytokine expression (Fold-change; 2 -ΔΔCt ) were analysed using a Kruskal-Wallis ANOVA with a Dunn's multiple comparisons test.Analysis was conducted using GraphPad Prism version 7.0 (GraphPad).In all statistical analyses, a P < 0.05 was considered significant.

SDHA and TBP were identified as the most suitable reference genes for RTqPCR analysis of Plasmodium liver-burden and host-cytokine response
To identify the most suitable reference genes for determining both Plasmodium liver burden and host-cytokine responses in the liver post-challenge, Ct values for each reference gene (Fig. 1), were ranked for their stability with the RefFinder software package 46 , combining geNorm, NormFinder and BestKeeper analysis (Table 2).SDHA and TBP were ranked as the most stable genes and had a combined geNorm M stability value of 0.14 which falls below the established threshold of 0.15 for requiring additional reference genes 47 .Therefore, the inclusion of further reference genes beyond the two genes SDHA and TBP was not required.Notably, all packages ranked SDHA and TBP as the most stable genes, and RPL13a as the least stable gene.The most widely used reference gene GAPDH 2 was ranked 6th, 8th, and 9th most stable by geNorm, NormFinder and BestKeeper, respectively.

The threshold for partial protection is defined as two standard deviations below the mean of the infection control
When infection was allowed to progress to the blood-stage all IC mice and one in five PyCSP mice developed parasitemia (Fig. 2A).Using RTqPCR relative quantification of parasite rRNA in the liver, we could determine both sterile protection (i.e., the absence of P. yoelii 18s (Py18s) rRNA) and a reduction in parasite burden indicating partial protection (Fig. 2B).A high liver-stage parasite burden was found in all IC mice (Py18s Ct mean = 24.78 with σ = 0.96; Fig. 2B).We defined the LOD of the assay as Ct = 35, which provided a fold-reduction dynamic range of the assay (relative to the IC) as 2 ΔΔCT = 1189 (Supp Protocol.1).Furthermore, we defined 2 σ from the Py18s Ct mean of the IC as the threshold of partial protection (i.e., Threshold Ct = 26.70 or 2 ΔΔCT = 3.77; Fig. 2B), which demonstrated five PyCSP mice were partially protected, and five PyCSP mice were sterilely protected (Fig. 2B).This RTqPCR protocol detected degrees of liver-stage parasite burden, allowing for the interpretation of partial protection.www.nature.com/scientificreports/

Whole-liver host-cytokine expression responds to both immunisation and challenge
To detect clinically relevant cytokines in the host-whole liver using relative mRNA quantification, we found the optimum concentration of cDNA in the qPCR to detect IFN-γ was 50 ng of cDNA per reaction (Supp Fig. 1B).

