Reverse Transcription Can Critically Impact the Diagnostic Outcome of BCR::ABL1 Quantitative Real-Time RT-PCR

Simple Summary Reverse transcriptases (RT) play a crucial role in BCR::ABL1 fusion transcript monitoring of chronic myeloid leukemia (CML). RT enzyme and reaction conditions may contribute to impairment of stoichiometry in cDNA synthesis potentially biasing in qRT-PCR data. We have comparatively investigated the performance of the MLV-RT and SuperScript IV with random-hexamer vs. target-specific priming by means of the Acrometrix™ BCR::ABL1 reference panel and 37 clinical specimens. Our experiments identified priming type and RT type as major factors for diagnostic data variation. Variation was mainly due to different efficacies of RT enzymes to process low- (<50) and high-copy targets. The impairment of the high-performing SuperScript IV in processing low- (BCR::ABL1) and high-copy-number (GUSB or ABL1) RNA targets equally was not reflected by the diagnostically relevant Log (BCR::ABL1/GUSB%) values. For improving BCR::ABL1 assay sensitivity with a faithful representation of diagnostic targets, increased RNA/cDNA amounts and distinct RT/priming combinations are highly recommended. Abstract Reverse transcriptases (RT) are essential tools in BCR::ABL1 fusion transcript monitoring in chronic myeloid leukemia (CML). The RT type and cDNA priming method may impair the stoichiometry of cDNA synthesis, thereby potentially introducing a bias in BCR::ABL1 qRT-PCR data. Using the Acrometrix™ BCR::ABL1 reference panel and 37 clinical specimens, we have comparatively investigated the performance of the RTs MLV and SuperScript IV with random hexamer vs. target-specific priming. Quantitative RT-PCR results identified the priming type and RT type as major factors for diagnostic data variation, mainly due to the different efficacies of processing BCR::ABL1 low-copy-numbers (<50) compared to GUSB or ABL1 high-copy targets. The impairment of SuperScript IV in processing low- and high-copy-number RNA targets equally was not reflected by the diagnostically relevant Log (BCR::ABL1/GUSB%) values. Therefore, the correct representation of housekeeping and BCR::ABL1 target genes should have priority when aiming at as high a number of housekeeping gene copies as possible. Our data suggest that for improving BCR::ABL1 assay sensitivity, increased RNA/cDNA amounts and the use of distinct RT/priming combinations are advantageous. However, for inter-laboratory harmonization, the proper conversion factor according to the CML international standard (IS) has to be reevaluated each time the grade of RT is changed.


Chronic Myeloid Leukemia (CML) and Assessment of Molecular Response (MR)
In the molecular diagnosis of CML and the monitoring of the molecular response to tyrosine kinase inhibitor (TKI) treatment, qRT-PCR is considered the gold standard [1,2]. Milestones defined by the ELN according to Log reduction have been used to define the optimal response or treatment failure. Moreover, deep molecular remission (DMR), defined as BCR::ABL1 < 0.01% IS lasting for at least 2 years, is crucial for the decision of TKI treatment discontinuation. A strict technical standardization of BCR::ABL1 transcript measurements is mandatory, which also includes the selection of the most suitable control genes [1,[3][4][5][6][7]. Employing ABL1 or GUSB as control genes, current state-of-the-art qRT-PCR techniques are able to detect up to a 5-Log reduction in BCR::ABL1. The CML Working Group within the European Leukemia Net (ELN) has established definitions of MR that also reflect the sensitivity of the respective molecular tests used. MR4 indicates 4-Log reduction (BCR::ABL1 IS 0.01%), MR 4.5 indicates 4.5-Log reduction (BCR::ABL1 IS 0.0032%), and MR 5 indicates 5-Log reduction (BCR::ABL1 international standard (IS) 0.001%). There, the absolute copy numbers of house-keeping genes serve as indicators of RNA quality and grant the highest degree of qRT-PCR fidelity, accuracy, and reproducibility [5,8,9]. Since all pre-PCR steps, starting from total RNA extraction to the quantitative co-amplification of BCR::ABL1 and an internal housekeeping gene (GUSB or ABL1), can significantly affect the results, all steps must be continuously checked by an adequate quality and validation management [8,[10][11][12]. A comprehensive analysis of factors contributing to cDNA synthesis efficacy was recently performed by Jeromin and co-workers, who found that the use of high-fidelity RT such as SuperScript IV (ThermoFisher Scientific, Waltham, MA, USA) can push the sensitivity of BCR::ABL1 monitoring to a higher level [13].

Polymerization Activities of RTs and Potential Impact on BCR::ABL1 Monitoring
Reverse transcriptases are the most relevant tools in gene expression analysis and life science research [14]. Under native conditions, retroviral RTs are able to proceed through relatively long stretches of RNA/DNA and DNA/DNA duplices [15,16]. Under native conditions, the strand displacement (SD) synthesis activity of RTs is necessary in vivo for the completion of the reverse transcription of retroviral RNA genomes, as demonstrated for MLV [17,18].
