(ERIC) Update on the International Harmonised Approach for Flow Cytometric Residual Disease Monitoring In

ORIGINAL ARTICLE Improving efﬁciency and sensitivity: European Research Initiative in CLL (ERIC) update on the international harmonised approach for ﬂow cytometric residual disease monitoring in CLL

AC Rawstron1,2,SBo¨ ttcher3, R Letestu4,5, N Villamor6, C Fazi7, H Kartsios1, RM de Tute1, J Shingles1, M Ritgen3, C Moreno8, K Lin9, AR Pettitt9, M Kneba3, E Montserrat6, F Cymbalista4,5, M Hallek10, P Hillmen11 and P Ghia7 on behalf of the European Research Initiative in CLL (ERIC)

Detection of minimal residual disease (MRD) in chronic lymphocytic leukaemia (CLL) is becoming increasingly important as treatments improve. An internationally harmonised four-colour (CLR) flow cytometry MRD assay is widely used but has limitations. The aim of this study was to improve MRD analysis by identifying situations where a less time-consuming CD19/CD5/k/l analysis would be sufficient for detecting residual CLL, and develop a six-CLR antibody panel that is more efficient for cases requiring full MRD analysis. In 784 samples from CLL patients after treatment, it was possible to determine CD19/CD5/k/l thresholds that identified cases with detectable MRD with 100% positive predictive value (PPV). However, CD19/CD5/k/l analysis was unsuitable for predicting iwCLL/NCI response status or identifying cases with no detectable MRD. For the latter cases requiring a full MRD assessment, a six-CLR assay was designed comprising CD19/CD5/CD20 with (1) CD3/CD38/CD79b and (2) CD81/CD22/CD43. There was good correlation between four-CLR and six-CLR panels in dilution studies and clinical samples, with 100% concordance for detection of residual disease at the 0.01% (10 À 4) level (n ¼ 59) and good linearity even at the 0.001–0.01% (10 À 5–10 À 4) level. A six-CLR panel therefore provides equivalent results to the four-CLR panel but it requires fewer reagents, fewer cells and a much simpler analysis approach.

Leukemia (2013) 27, 142–149; doi:10.1038/leu.2012.216 Keywords: chronic lymphocytic leukaemia; minimal residual disease; ﬂow cytometry

INTRODUCTION This method is not directly quantitative and has variable sensitivity The outcome of chronic lymphocytic leukaemia (CLL) has changed and specificity. However, in some situations the use of highly remarkably during the last decades1,2 with intensive and/or sensitive MRD flow cytometry may not be necessary, for example, combination therapies now capable of inducing long-lasting when there are still proportionally high levels of CLL cells in the clinical remissions in the majority of patients. As many patients presence of a normal lymphocyte count. In such situations CD19/ achieve complete remissions, the quantification of minimal CD5/k/l alone may be already informative. We therefore aimed at residual disease (MRD) gained importance. MRD levels during determining whether and when CD19/CD5/k/l analysis is and after therapy were shown to be independent predictors of sufficient to definitively confirm or exclude the presence of progression-free and overall survival in CLL.3–5 residual disease in situations where a quantitative result is not Quantification of MRD can be achieved using RQ-ASO IgH-PCR required. Moreover, with MRD data using CD19/CD5/k/l analysis or multi-parameter flow cytometry.6,7 An international available from recent clinical trials and series, it is important to standardised approach for flow cytometric four-CLR analysis has firmly establish the relationship between those analyses and the been developed that is applicable to all sample types and quantitative MRD methods. treatment regimes.8 However, the procedure is time consuming, Conversely, in those cases requiring a comprehensive MRD may be unnecessary for many patients with obvious CLL present evaluation, the increasing availability of cytometers that can and the interpretation of results may be difficult for non- analyse six or more fluorochromes in parallel makes it possible to specialised laboratories. reduce the cost and complexity by developing a protocol using six A simpler flow cytometric approach for MRD detection would or more CLRs. To this purpose we aimed at developing and include an initial screening for CD19/CD5 coexpression analysis standardising a six-CLR antibody panel to reduce sample coupled with clonality assessment using surface k/l expression.9–13 requirements and time taken for acquisition and analysis.

