Patient samples were collected and RNA was processed as part of a previous CTA study [21]. Ethics approval for the analysis of these samples was obtained from the Human Ethics Research Committee at the University of Cape Town, Faculty of Health Sciences (HREC REF: 194/2012). Briefly, PB samples (10 mL EDTA) were collected every three months from diagnosis for up to two years during treatment with basic chemotherapy regimens (Table 1). The PB mononuclear cells (PBMCs) were isolated using density centrifugation with Ficoll Histopaque (Sigma-Aldrich, St. Louis, MO, USA) and analyzed via flow cytometry to determine the MAGEC1- positive populations as previously described [21]. The RNA was extracted using the QIAamp RNA blood mini kit (Qiagen, Hilden, Germany), treated with the TURBO DNase-free Kit (Ambion, Waltham, MA, USA) to remove residual DNA, and checked for purity and integrity (quality: 260/280 nm ratios >1.8; RNA integrity number (RIN) values: 7–10 [22], MIQE guideline compliant [23]) as per a previous study [15]. The stored RNA from 10 patients was used in the present study. Serum M protein and Sb₂M levels were measured by the National Health Laboratory Service (NHLS)/GSH (South Africa) using accredited methodology. We compared our novel monitoring method with standardized hematological monitoring methods. Notably, the monitoring of serum free light chains was not available at our state hospitals during the study period.

Multiplex assays were developed and validated as per MIQE guidelines, including the assessment of amplification efficiency and specificity, and limits of detection. Reverse transcriptase (RT) reactions using 1.5 mg RNA and random hexamers were performed as previously described [21]. The primers for coamplifying MAGEA3/ABL, PRAME/ABL, and BAGE2/ABL are listed in Table 2. PCR reactions (25 mL) contained 0.8 mM dNTPs, clear GoTaq Flexi buffer, 1.5 U GoTaq Hotstart (Promega, Madison, WI, USA), 4 mL cDNA (20% of RT reaction), 4.5 mM or 2 mM (MAGEA3) MgCl₂, 0.4 mM of each CTA primer/0.2 mM CTA probe and 0.2 mM of each ABL primer/0.1 mM ABL probe. The cycling conditions were as follows: 95°C for 5 min, 40 cycles of 95°C/30 s, 60°C/40 s (57°C for BAGE2), and extension at 72°C/40 s (MAGEC1/ABL amplification was performed as per a previous study) [21]. All reactions were performed on the Rotor-Gene 6000 PCR machine (Qiagen) and analyzed using the Rotor-Gene 6000 series software 1.7. Cq values were determined using appropriate thresholds depending on the probe fluorescence. Target genes were quantified relative to ABL (reference gene) expression (normalized ratio) using a modified Livak method [24], as follows: Ratio (reference/target)=2^{Cq(reference)-Cq(target)} or ΔCq. All RNA samples were analyzed in duplicate (acceptable CV, <20%). The assays were able to detect >5 copies of each CTA transcript type; positive/negative (pos/neg) Cq cutoff values were based on assay-specific controls. Cq values generated from logarithmic curves were scored as positive. An ABL control indicative of 10,000 ABL copies/reaction was included in each experimental run; samples with ABL Cq<ABL Cq of controls (indicating potential RNA degradation) were excluded from further analysis.

Approximately 2×10⁶ ethanol-fixed PBMCs from two patients prepared during the previous studies [15, 21] were washed twice in 40 mL ice-cold 1× PBS/1% fetal calf serum (FCS), collected via centrifugation (1,500 rpm for 5 min), and re-suspended in 500 mL cold 1× PBS/0.5% bovine serum albumin (BSA) buffer. Approximately 100 mL of this cell suspension was incubated with the following antibody combinations for 30 min at 4°C in the dark: MAGEC1- Alexa488/CD34-PE/CD45-PerCP and MAGEA3-Alexa647 or PRAME-Alexa647 and MAGEA3-Alexa647/PRAME-Alexa488/ CD34-PE/CD45-PerCP; MAGEC1-Alexa488/CD138-PE/CD45PerCP and MAGEA3-Alexa647 or PRAME-Alexa647. All antibodies were used at a final concentration of 5 mg/mL. MAGEC1 (Abcam, clone ab115351), MAGEA3 (Abcam, clone ab38496) and PRAME (Abcam, clone ab135600) antibodies were fluorescently labeled with the Mix-n-Stain Alexa 647 kit (Sigma-Aldrich). The labeled cells were washed with 1 mL cold 1× PBS/0.5% BSA buffer at 1,500 rpm for 5 min; the pellet was re-suspended in 1 mL cold 1× PBS/0.5% BSA buffer and stored in the dark until analysis. The cells were analyzed on a BD FACSCalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) using the Cell Quest Pro software, applying parameters and gates that had been previously optimized to minimize spectral overlap and increase sensitivity [15, 21].