As widely acknowledged, in CML, an appropriate measure of molecular responses defined as BCR::ABL1/housekeeping gene (HKG) % levels on the IS is required. Therefore, it is preferable that the RT enzymes perform a stoichiometrically correct reverse transcription, i.e., for each RNA molecule, just one cDNA molecule is generated. Particularly, low-(BCR::ABL1) and high-copy-number (GUSB or ABL1) targets should be proportionally processed with identical efficacy within a defined range of sample RNA concentrations. However, there are numerous reports on how the yield of cDNA from clinical samples could be improved, thereby challenging the question regarding the desired 1:1 stoichiometry [11,13,[19][20][21][22]. In fact, it seems technically difficult to robustly judge different RTs and distinguish between errors found in the RNA preparations used as a template and those caused by the RT itself or the type of priming (specific, random, oligo-dT, or a mixture of the latter) [23].
Aiming at the improvement of diagnostic sensitivity in CML monitoring, we set out to comparatively investigate the impact of two different RTs (MLV RT vs. SuperScript IV RT) and the type of priming (random vs. specific priming) on BCR::ABL1 assay performance. After subsequent highly standardized TaqMan qPCR on the Acrometrix™ BCR::ABL1 panel and clinical specimens (n = 37), we report the resulting Log levels (BCR::ABL1/HKG%) and discuss the best possible authentic measurements of the BCR::ABL1 load by choosing the optimal combination of the RT enzyme and cDNA priming.

Clinical Samples, Controls, and Ethical Considerations
As one of the German national reference laboratories for CML diagnostics, the strict international guidelines for CML monitoring were applied in our study [5,9,24]. Total RNA was extracted from clinical samples, employing the automated Maxwell1MDx technology (Promega, Mannheim, Germany). Peripheral blood (PB) samples of healthy volunteers were included in all processes to serve as a negative control for spillover and/or crosscontaminations during the RNA extraction and cDNA synthesis steps. In case of the false positivity of the control samples, all steps were repeated, starting from frozen backup material. Control blood samples of healthy donors were derived, fully anonymized, from the local blood bank. Written informed consent was obtained in accordance with the declaration of Helsinki for all human specimens used in this study. The diagnostic analyses performed according to Good Clinical Practice (GCP) guidelines included a total of 37 CML patients from various clinical trials (Dasfree, 2012-001421-27; DasaHIT, 2015-003502-16; ENDURE CML, 2016-001030-94) and were approved by the local Ethics Committee (Medizinische Ethikkommision II der Medizinischen Fakultät Mannheim der Ruprecht Karls-Universität Heidelberg, 2013-509N-MA and 212-247-AWB-MA). In general, RNA leftovers or frozen backup material from routine testing was used for our investigation. In accordance with the declaration of Helsinki, written informed consent was obtained from all patients (Table 1). To simulate the different Log levels of BCR::ABL1 monitoring (% BCR::ABL1/ABL1 value (IS)) by qRT-PCR analysis, the AcroMetrix™ BCR-ABL Panel (ThermoFisher Scientific), the first commercially available cell-based secondary reference panel for BCR::ABL1 quantification on the IS [25], was used. The panel includes five vials, each containing a distinct lyophilized mixture of the human cell lines K562 (BCR::ABL1 e14a2 fusion gene positive) and HL60 cells (BCR::ABL1 negative). Each vial contains 1 × 10 6 cells, in total. Overall, the AcroMetrix™ panel comprises panels A (10% BCR-ABL/ABL), B (1% BCR::ABL1/ABL1), C (0.1% BCR::ABL1/ABL1), D (0.01% BCR::ABL1/ABL1), and E (0.0032% BCR::ABL1/ABL1) and is intended for being used as an external control panel for the analytical evaluation of BCR::ABL1 test methods. We have employed the AcroMetrix™ BCR::ABL1 reference panel (lot no. 043019, ref. no. 956980) for investigating the potential impact of different RT and priming combinations on diagnostic BCR::ABL1 monitoring data. The resuspendation of lyophilisate and the extraction of RNA were performed according to the manual of the manufacturer.

cDNA Synthesis Employing Random Hexamer and Single Specific Priming Using MLV Reverse Transcriptase
One commercially available RT/cDNA first-strand synthesis kit used was the MLV reverse transcriptase system (200 U/µL) (Invitrogen/Thermofisher Scientific, Waltham, MA, USA, cat. no. 28025013). This RT is also being used in our routine CML monitoring diagnostics. For random cDNA synthesis, 1 µg of total RNA was diluted in 19 µL of bidest. H 2 O. The solution was incubated at 65 • C for 10 min and then 3 min on ice. Consecutively, 21 µL of RT-Mix was added, containing 12 U/µL MLV RT, 1.2 U RNasin Plus (Promega), dNTPs (10 mM each), 2 mM DTT, 2 × first strand buffer, and 0.1 U A260 (400 ng =~200 pg pmol pdN6 random primer (Roche, Mannheim). The solution was incubated for 2 h at 37 • C, followed by 5 min at 65 • C, and then stored at −20 • C. For the specific priming of cDNA synthesis, 1 µg of total RNA was diluted in 19 µL of bidest. H 2 O. The solution was incubated at 65 • C for 10 min and then for 3 min on ice. The RNA was also transcribed in a reaction volume of 40 µL using the same MLV RT-Mix described for random priming above, containing 100 pmols of target-specific primers. The target-specific primers were: RT-GUSBreverse 5 -CTCGC AAAAG GAACG CTGC-3 ; RT-ABL1-reverse 5 -CCTCC CTTCG TATCT CAGCG-3 . After adding 21 µL of RT-Mix, the reaction mixture was incubated for 2 h at 37 • C, followed by 5 min at 65 • C, and then stored at −20 • C. For the subsequent qPCR, 3 µL of the final cDNA reaction mix was used per reaction and well.