1HMDS, St. James’s Institute of Oncology, Leeds Teaching Hospitals, Leeds, UK; 2Epidemiology and Cancer Statistics Group, University of York, York, UK; 3Second Department of Medicine, University Hospital Schleswig-Holstein, Campus Kiel, Kiel, Germany; 4AP-HP, Hoˆpital Avicenne, Service d’He´matologie Biologique, Bobigny, France; 5Universite´ Paris Nord, PRES Sorbonne Paris Cite´, UMR Inserm U978 )Adaptateurs de Signalisation en He´matologie*, UFR SMBH, Bobigny, France; 6Unitat d’Hematopatologia, Hospital Clıńic and Institut d’investigacions Biome`diques August Pi i Sunyer (IDIBAPS), Barcelona, Spain; 7Laboratory of B Cell Neoplasia and Unit of Lymphoid Malignancies, Division of Molecular Oncology and Department of Onco-Hematology, Universita` Vita-Salute San Raffaele and Ospedale San Raffaele, Milan, Italy; 8Department of Hematology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain; 9Department of Hematology, Royal Liverpool and Broadgreen University Hospitals NHS Trust, Liverpool, UK; 10Klinik I fu¨r Innere Medizin, University of Cologne, Ko¨ ln, Germany and 11Department of Clinical Haematology, St. James’s Institute of Oncology, Leeds Teaching Hospitals, Leeds, UK. Correspondence: Dr AC Rawstron, HMDS, St. James’s Institute of Oncology, Leeds Teaching Hospitals, Bexley Wing, Beckett Street, Leeds LS1 3EX, UK. E-mail: [email protected] Received 3 May 2012; revised 18 June 2012; accepted 25 June 2012; accepted article preview online 31 July 2012; advance online publication, 5 October 2012 ERIC update on CLL MRD AC Rawstron et al 143 PATIENTS AND METHODS leukocytes, (ii) the percentage of CD19 þ cells expressing CD5, (iii) the Patients percentage of CD19 þ and CD19 þ CD5 þ cells lacking surface immunoglobulins and (iv) the k/l ratio on CD19 þ and CD19 þ CD5 þ cells. For A total of 1155 samples from CLL patients were assessed in this study: 784 specificity/sensitivity analysis, the four-CLR flow cytometry MRD assay samples from patients during or after treatment for comparison of CD19/ quantitative result was used as the gold standard, with a CLL cell CD5/k/l analysis with four-CLR MRD flow cytometry; 304 samples from À 4 patients at presentation or relapse sent for routine diagnostic immuno- percentage of leukocytes at or above 0.01% (10 ) classified as positive. phenotyping (270 for hierarchical cluster analysis of surface antigen The PPV was therefore defined as the number of cases above the CD19/ expression, 6 samples for dilution studies to investigate the impact of CD3 CD5/k/l threshold for predicting presence of residual disease that had X0.01% (10 À 4) CLL detected by the four-CLR MRD flow panel (that is, true in different fluorochrome combinations, 22 samples for optimising positive) as a proportion of the total number of cases above the CD19/ fluorochromes used and 6 samples for dilution studies to compare four- k l CLR and six-CLR analysis); and 67 samples from patients with CLL after CD5/ / threshold for predicting presence of residual disease, indepen- treatment for comparison of four-CLR and six-CLR MRD flow cytometry. All dent of the four-CLR MRD flow panel result (that is, true positive þ false patients were diagnosed according to the iwCLL/NCI-WG criteria.14 positive). Patients were included before, during and after treatment with For evaluation of the requirement to use CD3 to exclude contamination with different antibody combinations (CD19/CD5/CD3 combined with chlorambucil, fludarabine-based combinations þ / À rituximab, 3,7,15–19 CD20/CD79b/CD38, or CD81/CD22/CD43, or CD81/CD79b/CD43), compar- alemtuzumab and autologous/allogeneic transplantation. All samples were taken with full-informed patient consent and analysis was ison of the numbers of events classified as CLL from FCS files without CD3 performed according to the requirements of the local ethics review data (electronically removed) against those with CD3 data was performed committee. using a paired t-test analysis. For assessment of the limit of detection in dilution studies, the error was defined as the difference between observed number of CLL cell events and Flow cytometry actual number of CLL cells events as a percentage of actual CLL cell events Samples were prepared using participating laboratories standard proce- where the latter was calculated from the known CLL dilution. dures as reported previously, either using ammonium chloride prestaining or FACSlyse post-staining.3,5,7,8,17–19 In six CLL cases, dilution studies have been carried on to assess the limit of detection by diluting CLL cells into RESULTS bone marrow or peripheral blood containing only normal B cells. An initial CD19/CD5 and clonality assessment for the quick identification of 1:10 dilution of CLL cells into normal blood/marrow and then six further samples with residual disease: a poor negative predictive value serial 1:3 dilutions were made to obtain a series of 42 samples with known concentration of CLL cells ranging from 0.001 to 1% of leukocytes. Peripheral blood and bone marrow samples from 784 patients For evaluation of the requirement to use CD3 to exclude contamination with CLL after treatment were assessed by a basic analysis with different antibody combinations, simulated minimal disease samples comprising antibodies to CD19, CD5, k and l immunoglobulin were prepared by diluting six CLL cases into bone marrow or peripheral light chains and compared with results derived from the complete blood containing only normal B cells. The cells were incubated with CD19/ harmonised four-CLR CLL MRD panel reported previously.8 The CD5/CD3 coupled with either CD20/CD79b/CD38, or CD81/CD22/CD43, or assessment of k and l ratios were standardised as outlined in CD81/CD79b/CD43. Data for 500 000 events from seven dilutions plus a Figure 1. CLL cells represented a median 0.5% of leukocytes, range sample containing only normal B cells were acquired in each case. From each dilution set, six files were generated that were predicted to contain o0.01–23.7%. CLL cells represented o0.01% of leukocytes in 212/ either no CLL events (n ¼ 2), 20–50 CLL events (n ¼ 2) and 50–100 CLL 784 (27%) of cases, 0.01–1.0% in 206/784 (26%) of cases and more events (n ¼ 2). An identical set of data without CD3 was prepared by than 1% in 366/784 (47%) of cases. electronically removing the information from the original FCS files. The files Clonality assessment is conventionally considered in terms of were blinded and the centres analysed the five-CLR data with no CD3 first, the ratio of k þ to l þ B cells with a ratio below 0.3:1 or above 3:1 reporting the number of CLL and total B-cell events. Once the participants typically considered to be abnormal,20 although other thresholds reported the five-CLR data they were sent the FCS files, which included may be more informative.21,22 The proportion of B cells expressing CD3 data. CD5 under normal circumstances is typically below 30%; in The antibody combinations used in the comparison of four-CLR and six- previous studies, threshold levels of 10(ref. 11)and 25%9,10 have CLR assays are shown in Table 1 and were supplied by BD Biosciences (Erembodegem, Belgium). The same four-CLR antibody combinations were been used to classify residual disease status but CD5 expression used in the assessment of CD19/CD5/k/l analysis with minor exceptions as on up to 90% of normal B cells has been reported in regenerating 23 previously reported.17 A uniform gating strategy for CD19/CD5 clonality blood after intensive therapy. However, MRD can still be assessment was used in all cases, shown in Figure 1. For comparison of detected in a high proportion of cases with a normal B-cell k:l CD19/CD5/k/l against four-CLR MRD detection and for comparison of ratio and CD5 coexpression levels.24 In this series, the four-CLR four-CLR against six-CLR MRD detection the data was acquired and MRD analysis detected residual CLL above the 0.01% (10 À 4) level analysed individually within each laboratory. Analysis of inter-laboratory in 42% (90/213) of individuals with a normal k:l ratio (1.0–2.1:1), variation was performed using flow cytometry data files. and in 31% (16/51) of individuals with o10% B-cell CD5 coexpression. Conversely, the four-CLR MRD analysis demon- Statistical analysis strated no evidence of residual disease in 5.2% (20/383) of The following parameters were calculated from the CD19/CD5/k/l tube: individuals with an abnormal B-cell k:l ratio (o0.3 or 43:1) and in (i) the CD19 þ and CD19 þ CD5 þ event number and percentage of 12% (68/545) individuals with more than 30% CD5 expression on