PRAME transcripts were detected in all the 10 patients at diagnosis, and were observed in 81% of all the RNA samples. This demonstrated the high prevalence of expression of this CTA and a possible role in early pathogenesis, in agreement with our previous findings [20]. Similarly, all patients showed MAGEC1 expression at diagnosis, and 85% of the RNA samples showed detectable transcripts [21]. In contrast, only five patients showed MAGEA3 expression at any time point (positivity rate, 20%). However, this was not unexpected, as MAGEC1 and PRAME expression has been linked to early stage disease [17, 18, 25], and MAGEA3 with more advanced disease [26, 27]. BAGE2 has been reported in advanced (stage 3) disease only [20]; in line with the above, we observed BAGE2 in three patients with clinically progressing disease.

PRAME and MAGEA3 expression levels varied widely, demonstrating positive normalized ratios over a 4-log range (0.00001–0.87 and 0.0005–4.31, respectively); these data were comparable to MAGEC1 data [21]. However, BAGE2 expression was much more limited, with positive results expressed in the 2-log range (0.0008–0.0206).

To investigate whether monitoring PRAME transcripts could provide clinically relevant information, the expression pattern of PRAME in each patient was correlated with biochemical parameters and MAGEC1 qRT-PCR results.

PRAME expression was detected in all patient samples where serum M protein and/or Sb₂M levels were above the clinical threshold [28] (indicating active disease), or had stabilized or returned to baseline levels following therapy (patient achieved CR); thus this new approach has higher sensitivity for detecting residual disease than the techniques currently employed at our facilities. Significant changes in the levels of these protein markers were at least mirrored, but more importantly often preceded by changes in PRAME expression (examples shown in Figs. 1, 2). The high frequency of detection of PRAME made it an attractive target to assess circulating disease, especially where patients appeared clinically normal following therapy, but later progressed.

An example of this is shown in Fig. 1 (MM4), where the patient had stage1 disease at therapy initiation; the serum M levels were always within the normal range (<8.9 g/dL) and the Sb₂M levels decreased to clinically normal levels (<1.8 g/dL) after therapy. While PRAME levels dramatically decreased initially, it was still detectable throughout therapy (while no clinical signs of disease were recorded); a large 2-log increase was observed six months prior to a 10-fold increase in paraprotein and the detection of further lytic lesions (at follow-up).

Even in cases with stable Sb₂M and serum M levels following therapy for a significant period of time (Fig. 1 MM6), monitoring the patient with qRT-PCR revealed that their condition deteriorated before clinical changes were observed; a 2.6-fold increase in PRAME transcript levels was observed at least 10 months prior to the development of significant anemia (indicative of BM failure) and renal damage. The patient died soon after the last time point, demonstrating that monitoring of the Sb₂M/serum M levels alone failed to predict the clinical complications early enough to avoid adverse outcomes.

PRAME and MAGEC1 expression followed similar patterns in this patient cohort; PRAME levels decreased simultaneously with MACE1 in response to therapy, and increased at the same time (Fig. 1 MM1/MM6, Fig. 2 MM10) or shortly after MAGEC1 (Fig. 1 MM4, Fig. 2 MM3/MM8). Monitoring PRAME levels in conjunction with MAGEC1 levels did not offer an earlier relapse prediction compared to monitoring MAGEC1 alone. However, the increases in PRAME levels were sometimes more pronounced than those in MAGEC1 levels, and served to confirm MAGEC1 data and disease progression. In line with the hypothesized occurrence of a CTA cascade with disease progression, PRAME expression may indicate early clonal changes that may influence malignant cell behavior.