cDNA Synthesis Employing Random Hexamer and Single Specific Priming Using SuperScript IV Reverse Transcriptase
The second commercially available RT/cDNA first-strand synthesis kit used was the SuperScript IV reverse transcriptase system (Invitrogen/Thermofisher Scientific, cat. no. 18090010), MLV mutant featuring enhanced fidelity and thermostability and reduced RNase activity. In total, 1 µg of total RNA was transcribed according to the manufacturer's instructions, using 10 U/µL SS IV RT and 200 ng (=100 pmol) pdN6 random primer (Roche) or 50 pmol target-specific primers. At first, the total RNA and primer/RT-Mix was incubated for 5 min at 65 • C; then, 7 µL of RT reaction mix was added. Afterwards, the reaction mixture was incubated at 52 • C for 10 min and 80 • C for 10 min for specific priming using the target-specific primers RT-GUSB-reverse 5 -CTCGC AAAAG GAACG CTGC-3 and RT-ABL1-reverse 5 -CCTCC CTTCG TATCT CAGCG-3 . The reaction mixture containing the random primer hexamers was incubated for 10 min at room temperature. For the subsequent qPCR, 3 µL of the final cDNA reaction mix with the volume of 40 µL was used per reaction and well.

Data Calculation and Diagnostic Outcome
For BCR::ABL1 routine diagnostics, duplicate qPCR analyses were performed with 3 µL of cDNA reaction mix per qPCR reaction and well (corresponding to a total of 150 ng of RNA in two wells). For the RT experiments described here, triplicates (matching 500 ng of RNA) or hexaplicates (entire 1 µg of sample RNA) were performed, depending on the amount of cDNA available. In the latter case, the entire cDNA reaction volume was distributed over six qPCR reaction wells, and the resulting values from the six measurements were added up so that the total number of transcripts from 1 µg RNA could be determined. When triplicates were performed, the sum of all measurements was also calculated, and the value was extrapolated to a virtual amount of 1 µg RNA for better comparability between the different RT experiments. The same holds true for the Acrometrix™ BCR::ABL1 reference panel experiments. For the calculation of the absolute transcript numbers of BCR::ABL1, GUSB, and ABL1 and the BCR::ABL1/GUSB or BCR::ABL1/ABL1 quotients, all measured values of the triplicates or hexaplicates of one sample were summed up; no means of triplicate and hexaplicate values were considered [9]. Furthermore, it was only possible to calculate the laboratory quotients Log (BCR::ABL1/HKG%) according to Cross et al., 2015 [5]. Log (BCR::ABL1/HKG %IS) could not be calculated due to the lack of appropriate conversions factors for MLV specific priming and the SuperScript IV enzyme system (random and specific priming). Although it is possible in principle to establish conversion factors using the Acrometrix panel, this was not accomplished in our experimental setting because, for the comparative results, it seemed not mandatory to us to stick with the quotient [% IS].

Performance Comparison of MLV-RT and SuperScript IV-RT Using the Acrometrix™ BCR::ABL1 Reference Panel
We used the Acrometrix™ BCR::ABL1 panel, the first commercially available cell-based secondary reference panel for BCR::ABL1 quantification on the IS [25], for the quality assurance and validation of our routine diagnostic workflow, including our actual laboratoryspecific conversion factor. Both are based on reverse transcription, employing a combination of MLV RT and random hexamer priming followed by TaqMan qPCR according to a standardized protocol, as described previously [8]. As shown in Figure  amount of cDNA available. In the latter case, the entire cDNA reaction volume was distributed over six qPCR reaction wells, and the resulting values from the six measurements were added up so that the total number of transcripts from 1 µg RNA could be determined. When triplicates were performed, the sum of all measurements was also calculated, and the value was extrapolated to a virtual amount of 1 µg RNA for better comparability between the different RT experiments. The same holds true for the Acrometrix™ BCR::ABL1 reference panel experiments. For the calculation of the absolute transcript numbers of BCR::ABL1, GUSB, and ABL1 and the BCR::ABL1/GUSB or BCR::ABL1/ABL1 quotients, all measured values of the triplicates or hexaplicates of one sample were summed up; no means of triplicate and hexaplicate values were considered [9]. Furthermore, it was only possible to calculate the laboratory quotients Log (BCR::ABL1/HKG%) according to Cross et al., 2015 [5]. Log (BCR::ABL1/HKG %IS) could not be calculated due to the lack of appropriate conversions factors for MLV specific priming and the Super-Script IV enzyme system (random and specific priming). Although it is possible in principle to establish conversion factors using the Acrometrix panel, this was not accomplished in our experimental setting because, for the comparative results, it seemed not mandatory to us to stick with the quotient [% IS].