Table 1. Antibodies used for four-CLR and six-CLR analysis

Test FITC PE PerCP-Cy5.5 PE-Cy7 APC APC-H7

4-CLR 1 l (1-155-2) k (TB28-2) CD19 (SJ25C1) — CD5 (L17F12) — 4-CLR 2 CD45 (2D1) CD14 (MOP9) CD19 (SJ25C1) — CD3 (SK7) — 4-CLR 3 CD20 (L27) CD38 (HB7) CD19 (SJ25C1) — CD5 (L17F12) — 4-CLR 4 CD81 (JS-81) CD22 (S-HCL-1) CD19 (SJ25C1) — CD5 (L17F12) — 4-CLR 5 CD43 (1G10) CD79b (SN8) CD19 (SJ25C1) — CD5 (L17F12) — 6-CLR 1 CD3 (SK7 CD38 (HB7) CD5 (L17F12) CD19 (SJ25C1) CD79b (SN8) CD20 (L27) 6-CLR 2 CD81 (JS-81) CD22 (S-HCL-1) CD5 (L17F12) CD19 (SJ25C1) CD43 (1G10) CD20 (L27) Abbreviation: CLR, colour.

& 2013 Macmillan Publishers Limited Leukemia (2013) 142 – 149 ERIC update on CLL MRD AC Rawstron et al 144 a b 250 105

200 104 150

103

Side Scatter 100 CD19 PerCP-Cy5.5

50 0

-472 M 50 100 150 200 250 -102 0102 103 104 105 -472 Forward Scatter (x 1,000) CD5 APC cde κ+λ− κ+λ− κ+λ− 105 105 105

104 104 104 PE PE κ κ 3 103 103 10 CD19 PerCP-Cy5.5 2 102 102 10 0 0 0 -102 -102 -102 κ−λ κ−λ κ−λ -472 -472 -472 M M M -1020102 103 104 105 -1020102 103 104 105 -1020102 103 104 105 -472 -472 -472 λ FITC λ FITC λ FITC Figure 1. Uniform gating strategy used for evaluation of CD19/CD5/k/l expression. (a) Mononuclear cells were identified according to light scatter characteristics (region ‘MNC’). (b) Within the mononuclear cell region, events were subclassified according to CD19 and CD5 expression with vertical and horizontal quadrants set to the edge of the CD19 À CD5 À population. (c) Quadrants for analysis of k and l expression were set using the CD19-CD5 þ T-cell population to define the lower boundaries for expression of k or l light chain, where B cells with light chain fluorescence within the limits of T-cell fluorescence levels were considered to lack expression of that light chain. The quadrants for k and l expression were then applied to CD19 þ CD5 À (d) and CD19 þ CD5 þ (e) mononuclear cells. At least 50 events in each region were required to calculate a parameter (for example, to calculate the CD19 þ k:l ratio would require at least 50 CD19 þ k þ l À and 50 CD19 þ k À l þ events). In this case, the B cells represent 1.0% of leukocytes, the CD19 þ k:l ratio was 1.0:1; 29% of CD19 þ cells expressed CD5; the CD19 þ CD5 þ k:l ratio was 0.33:1 and 11% of CD19 þ CD5 þ were sIg À . The four-CLR MRD analysis detected CLL cells representing 23% of total B cells; the CLL cells had weak sIgl expression and because of the background of polyclonal CD19 þ CD5 À and CD19 þ CD5 þ B cells, the only parameter to show significant skewing was the CD19 þ CD5 þ k:l ratio, although this did not reach a level consistent with 100% PVV for the presence of MRD.