Discussion
Herein, we describe a Plasmodium yoelii 18s (Py18s) rRNA-specific RTqPCR-based detection strategy with an optimised reference gene selection.This protocol defines 'partial protection' in the liver-stage following a sporozoite challenge and allows matched quantification of host whole-liver cytokine responses.Our assay provides an important update for pre-erythrocytic stage whole-liver Plasmodium parasite burden molecular diagnostics.The inappropriate selection of reference genes is a major contributor to the lack of reproducibility of RTqPCR data 20,21 .Previously published RTqPCR-based relative quantification strategies of Plasmodium liver burden are derived from a single reference gene 30 .Indeed, routine or habitual RTqPCR reference gene selection is common across multiple disciplines 48,49 .Using inappropriate reference genes for normalisation may result in the incorrect identification of fully or partially protected animals and misrepresentation of cytokine expression profiles.By analysing the variability of Ct values from 12 commonly cited host reference genes, we identified half were differentially expressed following immunisation and infection, emphasising the importance of systematic reference gene assessment.Although we have established that TBP and SDHA as highly suitable for RTqPCR relative normalisation in our model of P. yoelii sporozoite challenge and adenovirus vector-based 'Prime-Target' immunization, it is likely that other stably expressed reference genes may be identifiable with unbiased screening in other models [35][36][37] .TBP and SDHA have been identified as stable reference genes for human leukocyte RTqPCR analysis studies 22 .
It is widely acknowledged that reference gene expression stability testing must include all experimental conditions 20,21 , including vaccination 25 and challenge 50 , as these influence reference gene expression.Our study found statistically significant whole-liver reference gene expression instability in the expression of the commonly cited reference gene β-actin (βACT ) following a 'Prime & Target' immunisation regimen and parasite challenge.βACT can be differentially expressed under inflammatory conditions 51 .therefore, we speculate that the differential expression we observed in whole-liver βACT expression may result from antigen-independent adenovirus vector-based inflammation 41 .We found several reference genes were differentially expressed in infection control mice (i.e., IPO8, PGK1 and ALAS1) in response to sporozoite challenge.While many key host-parasite liver-stage immunological interactions remain unresolved 52 , innate or innate-adaptive interface immune responses to P. yoelii challenge may be driving differential whole-liver cytokine expression.Whilst we have assessed a 'Prime-Target' regimen followed by a 1000 P. yoelii sporozoite challenge, other vaccine regimens or challenges involving different sporozoite species or numbers, or other mouse strains will likely require an independent assessment of reference gene expression stability.
A significant advantage of liver-stage parasite burden RTqPCR quantification is the determination of the degree of pre-erythrocytic stage non-sterile protective immunity following sporozoite challenge.To provide a robust method to analyse parasite liver burden data and define 'partial protection' from sporozoite challenge, we made several key adaptations to the standard fold change (2 -ΔΔCt ) method.The first adaptation was to invert the method from fold-change to fold-reduction (2 -ΔΔCt vs 2 ΔΔCT ) relative to the infection control group.The second adaptation was to define the LOD as Ct = 35.The theoretical LOD (i.e., the lowest amount of measurable analyte) of qPCR is between one and three copies 21 , which under ideal conditions (i.e., a reaction efficiency of 100%) typically reaches cycle threshold (Ct) around cycle 35.Including results > 35 Ct significantly increases the falsepositive rate when performing pathogen detection qPCR 24,45 .We set our fold-reduction (i.e., 2 ΔΔCT ) data analysis strategy LOD to Ct = 35 and found a 1189-fold dynamic range from the mean of the IC.The third adaptation was to use this calculated fold dynamic range to provide values to samples from which no amplification occurred.As fold-change RTqPCR analysis is incapable of including 'undefined' samples 24 , an 'undefined' or 2 ΔΔCT > 1189 result was therefore given a value of 2 ΔΔCT = 1189.This strategy (Supp.protocol 1) avoids the use of setting non-detect Ct values to a threshold (i.e., Ct = 35), which can introduce substantial bias during normalisation 53 .
The final method adaption was to define a threshold of partial protection.We found the standard deviation (σ) of IC liver parasitemia (Py18s) was 0.96 Ct, and the Py18s of the IC and the reference genes of all groups were normally distributed.This consistency suggested the I.V. transmissibility of the sporozoites was high.Two standard deviations from the mean typically cover 95% of all intra-group data when normally distributed.Therefore, we defined 'partial protection' as 2 σ from the mean Py18s of the IC group to appropriately identify mice with a clinically relevant reduction of parasite liver burden.We did not employ any method to remove technical replicate outliers, as no obvious inappropriate technical variability was observed.However, care must be taken to ensure that results are not biased by high replicate variability, and methods to identify and remove replicate outliers must be reported 21 .
We have optimised a SYBR®-chemistry fold change (ΔΔCt)-based strategy to quantify the whole-liver expression of immunologically important cytokines to facilitate matched host-response and parasite burden assessment.We found the Py18s rRNA-specific SYBR® qPCR primers amplified non-specifically in the absence of Py18s in the sample.As reported, these readings were excluded based on an incorrect melt curve.It is likely a TAQpolymerase probe-based assay could eliminate the detection of this non-specific amplification, however, these results demonstrate that careful optimisation is required to ensure the probe does not bind to the non-specific amplicon.We utilised the ΔΔCt method due to its prevalent use in evaluating whole liver parasitemia 2,5,6,[26][27][28][29][30][31] .Although, normalisation methods like the Pfaffl method, which account for primer efficiency, could offer a more rigorous analysis of gene expression data 54 .
The simultaneous quantification of liver parasite burden and host-cytokine response in a standardised protocol is an important addition to pre-erythrocytic stage vaccine development, as this technique will increase the reproducibility of studies investigating the host immune response elicited during the pre-erythrocytic stage to vaccination and challenge 40 .The critical effector molecule of adaptive immunity to sporozoite challenge appears to be Interferon-gamma (IFN-γ) released by CD8 + T cells 34,55 , and Th1 CD4 + T cells secreting IFN-γ and Interleukin-2 (IL-2) 32,33,56 .The mRNA expression profiles of IFN-γ and many other rapidly produced and secreted cytokines are relatively highly correlated to protein production 43 .Therefore, transcriptomic quantification of host-cytokine responses will inform functional efforts to understand the immunological response following vaccination.A protocol that can evaluate mRNA expression of essential host effector genes following a challenge of 1,000 P. yoelii sporozoites is expected to provide the sensitivity required for most P. yoelii vaccine challenge models.Furthermore, it is anticipated that a similar strategy to quantify whole organ cytokine response could be applied to other immunisation, mouse strain, Plasmodium species, or other pathogen challenge rodent models.
Here, we present a protocol for the robust analysis of primary liver-stage Plasmodium infection and pre-erythrocytic stage immunity burden testing.We demonstrate that P. yoelii infection and 'Prime-Target' immunisation influence reference gene expression and identify SDHA and TBP as optimal reference genes for relative RTqPCR normalisation.We have established a criterion for defining partially protective immunity to infection and provide a customised fold-reduction method to provide a LOD and account for 'undefined' measurements.This assay is suitable for studying whole-liver host-cytokine mRNA responses, which are matched with a parasite-burden readout.This protocol is designed to be broadly adaptable across various murine models.While we anticipate the need for reference gene optimization may vary depending on the specific model, the protocol presented herein offers a systematic framework for identifying stable RTqPCR reference genes in mouse whole liver, determining 'partial' and 'sterile' protection, and assessing the expression of critical matched host immunomodulatory genes.This report provides an important update for further trials evaluating pre-erythrocytic stage whole-liver Plasmodium parasite burden and host response and highlights the importance of thorough selection of reference genes for RTqPCR.