Performance Comparison of MLV-RT and SuperScript IV-RT Using the Acrometrix™ BCR::ABL1 Reference Panel
We used the Acrometrix™ BCR::ABL1 panel, the first commercially available cellbased secondary reference panel for BCR::ABL1 quantification on the IS [25], for the quality assurance and validation of our routine diagnostic workflow, including our actual laboratory-specific conversion factor. Both are based on reverse transcription, employing a combination of MLV RT and random hexamer priming followed by TaqMan qPCR according to a standardized protocol, as described previously [8]. As shown in Figure   Since, obviously, the Acrometrix™ BCR::ABL1 panel is an excellent tool for monitoring BCR::ABL1 assay performance, we used it to comparatively investigate the impact of two different reverse transcriptases (MLV RT vs. SuperScript IV RT) on BCR::ABL1 assay performance. Furthermore, the influence of the cDNA priming method was analyzed by using either target-specific primers (each one for GUSB and ABL1) or a mixture of random hexamers, as commonly recommended by the majority of RT manufacturers and by the BCR::ABL1 secondary reference panel protocol [25]. The total RNA extracted from each of the reference panel vials A-E was equally split into four portions. Four reverse transcription reactions were performed in parallel with either MLV RT (random vs. specific priming) or SuperScript IV RT (each enzyme with random vs. specific priming). The resulting molecule numbers detected by TaqMan qPCR were upcalculated to a 1 µg RNA input for better data comparability. Two representative panels, C (0.1% = MMR) and D (0.01% = MR4), are depicted in Figure 2A and 2B, respectively. Validation of our routine diagnostic workflow using the Acrometrix™ BCR::ABL1 panel that consists of 5 vials, here termed A, B, C, D and E. (A) TaqMan qRT-PCR employing our standardized reverse transcription protocol (MLV RT combined with random hexamer priming) revealed great linearity across the expected MR range (BCR::ABL1 10% down to 0.0032%). (B) The laboratoryspecific conversion factors 1.147 and 0.476 were used to achieve the acceptable range of IS accuracy when GUSB and ABL1 served as housekeeping genes (HKG), respectively. All data are based on triplicate assays. Abbreviations: HKG, housekeeping gene; IS, international scale. A Log (BCR::ABL1/GUSB %IS) of 0.1 corresponds to MMR.
Since, obviously, the Acrometrix™ BCR::ABL1 panel is an excellent tool for monitoring BCR::ABL1 assay performance, we used it to comparatively investigate the impact of two different reverse transcriptases (MLV RT vs. SuperScript IV RT) on BCR::ABL1 assay performance. Furthermore, the influence of the cDNA priming method was analyzed by using either target-specific primers (each one for GUSB and ABL1) or a mixture of random hexamers, as commonly recommended by the majority of RT manufacturers and by the BCR::ABL1 secondary reference panel protocol [25]. The total RNA extracted from each of the reference panel vials A-E was equally split into four portions. Four reverse transcription reactions were performed in parallel with either MLV RT (random vs. specific priming) or SuperScript IV RT (each enzyme with random vs. specific priming). The resulting molecule numbers detected by TaqMan qPCR were upcalculated to a 1 µg RNA input for better data comparability. Two representative panels, C (0.1% = MMR) and D (0.01% = MR4), are depicted in Figures 2A and 2B, respectively. As shown in Figure 2, where the measured copy numbers of BCR::ABL1, GUSB, and ABL1 detected by the four experimental settings were comparatively analyzed, we found that the number of targets (low-or high-copy) affects the efficacy of the respective priming type (random vs. specific). Moreover, the type of RT (MLV RT vs. SuperScript IV RT) has a crucial impact. So, when looking at MLV RT only, there was no (A = reference panel C) or only a minor effect (B = reference panel D) in the detection of BCR::ABL1 low copy numbers (n < 800), irrespective of the priming type applied. In contrast, for the high-copy targets, GUSB/ABL1 random priming revealed a 0.39-fold/0.57-fold (reference panel C) and 0.37-fold/0.52-fold (reference panel D) decrease in the number of detected molecules, respectively, when compared to specific priming. Thus, specific priming seems superior when it comes to the measurement of the HKG GUSB or ABL1, in contrast to the detection of BCR::ABL1 targets. The respective fold changes are shown in Figure 2C.
However, the high-fidelity RT enzyme SuperScript IV displays completely different reverse transcription characteristics, as, for all targets, random priming turned out to be superior to specific priming, irrespective of the target copy number and reference panel tested. This feature is reflected by the FC (random vs. specific priming) given in Figure 2C, where the FCs for the SuperScript IV RT range between 2.40 and 3.52 for both reference panels. This is in strong contrast to the performance of MLV RT that features FCs between 1.01 and 0.37 (random priming was inferior to specific priming).