the B cells. This datum confirms that the conventional CD19/CD5/ CD19/CD5/k/l results cannot exclude the presence of residual k/l thresholds have a poor negative predictive value because disease, it was possible to determine threshold levels for there is a high proportion of cases with residual disease but normal parameters within the CD19/CD5/k/l analysis that could identify CD19/CD5/k/l levels. However, the positive predictive value (PVV) samples with residual disease detectable through the standar- of CD19/CD5/k/l could be informative if it is possible to optimise dised four-CLR assay with 100% PVV. Some of these thresholds the thresholds used to identify cases with residual disease. (see Table 2) when applied to the validation set were able to confirm a 100% PVV for the presence of residual disease. To obtain such a high PVV, the selected parameters had to be heavily CD19/CD5 and clonality assessment for the quick identification of skewed from normal situation. Thus, coexpression of CD5 on more samples with residual disease: a 100% positive predictive value than 82% of B cells, a k:l ratio on the CD19 þ CD5 þ B cells To identify the optimal thresholds for positive prediction of the p0.05:1 or X32:1 or more than 54% of CD19 þ CD5 þ cells having presence of MRD, the series was divided into a training set and a no detectable surface immunoglobulins were required to predict validation set comprising cases matched for the percentage of the presence of MRD with a 100% PPV. Though using these very B cells and the proportion of cases with X0.01% (10 À 4) CLL cells restrictive criteria, this approach was informative in a high by four-CLR (CLR) flow cytometry (training set: n ¼ 392, B-cell proportion of samples as the presence of residual disease percentage median 3.93%, range 0.0041–97.9%, 289/392 with detectable by the four-CLR MRD flow panel could be predicted X0.01% CLL; validation set: n ¼ 392, B-cell percentage median in 185/392 (47%) samples by simply using CD19/CD5/k/l analysis. 3.92%, range 0.0038–97.4%, 283/392 with X0.01% CLL). Specificity and sensitivity analysis was performed on the CD19/CD5 and clonality assessment for the quick identification of training set (see Figure 2), and informative thresholds were samples with residual disease: correlation with clinical response evaluated on the validation set. As expected, there were no The correlation between CD19/CD5/k/l analysis and iwCLL/NCI parameters for which a threshold could be identified that showed response status was evaluated in 27 bone marrow samples from a 100% negative predictive value for the presence of residual single centre to determine whether it may be possible to reduce disease because a high proportion of cases with normal CD19/ the number of biopsies required for response assessment. Ten of CD5/k/l have low levels of residual disease. Although a normal the 27 cases had CD19/CD5/k/l parameters within levels

Leukemia (2013) 142 – 149 & 2013 Macmillan Publishers Limited ERIC update on CLL MRD AC Rawstron et al 145 Optimal threshold Specificity: TN/(TN+FP) 0.7 1 0.9 0.6 0.8 0.5 0.7 0.6 0.4 0.5 0.3 0.4 0.2 Specificity 0.3 0.2 0.1

Sensitivity + specificity -1 0.1 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Percentage of CD19+ B-cells expressing Percentage of CD19+ B-cells expressing CD5 CD5

Optimal threshold Specificity: TN/(TN+FP) 0.8 1 0.7 0.9 0.8 0.6 0.7 0.5 0.6 0.4 0.5 0.3 0.4 Specificity 0.3 0.2 0.2 Sensitivity + specificity -1 0.1 0.1 0 0 0.01 0.1 1 10 100 0.01 0.1 1 10 100 CD19+CD5+ B-cell κ:λ ratio CD19+CD5+ B-cell κ:λ ratio Figure 2. Sensitivity and specificity analysis for identifying residual disease based on the percentage of B cells expressing CD5 in plots (a and b), and the B-cell k:l expression ratio in plots (c) and (d). Plots (a and c) show the sensitivity þ specificity-1 for identifying the presence of residual disease above a 0.01% (10E-4) threshold by the standardised four-CLR cytometry method, where a value of 1 equates to a threshold with 100% specificity and 100% sensitivity. Plots (b and d) show the specificity for identification of a sample with X0.01% (10E-4) residual disease.

Table 2. Thresholds for predicting the presence of residual CLL using CD19, CD5, k and l expression

Training set: optimal threshold Training set: threshold for Validation set: positive for speciﬁcity vs sensitivity 100% positive predictive value predictive valuea

Percentage of total leukocytes expressing CD19 45.5 (n ¼ 165) 48.9 (n ¼ 134) 99.2% (n ¼ 132) CD19 þ k:l ratio o0.5: 1 or 48.2:1 (n ¼ 198) o0.04:1 or 461:1 (n ¼ 91) 98.8% (n ¼ 85) Percentage of CD19 þ expressing CD5 470% (n ¼ 185) 482% (n ¼ 160) 100% (n ¼ 144) CD19 þ CD5 þ k:l o0.6:1 or 45.7:1 (n ¼ 218) o0.05:1 or 432:1 (n ¼ 146) 100% (n ¼ 128) Percentage of CD19 þ CD5 þ lacking sIg 421% (n ¼ 123) 454% (n ¼ 43) 100% (n ¼ 40) The optimal threshold for specificity vs sensitivity was defined as the parameter value which provided the highest level for sensitivity þ specificity-1. aThe positive predictive value for the validation set was determined using the threshold for 100% positive predictive value identified in the training set. The values in brackets are the numbers of cases exceeding the threshold levels for the relevant parameter (training set total n ¼ 392, validation set total n ¼ 392). associated with a low predictive value for the presence of residual CD19/CD5/k/l analysis therefore has limited potential to disease (CD19 þ o5.5% of total leukocytes, CD19 þ k:l ratio of predict the likelihood of morphologically detectable disease in 0.5–8.2:1, and in cases with sufficient CD5 þ B cells for the bone marrow because the analysis may detect very low levels enumeration o70% of B cells coexpressing CD5, CD19 þ CD5 þ of disease, when normal B cells are absent, and conversely may k:l ratio of 0.6–5.7:1 and o21% of CD19 þ CD5 þ lacking surface not detect residual disease, when there is an excess of normal immunoglobulin). Of these cases, 9/10 were in CR and one had B cells. However, we again confirm that CD19/CD5/k/l may be of morphological evidence of disease (nodular PR). Six of the 27 value in identifying cases that do not require extensive analysis for cases had CD19/CD5/k/l parameters above the thresholds detection of residual disease. indicative of the presence of residual disease but below levels with 100% PVV, of which 3/6 were in CR (1 CRi) and 3/6 were in PR. CD19/CD5/k/l parameters were over the thresholds indicative of Development of a six-CLR assay: identifying appropriate marker 100% PVV for detection of residual disease in 11/27 cases (CD19 þ combinations 48.9% of total leukocytes, CD19 þ k:l ratio o0.0.4:1 or 461:1, CD19/CD5/k/l analysis may provide sufficient information in and in cases with sufficient CD5 þ B cells for enumeration 482% some cases, but in a large proportion of samples a full MRD of B cells coexpressing CD5, CD19 þ CD5 þ k:l ratio of o0.05:1 or analysis will still be required. We next investigated the possibility 432:1, or 454% of CD19 þ CD5 þ lacking surface immunoglobu- of using six-CLR combinations to minimise the number and lin). All 11 of these cases had evidence of residual disease by flow complexity of the component tests in the full MRD assays. cytometry but only 9/11 had morphological evidence of disease. To identify optimal recombination of the markers present in the