Although the frequency of MAGEA3 detection in this MM patient cohort was only 20%, any detectable expression was linked to the development of anemia (Hb levels <12/13 g/dL), with an average Hb level of 10.6 g/dL, as previously observed [20]. Its expression was also strongly linked to clinical progression of the disease (plasmacytoma, severe anemia, and death) (see Figs. 2, 3). Notably, in patient MM 8, MAGEA3 expression provided valuable information about disease states and imminent disease progression, which was not apparent with standard biochemical monitoring. At diagnosis, the biochemical parameters were significantly deranged, and the patient was diagnosed with stage3 (ISS) disease; elevated renal failure markers (creatinine), anemia, and bone lesions (RAB) were also observed. As per a previous report linking reduced Hb levels with MAGEA3 expression [20], this CTA was detected at diagnosis (Hb level at diagnosis, 6 g/dL). Following chemotherapy, the patient appeared to improve clinically (serum M and anemia). However, MAGEA3 remained detectable and transcript levels increased significantly during the last three months of the patient’s disease leading up to the detection of a plasmacytoma (extramedullary disease), and the subsequently died.

The relative expression of MAGEA3 was also compared to that of MAGEC1 and PRAME in the relevant patients (MM 3, 4, 8, 9, and 10); while the increases in both MAGEA3 and MAGEC1 transcript levels correlated with imminent death or clinical relapse, the increase in MAGEC1 was detected earlier, and there were no time points at which MAGEA3 was expressed in the absence of MAGEC1. MAGEA3 expression increased at the same time (Fig. 2 MM3 and MM8) or after PRAME expression increased (Fig. 2 MM10, Fig. 3 MM9). Although the sample size was small, overall MAGEA3 expression data offered no additional advantage over MAGEC1 data for cell-level monitoring, but could be used to detect clonal changes that may affect therapy responses and possible spread of disease.

Although BAGE2 expression in our patient cohort was extremely limited, its expression was clearly associated with known indicators of stage 3 disease (as per the DSS and ISS systems), namely elevated creatinine levels (>90 mmol/L), severe anemia (<10 g/dL), and/or very high Sb₂M levels (>5.5 mg/L) [29, 30]. Fig. 3 shows the changes in BAGE2 expression in patients where it was detected (MM6, 8, and 9). Although detection of BAGE2 transcripts was always associated with a prior increase in MAGEC1 transcripts, its detection was associated specifically with advanced clinical disease indicators such as elevated creatinine (indicating organ damage), rather than just elevated PC levels (increased paraprotein). This could possibly indicate advanced clonal changes to the malignant cells in a cascade-like manner, which could potentially affect the proliferation rate, resistance to apoptosis, and thus overall drug responsiveness; these features can guide treatment for the predicted relapse. The CTA expression patterns in patient MM9 (Fig. 3) were in line with this cascade theory, and additional CTAs (MAGEC1-PRAME-MAGEA3-BAGE2) were expressed within the PBMC component as anemia, renal damage, and M protein levels increased.

The co-expression of multiple CTAs in this patient group indicated a cascade-like pattern as disease progressed, as previously inferred [17-20]. However, as qRT-PCR was performed on enriched PBMCs and not isolated cell populations, it was not possible to establish whether cells were undergoing sequential clonal changes, or whether different clones, or even different cell types, expressed single CTAs. Using limited residual ethanol-fixed PBMCs from two patients from the initial studies (MM3 and MM4) [15, 21] where MAGEC1-expressing cells have been extensively characterized, we additionally analyzed the expression of PRAME and MAGEA3 antigens in the matched samples (MM3 and MM4); co-expression of MAGEC1/PRAME/MAGEA3 mRNA was observed in these samples. While this study offers only preliminary results due to the limited sample pool and cell numbers, the results indicated that PRAME and MAGEA3 were only associated with the populations previously shown to express MAGEC1, namely a pro-B/pre-B progenitor population (CD34/CD19) and a PC sub-fraction (CD138+) (Table 1) [15]. We detected the co-expression of PRAME/MAGEC1 and MAGEC1/MAGEA3 in the CD34+ population and PRAME/MAGEC1 in the PC fraction (previously shown to be a proliferating population) [15]. No cells were found to express PRAME or MAGEA3 only; only small numbers of MAGEA3-positive stem cells were detected, and they were all found to be PRAME-positive, implying sequential antigen expression as opposed to activation in separate clones, although larger cell populations are required to confirm these findings.