What may this mean for the resulting clinical diagnostic outcome? The Log (BCR::ABL1/HKG %) is calculated according to the formula X [%] = [total number of molecules BCR::ABL1]/[total number of molecules HKG] × 100, in compliance with the guidelines of Foroni et al., 2011 [9]. Therefore, changes in the denominator corresponding to the number of detected HKG targets display a higher impact on the diagnostic outcome than the numerator (i.e., detected BCR::ABL1 copies). This may lead to a diagnostic bias. As shown in Figure 2A, in the case of MLV-RT (reference panel C), about equal BCR::ABL1 copies (=numerator) were detected, irrespective of the priming type, while the detection of the corresponding HKG GUSB turned out superior when specific priming was used. This yielded lower Log (BCR::ABL1/GUSB%) values when specific priming in combination with MLV RT was applied for reverse transcription ( Figure 2D). The same held true when ABL1 served as HKG. Thus, paradoxically, assaying clinical samples with MLV RT and specific priming may result in better-responding patients (i.e., lower Log (BCR::ABL1/GUSB%) values) compared to random priming. In the case of SuperScript IV RT, we found that the overall performance superiority of random priming affects both the numerator and denominator quite equally, and therefore, the cDNA priming type may account for only minor changes in the diagnostic outcome.

Performance Comparison of MLV RT and SuperScript IV RT Using Clinical Samples
To verify the Acrometrix™ BCR::ABL1 reference panel-based findings under routine diagnostic conditions, we performed analogous experiments with clinical specimens derived from routine laboratory BCR::ABL1 monitoring. Eleven samples (patients 1-11; for clinical data, see Table 1) were assayed using MLV RT in combination with random vs. specific priming ( Figure 3A-C). Due to the limitation of the clinical sample material, 10 different clinical RNA specimens (patients 12-21) had to be used for testing the SuperScript IV RT in an analogous manner ( Figure 4A-C).
The detection of BCR::ABL1 targets ( Figure 3A, copy range per 1 µg RNA: 12 (patient 8) to 375 (patient 1) when random-primed) after MLV RT-based reverse transcription (random vs. specific priming) revealed that random priming was superior to specific priming if the number of BCR::ABL1 targets exceeded the number of about 70. This resulted in an overall FC of 1.82 (ratio random/specific priming) for the BCR::ABL1 target. At target copy numbers below 70, specific priming seems to become more and more superior ( Figure 3A, patients 3,5,7,8,9), indicating a strong reciprocal relationship of target copy numbers with the priming type efficacy, as emphasized in Figure 3B, where the corresponding fold changes (FC: random vs. specific priming) are shown. For both the high-copy-number targets GUSB and ABL1, the MLV RT in combination with specific priming detected nearly twice as many target molecules when compared to random priming, resulting in a mean FC (ratio of random vs. specific priming) of 0.43 and 0.49 for GUSB and ABL1 targets, respectively ( Figure 3B). Since, consequently, MLV RT in combination with specific priming leads to a twice-as-high denominator as with random priming, the resulting Log (BCR::ABL1/GUSB%) values suggest a better diagnostic outcome than when patients are monitored with random-primed assays ( Figure 3C). In the extreme case (patient 1), this accounts for one entire BCR::ABL1 Log reduction step (random 0.1 = MMR vs. specific 0.0109 = MR 4 ). Since at BCR::ABL1 target copy numbers below 70, random and specific priming seem to generate converging target numbers ( Figure 3A, patients 3, 5, 7, 8, 9), the resulting Log (BCR::ABL1/GUSB%) reduction differences (random vs. specific priming) becomes less and less prominent (mean random 0.0074 (range 0.0034 to 0.0111 vs. mean specific 0.0044 (range 0.0030 to 0.0064)). This is most visible for patient 8, for whom an identical Log (BCR::ABL1/GUSB%) of 0.0034 could be stated, irrespective of the type of priming. However, except for patient 1, the approximate two-fold differences observed in the Log (BCR::ABL1/GUSB) are in the acceptable range (up to three-fold) of variability observed when using a conversion factor (CF) [26].    The detection of BCR::ABL1 targets ( Figure 3A, copy range per 1 µg RNA: 12 (patient 8) to 375 (patient 1) when random-primed) after MLV RT-based reverse transcription (random vs. specific priming) revealed that random priming was superior to specific priming if the number of BCR::ABL1 targets exceeded the number of about 70. This resulted in an overall FC of 1.82 (ratio random/specific priming) for the BCR::ABL1 target. At target copy numbers below 70, specific priming seems to become more and more superior ( Figure 3A, patients 3, 5, 7, 8, 9), indicating a strong reciprocal relationship of target copy numbers with the priming type efficacy, as emphasized in Figure 3B, where the corresponding fold changes (FC: random vs. specific priming) are shown. For both the high-copy-number targets GUSB and ABL1, the MLV RT in combination with specific priming detected nearly twice as many target molecules when compared to random priming, resulting in a mean FC (ratio of random vs. specific priming) of 0.43 and 0.49 for GUSB and ABL1 targets, respectively ( Figure 3B). Since, consequently, MLV RT in combination with specific priming leads to a twice-as-high denominator as with random priming, the resulting Log In the case of SuperScript IV RT, we confirmed the overall performance superiority of random priming over specific priming, as already observed for the Acrometrix™ BCR::ABL1 reference panel (Figure 2). For the low-copy target BCR::ABL1, random priming detected about 1.88-fold (mean of patients #15-19, #21) more molecules per 1 µg of RNA compared to specific priming ( Figure 4A). At molecule numbers below 250 (patients #12-14, #20), specific priming becomes superior, featuring an FC of 0.56 (mean) for BCR::ABL1. For the high-copy targets, random priming, when compared to specific priming, revealed a 1.70-fold and 2.43fold increase (mean) in the number of detected molecules for GUSB and ABL1, respectively. Thus, except for patients #12-14 and #20, the absolute assay sensitivity (per 1 µg RNA) seems to benefit from the combination of SSIV RT and random priming ( Figure 4B). This affects both the numerator and denominator quite equally, and therefore, despite a weak trend towards random priming superiority, the cDNA priming type accounts only for minor changes in the diagnostic outcome, which are within the acceptable range (up to threefold) of variability [26] (Figure 4C). However, for clinical samples with BCR::ABL1 target numbers below 250 (patients #12-14, #20), a 3.1-fold (mean random: 0.0114 vs. mean specific: 0.0353) Log (BCR::ABL1/GUSB%) reduction was observed. This points to a general trend where the corresponding patients seem to respond better when assayed with specific priming (i.e., lower Log (BCR::ABL1/GUSB%) values) than when specific priming was applied in combination with SuperScript IV (Figure 4C).