& 2013 Macmillan Publishers Limited Leukemia (2013) 142 – 149 ERIC update on CLL MRD AC Rawstron et al 146 four-CLR assay, we first reviewed data on the expression of the key increasing to 100–300 units for reagent sets prepared between 1 markers in a series of 270 CLL cases (147 peripheral blood, 116 and 28 days before sample preparation. This change was relatively bone marrow and 7 lymph node biopsy samples) to determine minor compared with the background fluorescence seen on other whether there were any redundant combinations. Hierarchical cell populations (for example, monocyte non-specific signal of cluster analysis was performed with the aim of identifying markers 300–500) and there was no apparent loss of specific PE-Cy7 or that clustered separately within CLL and may therefore be useful APC-H7 signal on B cells. As a result of these differences, a reagent to combine together, but aside from CD20 and CD79b expression set was prepared comprising two mixtures of the FITC, PE, PerCP- which showed a degree of correlation, all the other markers of Cy5.5 and APC reagents with CD19 PE-Cy7 and CD20 APC-H7 as interest showed independent variation (data not shown). separate reagents to be added at the time of sample preparation. We next determined whether CD19 þ CD3 þ contaminating events, which cause difficulties in the four-CLR analysis because Dilution study analysis they have a similar expression profile to CLL cells, could be To identify the optimal approach for analysis of both four-CLR and excluded without requiring the CD3 marker with certain six-CLR data, serial dilutions of peripheral blood leuocytes from six combinations of antibodies. This was achieved by comparing patients with CLL into normal leukocytes were prepared, the number of events classified as CLL by operators in three providing samples with a known level of CLL cells representing centres analysing files generated by dilution of peripheral blood a low percentage of total leukocytes (between 0.005 and 1.1%). leuocytes from six patients with CLL diluted into normal Although 42 samples were prepared, it was difficult to obtain leukocytes. The files contained low numbers ( 150) of CLL o sufficient events in each of the five four-CLR and both six-CLR data events stained with a combination of CD19/CD5/CD3 with CD20/ files, and only nine samples were suitable for further analysis. Flow CD79b/CD38, CD81/CD22/CD43 or CD81/CD79b/CD43. Results cytometry data files were distributed between participating from analysis of files with the CD3 data electronically removed centres and the files were analysed without knowing the were compared against unmanipulated files that were identical percentage of CLL cells in each file and the results returned for apart from the inclusion of CD3 data. The results demonstrate that central analysis. there is a small though significant reduction in the number of We investigated whether using a minimum number of events events classified as CLL when CD3 data is included in the gating required to define a population of CLL cells (no minimum, 20 strategy for a tube containing CD19/CD5/CD20/CD79b/CD38 events, or 50 events) impacted on the coefficient of variation for (n ¼ 18, paired t-test P ¼ 0.019, regression slope for analysis of each dilution and on the percentage error. As all participants data with CD3 vs data without CD3 of 1.13). More interestingly, the analysed the same FCS files, the CV obtained is a measure of the inclusion of CD3 becomes relevant particularly for samples with inter-observed variation and does not reflect the variability that residual disease near the limit of detection using this combination, might be introduced by repeatedly staining the same aliquot. where it equates to approximately 10–15% reduction in the As expected, there was a good correlation between four-CLR number of events classified as CLL. In contrast, there was no and six-CLR panels in dilution studies. There was good linearity difference for analyses containing CD19/CD5/CD81/CD22/CD43 or even below 0.01% (10 À 4) but the percentage error (defined as the CD19/CD5/CD81/CD79b/CD43, indicating that CD3 should be difference between observed number of CLL cell events and included in combinations that do not contain CD81/CD43. actual number of CLL cells events as a percentage of actual CLL cell events) on results was on average greater than 15% Development of a six-CLR assay: optimising fluorochromes for uncorrected data in samples with CLL cells levels between 0.005–0.05%. Assessment of different approaches to control We next tested a different antibody/fluorochrome combinations for contamination were assessed by correcting for (that is, to determine optimal signal:noise ratio and stability of reagent subtracting) the numbers of CD19-gated events binding CD3, or sets. CD19 and CD5 were included in all combinations with CD79b, by using CD3 as a threshold to determine the limit of detection CD43, CD20, CD38 tested in different formats in 13 peripheral (that is, only classifying a sample as having detectable residual blood and 9 bone marrow samples (data not shown). It was disease if the number of CLL events is greater than the number of decided to use CD20 APC-H7 in all combinations for three reasons CD19 þ CD3 þ events). The results are shown in Table 3. The most that are as follows: (i) CD20 as a single marker provides the most reproducible results with the lowest difference from expected powerful separation of CLL cells from normal B cells;24 values for both four-CLR and six-CLR analyses were obtained when (ii) combinations that included CD20 did not correlate worse to a minimum of 50 events were utilised to define a CLL population, RQ-PCR than the combination without CD20 in patients treated or when the number of contaminating CD19 þ CD3 þ events was with rituximab-containing regimens;17 and (iii) most other used to define a limit of detection for that sample with a minimum antibodies available at the time for testing provided a poor of 20 events utilised to define a population. signal:noise ratio when conjugated to this fluorochrome. The remaining antibodies in the combinations were chosen to allow the direct or indirect exclusion of contaminating T cells in both Comparison of four-CLR and six-CLR assays in clinical samples tubes. As a result, CD19, CD5, CD20 on PE-Cy7, PerCP-Cy5.5 and A series of 67 samples (54 peripheral blood and 13 bone marrow) APC-H7, respectively, were combined with CD3 FITC, CD38 PE and were analysed using the harmonised four-CLR MRD assay and the CD79b APC in the first combination and CD81 FITC, CD22 PE and six-CLR assay in parallel, with the aim of testing the reagent set in CD43 APC in the second combination (see Table 1). samples with a low proportion of CLL cells, preferably less than 0.1%. FITC, PE, and APC reagents are stable in combination, and An example of the six-CLR analysis is shown in Figure 3. The median PerCP-Cy5.5 stability is not an apparent issue because both PerCP CLL cell percentage reported using the four-CLR assay was 0.008%, and Cy5.5 emissions are measured by the same detector. and with the six-CLR assay 0.006%. As in the dilution study analysis, However, the stability of tandem conjugates PE-Cy7 and the use of a 50 event minimum or the number of CD19 þ CD3 þ particularly APC-H7(ref. 25) is less evident and so this was tested events as a threshold for the limit of detection resulted in the best over a 28-day period. When all the reagents were combined there concordance between four-CLR and six-CLR approaches. Overall was a slight increase in the non-specific APC-H7 signal detected there was greater than 98.4% concordance, defined as the detection on APC þ cells. This was only detectable in the second of MRD at either the 0.01% (10 À 4) level or higher, or below 0.01% combination on T cells with strong CD43-APC binding, which (10 À 4) in both four-CLR and six-CLR assays. showed a median CD20 APC-H7 fluorescence intensity of 10–20 Ideally 500 000 events would be acquired in each test, but units for reagents sets prepared at the time of sample preparation because of the difficulties in obtaining sufficient sample to acquire