Figure 1 .
Figure 1.Reference gene expression following 'Prime-Target' Immunisation and Plasmodium sporozoite challenge.Groups of BALB/c mice (n = 10/group) included naïve unchallenged (Naïve), naïve with sporozoite challenge (Infection Control), and Plasmodium yoelii circumsporozoite protein immunised and sporozoite challenged (PyCSP).Where appropriate, mice were intramuscularly immunised with plasmid DNA (Prime), followed 12 days later with an intravenous injection of the respective AdHu5 virus (Target).RNA was extracted from the homogenised whole liver at 5 weeks post-immunisation and 42 h post-challenge with 1,000 Py17XNL sporozoites.Cycle threshold (Ct) values were determined from the mean of triplicate replicate qPCR reactions, with the threshold set in exponential phase amplification at ΔRn0.3.Ct values of twelve candidate reference genes from two independent experimental replicates (n = 5/replicate) are shown.Data were analysed using oneway ANOVA with a Bonferroni-corrected multiple comparisons test comparing each group to the naïve mice (* P < 0.05).

Figure 2 .
Figure 2. Determination of the threshold of 'partial protection' in the liver stage.BALB/c mice (n = 10/ group) were immunised and challenged as described above (Fig. 1 legend): naïve (brown), infection control (IC) sporozoite challenged (red), and PyCSP immunised and sporozoite challenged (blue).Parasitemia over the duration of infection and at day 12 post-challenge (A) was measured by flow cytometry using the FCAB assay with mean ± technical SEM shown.Liver-stage parasite burden of individual mice was measured at 42 h post-challenge by technical triplicate RTqPCR (B).The Ct was determined from the mean of triplicate replicate, with data calculated based on Fold-reduction (2 ΔΔCt ) relative to the Ct geometric mean of the reference genes TBP and SDHA.Protection was defined as two standard deviations (2 σ = 1.91 Ct) below the mean Ct of the IC (dotted line).The limit of detection (LOD) was Py18s Ct = 35 (2 ΔΔCt = 1189).Data are pooled from two independent experimental replicates (n = 5/replicate).

Figure 3 .
Figure 3. Cytokine expression following Prime-Target immunisation and Plasmodium yoelii sporozoite challenge.BALB/c mice (n = 10/group) were immunised with a Prime-Target regimen and challenged with Py17XNL sporozoites as previously described (Fig. 1 legend).mRNA expression was assessed by RTqPCR in liver extracts harvested at 42 h post-challenge from naïve, infection control (IC) sporozoite-challenged (red), PyCSP-immunised and sporozoite-challenged (blue) BALB/c mice (n = 10/group).Data are pooled from two independent experimental replicates (n = 5/replicate).Fold-change was determined within each experiment with the delta-delta cycle threshold (2 -ΔΔCt ) method relative to the Ct geometric mean of the reference genes TBP and SDHA.Data were compared with a non-parametric Kruskal-Wallis one-way ANOVA with P-value displayed and a post-hoc Dunns-corrected multiple comparisons test comparing test groups to the mean of naïve mice (*, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001).

Table 1 .
Primer characteristics.Gene-specific forward and reverse primers acquired from Primer Bank™ database or literature.Assay performance determined as per MIQE guidelines.E' Reaction efficiency, R 2 standard coefficient of determination, (Mu)

Table 2 .
Reference gene expression stability as determined by RefFinder.