SuperScript IV in Combination with Specific Priming Grants the Highest Sensitivity for BCR::ABL1 Detection at Target Copy Numbers <50
In order to identify the best-performing reverse transcription procedure for monitoring therapy-free remission (TFR) and patients before stopping TKI treatment decisions, we performed analogous experiments with clinical specimens (n = 16) that are representative for the types of samples regularly encountered by a routine diagnostic workflow. Sixteen samples (patients #22-#37; for clinical data, see Table 1) were assayed using SuperScript IV (SSIV) in combination with random vs. specific priming ( Figure 5A-C). BCR::ABL1 copy numbers, as shown in Figure 5A in decreasing order, ranged between 135,000 (pat #22) and 6 (pat #37) when random priming was applied. According to our lab-specific workflow, only GUSB was employed as HKG (simultaneously tested with BCR::ABL1 within the same well) in this experiment. This enabled the execution of hexaplicate assays so that the complete amount of cDNA corresponding to 1 µg of the sample RNA was used up.
While for GUSB, a similar superiority of random over specific priming was detected (mean FC of 1.28, SD 0.011) as in the former experiment in Figure 4, an analysis of BCR::ABL1 confirmed our former finding that, below a certain copy number cut-off (here, n = 50), specific priming becomes superior to random priming for BCR::ABL1. This is depicted in Figure 5B, where patients #22 to #29 (n = 8) show random priming superiority (mean FC of 3.02, SD 1.64), while at copy numbers <50 (patients #30 to #37), specific priming became more efficient (mean FC 0.90, SD 0.22). For GUSB, random priming always shows a higher efficacy in terms of target molecule detection ( Figure 5B, blue columns). Within the range of the cut-off, two outliers (patient #27 and patient #31, depicted by open triangles) suggest that the exact cut-off may be slightly floating depending on experimental conditions. Thus, in the former experiment in Figure 4, the cut-off was at a BCR::ABL1 copy number <250, where specific priming outnumbered random priming (compare Figure 4B).
These findings find expression in the diagnostic important Log (BCR-ABL1/GUSB%) level quotient that clearly demonstrates that, in this experiment, the numerator (BCR::ABL1) and denominator (GUSB) follow different kinetics in reverse transcription depending on the priming type. At BCR::ABL1, copy numbers below the cut-off random priming result in lower Log (BCR::ABL1/GUSB%) quotients, making patients appear better-responding (mean FC 0.67, SD 0.16, range 0.46-1.02), while at copy numbers above the cut-off, an opposite trend could be observed (mean FC 2.48, SD 1.4, range 0.92-5.62). Although the acceptable range of variability observed when using a CF has been described approximately three-fold [26], the absolute number of BCR::ABL1 molecules detectable in a clinical specimen (per 1 µg RNA) may serve as an excellent basis for clinical decision making. Cancers 2023, 15, x FOR PEER REVIEW 12 of 18 While for GUSB, a similar superiority of random over specific priming was detected (mean FC of 1.28, SD 0.011) as in the former experiment in Figure 4, an analysis of BCR::ABL1 confirmed our former finding that, below a certain copy number cut-off (here, n = 50), specific priming becomes superior to random priming for BCR::ABL1. This is depicted in Figure 5B, where patients #22 to #29 (n = 8) show random priming superiority

Optimization of Reverse Transcription Proves Advantageous in Terms of TFR Monitoring
In order to prove the potential benefit of the best possible RT efficacy in terms of BCR::ABL1 detection, we have compared the diagnostic routine laboratory data ((Log BCR-ABL/GUSB %IS) shown in Table 1) of 26 patients (patient ID: 12 to 37) to an optimized RT protocol. In contrast to the routinely used method (MLV RT, random priming), this protocol combined SuperScript IV RT with random or specific priming. Quantitative PCR was performed on a 1 µg RNA/cDNA equivalent instead of 0.150 µg (1/6 of 1 µg RNA) only, the amount that applies for regular BCR::ABL1 routine qRT-PCR assaying (for details, see Materials & Methods). As shown in Table 2, the resulting higher sensitivity led to 12 individual Log level changes each for random and specific priming. For random priming, six patients (#14, #29-32, #36) stepped up one Log level (yellow), and five patients (#12, #13, #26-28) stepped up two Log levels (orange). One patient (#37) stepped down one Log level (MR 4.5 -> MR 5 ). Specific priming performed quite similar (stepped up one Log level, n = 7; two Log levels, n = 4). One patient (#21) reached MMR, although no MMR was diagnosed in routine diagnostics. Obviously, these Log level shifts were due to a higher cDNA amount used in general but also to the priming type-dependent RT reaction kinetics differing for low-(BCR::ABL1) and high-copy