Leukemia (2013) 142 – 149 & 2013 Macmillan Publishers Limited ERIC update on CLL MRD AC Rawstron et al 147 Table 3. Assessment of different approaches to reducing error and variation between the observed and expected CLL cell levels in the dilution tests

Minimum events to deﬁne Adjustment for 4-CLR average 4-CLR average 6-CLR average 6-CLR average a CLL population contaminating CD3 þ events % error (range) % CV (range) % error (range) % CV (range)

50 Events No correction 11 (1-18) 8 (0–18) 13 (0–45) 10 (4–23) 20 Events No correction 25 (1–160) 15 (0–40) 15 (0–45) 18 (4–43) No minimum No correction 17 (1–160) 13 (0–43) 17 (0–49) 20 (4–43) No minimum Subtract CD19 þ CD3 þ 33 (4–195) 16 (0–68) 33 (1–150) 17 (4–55) 20 Events Exclude if CLLoCD19 þ 3 þ events 10 (1–58) 9 (0–26) 14 (0–45) 10 (4–23 Abbreviations: CLL, chronic lymphocytic leukaemia; CLR, colour. enough events for seven different tubes in post-treatment From these results it could be argued that in such cases the use samples, the reproducibility was limited by a lack of events in of the full MRD panel would add significant cost and labour time some cases. In cases with at least 200 000 events in each data file, to obtain a quantitative result, when a qualitative evaluation is (n ¼ 59) there was 100% concordance between four-CLR and six- sufficient in some settings. Therefore, the entire full MRD panel CLR assays in samples. The results are shown in Figure 4. In seven could then be more appropriately restricted to those cases which samples (three from Paris where the median number of events do not meet the above criteria. Using a CD19/CD5/k/l screening captured was over 1 million, and four from Barcelona where at approach provided a positive result in 47% of all cases in this least 500 000 events were captured and the contamination rate series and therefore has the potential to nearly halve the amount was consistently below 20 events), it was possible to compare the of full MRD analyses required in patients receiving standard detection of CLL phenotype cells in the 0.001–0.01% (10 À 5–10 À 4) chemo-immunotherapy protocols. However, CD19/CD5/k/l ana- range by both four-CLR and six-CLR assays. Although the number lysis was most effective in samples with higher levels of CLL, of samples is small there was good correlation (Pearson R ¼ 0.832) predicting the presence of residual disease in 72% (166/232) of with no obvious bias (Bland–Altman limits of agreement cases in the validation set with more than 1% CLL but in only 37% 0.0011±0.0028). (19/51) of cases with p1% residual CLL. A further limitation of the approach is that it is qualitative only, whereas recent observations clearly demonstrated the added value of MRD quantification DISCUSSION within the range of MRD positivity.5,7 The use of MRD analysis in CLL is becoming increasingly common for Further time savings may be made by taking advantage of assessment of response, identifying the kinetics of tumour depletion the newer flow cytometry equipment that offers more parameters and relapse, and it is proposed as a potential surrogate end point in and the potential to reduce the number of individual assessments clinical trials.5 The harmonised four-CLR assay developed some years required for each case. We have designed a six-CLR assay that ago is highly reproducible and has been validated in multi-centre incorporates the same markers as the harmonised four-CLR assay randomised clinical trials.5,26 However, in the four-CLR version the but halves the number of tubes required for quantification of assay can be time consuming, particularly in the analysis of minimal the proportion of B-cell which have a CLL phenotype. This can levels of disease. In addition the full assay may be unnecessary in confer modest benefits in terms of the amount of time and cost samples with higher levels of residual disease. required to set-up but provides a 50% reduction in the time We have investigated the potential for using the basic CD19/ required to acquire information and provides a major advantages CD5 and clonality assessment to simplify MRD analysis. It was not in the time and complexity of analysis, as there are fewer files to possible to identify thresholds with a high negative predictive analyse in parallel for each case. Also, because the sample is value for the presence of MRD, however, CD19/CD5/k/l analysis distributed between fewer component tests, the use of a six-CLR could be used to identify samples with a very high probability of analysis permits acquisition of 500 000 events or more even in containing residual disease. In cases with more than 82% of B cells post-treatment samples, which are often poorly cellular. coexpressing CD5, and with a highly skewed k:l ratio (o0.05:1 or The results in this study show that when larger numbers of 432:1) or very weak surface immunoglobulin on the CD5 þ B cells events are analysed, both four-CLR and six-CLR assays can detect (454% sIg À ), it was possible to demonstrate in a validation set of CLL cells in the 0.001–0.01% (10 À 5–10 À 4) range with good 392 cases that there was 100% PVV for the presence of residual concordance. This places the detection limit of residual disease by disease at the 0.01% (10 À 4) level or higher. Very large deviations flow cytometry in the same level as RQ-ASO IGH-PCR and from normal are required for confidently predicting the presence high-throughput sequencing.17,27 The ability to reproducibly of residual disease for several reasons. First, regenerating B cells detect CLL cells at this level would need to be tested in a multi- have high levels of CD5 expression and even the 82% threshold centre setting, and until such studies are reported it may be may be too low after transplantation (SB and NV unpublished preferable to report results of residual disease under the 0.01% observation and Bomberger et al.23). Second, highly skewed k:l (10 À 4) threshold as being below the known quantitative range. A ratios are frequently seen in the early stages after treatment potential limitation of this study is that the inter-laboratory because the numbers of B cells expressing surface variation was studied using flow cytometry data files rather than immunoglobulin are low and therefore oligoclonal-reactive primary samples. However, by analysing the same sample at the expansions can have a major impact on the k:l ratio. Lack of same time in multiple laboratories we have previously surface immunoglobulin expression, (which in this study was demonstrated that differences in laboratory preparation do not defined using T cells as an internal control) on the CD5 þ B cells add significantly to the level of inter-laboratory variation may be most reliable as normal CD5 þ B cells should have generated by differences in the approach to data analysis.8 The moderate surface immunoglobulin levels, but the levels of close correlation between laboratories using the four-CLR assay8 immunoglobulin on CLL cells vary significantly between patients. coupled with previous reports of detecting residual disease in the It is therefore recommended that an excess of CD5 þ B cells 0.001–0.01% (10 À 5–10 À 4) range under theoretical conditions28 should only be considered as an indicative of the presence of indicates that quantitative assessment of residual disease below residual disease if combined with a heavily skewed k:l ratio and/ the 0.01% (10 À 4 CLL cell in 10 000 leukocytes) threshold is or a lack of surface immunoglobulin. achievable.