targets (GUSB). For example, in the case of patient #32 (compare Figure 5B), the fold changes (random vs. specific priming) for the reverse transcription of GUSB (FC = 1.26) differ considerably from those of BCR::ABL1 (FC = 0.58). This concurs with the "better" diagnostic outcome (patient #32 in Table 2, MR 4.5 ) when random priming was applied compared to specific priming. Specific priming increased the number of detected BCR::ABL1 molecules, resulting in a Log level shift MR 4.5 -> MR 4 . This is an excellent example of how alternative priming can impact the Log level calculation, making the patient's residual tumor load appear more or less eminent. Table 2. Log (BCR::ABL1/GUSB%) level changes when comparing data derived from routine laboratory diagnostics (MLV RT, random priming) to corresponding SS IV RT experiments (random/specific priming) from the same clinical sample, as depicted in Figure 4 (n = 10, patient IDs: 12-21) and Figure 5 (n = 16, patient IDs: [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37]

Discussion
In diagnostic applications, it is assumed that RT enzymes perform a stoichiometrically correct reverse transcription. Low-(BCR::ABL1) and high-copy-number (GUSB or ABL1) targets should be proportionally processed with identical efficacy within a defined range of sample RNA concentrations. In particular, this is of great importance in CML routine BCR::ABL1 monitoring, as the number of BCR::ABL1 cDNA targets is considered to exactly reflect the tumor load.
Our experiments with MLV RT and SuperScript IV on the Acrometrix™ BCR::ABL1 reference panel and 37 clinical specimens revealed that the question regarding the desired 1:1 stoichiometry has to be discussed, as already addressed by the findings of others [11,13,[19][20][21][22]. We found that the RT type (MLV vs. SuperScript IV), type of priming (random vs. specific), and target copy number (> or <50 copies of BCR::ABL1) are major factors for diagnostic data variation, and it became clear that the influence of RT on the diagnostic outcome has been clearly underestimated so far.
The key difference between both tested RT enzymes was their divergent processivity of high-and low-copy RNA targets with respect to the priming type. When the BCR::ABL1 RNA target copy number fell below a distinct threshold (<200 in Figure 4; <50 in Figures 3 and 5), specific priming tended to become superior over random priming. Obviously, the kinetics of both enzymes harmonize when very-low-copy targets are present in a high background of human total RNA. Thus, specific priming can be desirable when it comes to the detection of very few BCR::ABL1 targets, e.g., in the case of TFR monitoring or therapy decisions regarding the participation of CML patients in TKI cessation studies, as suggested previously [13].
The different properties of the two enzymes MLV and SuperScript IV may best be explained by comparing the respective working protocols given by the manufacturers. While MLV RT-driven cDNA synthesis takes place at 37 • C for 2 h, SuperScript IV RT performs at 52 • C for 10 min, pointing to a remarkable temperature resistance (steady ternary structure of polypeptide) combined with a much higher processivity according to the temperature coefficient Q10 rule given by the van't Hoff equation, saying that the reaction rate can double for each 10 • C rise in temperature. Recombinant RTs that can work at elevated reaction temperatures were found to be superior to low-temperature-processing enzymes and have been recommended for low-RNA-input applications [27]. Irrespective of the applied priming type, one should expect nearly identical results for the same RNA sample and RT enzyme. However, our data showed that the reverse transcription of highand low-copy-number target RNA is divergent depending on the priming type. Since the RNA target numbers in all reactions were identical, there must have been more successful BCR::ABL1-specific initiations in the random primed samples than in the specific primed sample. For specific priming, just one initiation per target molecule may be expected theoretically. It seems that random priming can lead to multiple initiations on one single RNA target that results in multiple cDNA molecules per RNA template molecule. This mechanism is supported by SD synthesis activity, known to be a natural feature of RT enzymes essential for retroviral replication in vivo. The SD synthesis activity of SuperScript IV in combination with random priming under high reaction temperatures may favor multiple initiations per RNA template, entailing some sort of linear "pre-amplification" of targets during cDNA synthesis. As a consequence of this RT side activity, it leads to an overestimation of themeasured targets in qPCR. It cannot be excluded that specific primers also contribute to SD synthesis activity by the unspecific binding of their 3 ends in a less random manner.