& 2013 Macmillan Publishers Limited Leukemia (2013) 142 – 149 ERIC update on CLL MRD AC Rawstron et al 148 abc 250 105 105

(x 1,000) 200 104 104 150

103 103 100 Side Scatter CD19 PE-Cy7 CD19 PE-Cy7

0 50 0

-974 -974 M M 50 100 150 200 250 50 100 150 200 250 -102 0102 103 104 105 Side Scatter(x 1,000) Forward Scatter(x 1,000) -472 CD3 FITC def 105 105 105

4 4 10 10 104

3 10 103 103 CD38 PE CD5 PerCP-Cy5.5 102 102 CD5 PerCP-CY5.5 0 0 0 -102 -102 -472 -472 -472 MMM-102 0102 103 104 105 -102 0102 103 104 105 -102 0102 103 104 105 -472 -472 -472 CD79b APC CD20 APC-H7 CD20 APC-H7 ghi 105 105 105

4 10 104 104

103 103 103 CD43 APC

2 CD5 PerCP-Cy5.5 10 CD5 PerCP-Cy5.5 0 0 0 -102 -472 -974 -974 MMM -102 0102 103 104 105 -102 0102 103 104 105 -102 0102 103 104 105 -472 -472 -472 CD81 FITC CD22 PE CD20 APC-H7

CD3/CD38/CD79b statistics CD81/CD22/CD43 statistics Population #Events %Total Population #Events %Total

All Events 500,000 100.000 All Events 500,000 100.000 Total B-cells 7,997 1.599 Total B-cells 4,460 1.692 CLL cells 565 0.113 CLL cells 616 0.123 B-cells excluding CLL 7,432 1.486 B-cells excluding CLL 7,844 1.569 CD19+CD3+ events 34 0.007

Figure 3. Example of the approach to analysis of samples for the presence of minimal residual disease using two six-CLR assays. A B-cell gate is defined by the combination of two regions according to CD19 expression and light scatter characteristics (a, b). In the first tube, the number of CD3 þ events within the B-cell gate are calculated (c) and the numbers of CLL phenotype events (shown in dark red) are identified by the combination of three regions according to CD5, CD20, CD79b and CD38 expression in (d), (e) and (f). In the second tube, the same B-cell gating approach is applied, and the numbers of CLL phenotype events (shown in dark red) are identified by the combination of three regions according to CD5, CD20, CD22, CD43 and CD81 expression in (g), (h) and (i).