Moreover, primer binding kinetics are highly dependent on the reaction temperature. At 37 • C, random primer target binding may be comparable to specific primer binding, whereas at 52 • C, random priming can be expected to be much less efficient (Tm value and Gibbs free energy (∆G)). Thus, during cDNA initiation with MLV RT, the abundancy of template-bound random hexamers (200 ng = 100 pmol) may outnumber the number of functional RT molecules and lead to a general underrepresentation of processed RNA targets (RT is a limiting factor). Of course, the amount of the primer and enzyme, i.e., the relation of the primer and enzyme, could be optimized with regard to the expected number of respective templates. The impairment of MLV RT and SuperScript IV RT in combination with specific or random priming may not be reflected by the resulting diagnostically relevant Log (BCR::ABL1/HKG%) value, since a small numerator (BCR::ABL1) can be easily compromised by changes in the denominator (GUSB or ABL1). Patients with small BCR::ABL1 target numbers show deeper molecular responses (DMR) when assayed with random priming (i.e., lower Log (BCR::ABL1/GUSB%) values) than when specific priming was applied in combination with SuperScript IV ( Figure 4C), even if the number of detected BCR::ABL1 molecules was higher with random-primed cDNA ( Figure 4A, patients #15-#19). As a consequence, if an optimized protocol is applied (SuperScript IV combined with a random hexamer or specific priming depending on the BCR::ABL1 target copy number> or <50, respectively), the resulting higher sensitivity can lead to individual Log level changes. Alternative priming may have an impact on the Log level calculation, making the patient's residual tumor load appear more or less eminent. In other words, the goal must not be to have the highest amount of the housekeeping gene but to have the correct relation of the housekeeping gene and target gene. In our regular routine monitoring, this may be of minor importance, as a higher sensitivity of a lab-specific method will be corrected by the proper CF factor. The limits of variation can be up to three-fold when using a CF [26].
The sensitivity of a diagnostic laboratory is always normalized to that of the corresponding reference laboratory or its given standards-here, the Acrometrix™ BCR::ABL1 reference panel that was also established using random priming combined with the ABI High-Capacity cDNA reverse-transcription kit (ThermoFisher Scientific, order no. 4368814; [25]). However, when it comes to TFR monitoring or TKI discontinuation decisions, an increase in assayed RNA amounts and the use of distinct RT/priming combinations may be advantageous for achieving the highest BCR::ABL1 assay sensitivity possible. For us, it seems advantageous to measure BCR::ABL1 copy numbers additionally in a distinct amount of RNA (e.g., 1 µg), thereby omitting the influence of denominators (HKG) varying with the RT and priming type. This approach would require the exact determination of the RNA quantity and quality, the latter to be analyzed by chip-based methods calculating RNA integrity numbers (e.g., RIN calculated by an Agilent 2100 Bioanalyzer System). Since the detection of single or sets of housekeeping genes also depends on the global RNA quality, more efficient and effective methods are imaginable, such as looking at the total levels of gene expression across all genes, as suggested by others [28].
The ELN panel agrees that TFR is the significant goal in CML management and that TKI treatment discontinuation should be considered in patients with durable DMR [2,[29][30][31]. Multiple cohort studies imply relapse (loss of MMR) in about 50% of patients after TKI cessation, regardless of the TKI used [32,33]. Most molecular recurrences occur within the first 6-8 months after TKI discontinuation and trigger the restart of TKI therapy [30]. Since the duration of TKI therapy and DMR are considered to be the most important prognostic factors [3,34] for TFR success, the application of the BCR::ABL1 monitoring method with the highest sensitivity is crucial. Our patients ( Table 2, #12, #13, #30, #31), after TKI discontinuation (MR 4.5 ), stepped up one or two Log BCR-ABL/HKG % IS values when a more sensitive method was applied in monitoring after TKI discontinuation. Since, for these patients of our cohort, the real tumor load may be higher than that monitored by routine BCR::ABL1 assays (MR 4.5 ), future follow-up may be informative. On the record, all four patients are still in TFR.

Conclusions
The employment of modern, convenient-to-use cDNA first-strand synthesis kits has led to a common disregard for the influence of cDNA synthesis in molecular methods. Varying efficacies of different RT enzyme grades for processing high-and low-copy-number RNA targets depend on the priming type and can introduce a bias into the resulting gene expression data, shifting Log BCR::ABL1 % IS values and potentially impacting the monitoring outcome if the proper CF will not be applied. In the case of BCR::ABL1 monitoring for TKI discontinuation, we consider BCR::ABL1 assays with the best possible sensitivity crucial but also underline the importance of inter-laboratory harmonization and the permanent validation of laboratory methods on the basis of a series of highly standardized reference samples (i.e., the Acrometrix™ BCR::ABL1 reference panel). However, for inter-laboratory harmonization, the proper conversion factor according to the CML international standard (IS) has to be reevaluated, especially if the RT type or cDNA priming is changed. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: All data generated in this study have been included in the published article.