In summary, we present an approach that can signiﬁcantly cases that eventually need to go through the full MRD analysis, reduce the amount of time and sample required for the with additional savings in terms of labour, time and reagents. enumeration of residual disease in CLL, which can be reliably set-up and run in most diagnostic laboratories with less burden in terms of cost and labour as compared with the current four-CLR CONFLICT OF INTEREST ﬂow-cytometric protocol. The use of a simple clonality assessment The antibodies used for MRD analysis in this study were provided by BD Biosciences. upfront can also help reducing by around 50% the number of AR has received royalties from BD for an unrelated product (IntraSure). Other

Leukemia (2013) 142 – 149 & 2013 Macmillan Publishers Limited ERIC update on CLL MRD AC Rawstron et al 149 9 Lenormand B, Bizet M, Fruchart C, Tilly H, Daliphard S, Thouret F et al. Residual disease in B-cell chronic lymphocytic leukemia patients and prognostic value. Leukemia 1994; 8: 1019–1026. 10 Vuillier F, Claisse JF, Vandenvelde C, Travade P, Magnac C, Chevret S et al. Evaluation of residual disease in B-cell chronic lymphocytic leukemia patients in clinical and bone-marrow remission using CD5-CD19 markers and PCR study of gene rearrangements. Leuk Lymphoma 1992; 7: 195–204. 11 Robertson LE, Huh YO, Butler JJ, Pugh WC, Hirsch-Ginsberg C, Stass S et al. Response assessment in chronic lymphocytic leukemia after fludarabine plus prednisone: clinical, pathologic, immunophenotypic, and molecular analysis. Blood 1992; 80: 29–36. 12 Tam CS, O’Brien S, Wierda W, Kantarjian H, Wen S, Do K-A et al. Long-term results of the fludarabine, cyclophosphamide, and rituximab regimen as initial therapy of chronic lymphocytic leukemia. Blood 2008; 112: 975–980. 13 Hillmen P, Skotnicki AB, Robak T, Jaksic B, Dmoszynska A, Wu J et al. Alemtu- zumab compared with chlorambucil as first-line therapy for chronic lymphocytic leukemia. J Clin Oncol 2007; 25: 5616–5623. 14 Hallek M, Cheson BD, Catovsky D, Caligaris-Cappio F, Dighiero G, Do¨hner H et al. Guidelines for the diagnosis and treatment of chronic lymphocytic leukemia: Figure 4. Comparison of four-CLR and six-CLR analysis for the a report from the International Workshop on Chronic Lymphocytic Leukemia detection of minimal residual disease in samples from patients with updating the National Cancer Institute-Working Group 1996 guidelines. Blood CLL after treatment (n ¼ 59, of which CLL cells represented p0.001% 2008; 111: 5446–5456. of leukocytes in 25 cases). 15 Catovsky D, Richards S, Matutes E, Oscier D, Dyer MJS, Bezares RF et al. Assess- ment of fludarabine plus cyclophosphamide for patients with chronic lympho- companies were contacted at the time of publication of the international cytic leukaemia (the LRF CLL4 Trial): a randomised controlled trial. Lancet 2007; standardised approach8 but did not express an interest in providing antibodies for 370: 230–239. development of MRD detection. BD Biosciences had no scientific input into the 16 Kennedy B, Rawstron A, Carter C, Ryan M, Speed K, Lucas G et al. Campath-1H and project and there were no significant conflicts of interest that could be perceived to fludarabine in combination are highly active in refractory chronic lymphocytic bias the work. The remaining authors declare no conflict of interest. leukemia. Blood 2002; 99: 2245–2247. 17 Bo¨ttcher S, Stilgenbauer S, Busch R, Bru¨ggemann M, Raff T, Pott C et al. Standardized MRD flow and ASO IGH RQ-PCR for MRD quantification in CLL patients after rituximab-containing immunochemotherapy: a comparative ACKNOWLEDGEMENTS analysis. Leukemia 2009; 23: 2007–2017. We are grateful to BD Biosciences, particularly Frans Nauwelaers and Lucia Testolin, 18 Bosch F, Abrisqueta P, Villamor N, Terol MJ, Gonza´lez-Barca E, Ferra C et al. for provision of antibodies used in this study. PG supported by Cariplo Foundation Rituximab, fludarabine, cyclophosphamide, and mitoxantrone: a new, highly (Milan, Italy), Program Molecular Clinical Oncology-5 per mille number 9965 and active chemoimmunotherapy regimen for chronic lymphocytic leukemia. J Clin Investigator Grant, Associazione Italiana per la Ricerca sul Cancro (AIRC Milano, Italy), Oncol 2009; 27: 4578–4584. Progetti di Ricerca di Interesse Nazionale (PRIN - Ministry of education, University and 19 Bosch F, Ferrer A, Villamor N, Gonza´lez M, Briones J, Gonza´lez-Barca E et al. Flu- Research - MIUR, Rome, Italy) and Ricerca Finalizzata 2010 (Ministry of Health, Rome, darabine, cyclophosphamide, and mitoxantrone as initial therapy of chronic Italy); AR supported by LLR (www.leukaemialymphomaresearch.org.uk). We are lymphocytic leukemia: high response rate and disease eradication. Clin Cancer Res grateful to Linda Falck, Elke Harbst, Jamileh Hanani, Lada Henseleit, Heike Hinrichs 2008; 14: 155–161. and Abdelmalek Dahmani for excellent technical support. 20 Marti GE, Rawstron AC, Ghia P, Hillmen P, Houlston RS, Kay N et al. Diagnostic criteria for monoclonal B-cell lymphocytosis. Br J Haematol 2005; 130: 325–332. 21 Bo¨ttcher S, Ritgen M, Buske S, Gesk S, Klapper W, Hoster E et al. Minimal residual REFERENCES disease detection in mantle cell lymphoma: methods and significance of 1 Gribben JG, O’Brien S. Update on therapy of chronic lymphocytic leukemia. J Clin four-color flow cytometry compared to consensus IGH-polymerase chain reaction Oncol 2011; 29: 544–550. at initial staging and for follow-up examinations. Haematologica 2008; 93: 2 Abrisqueta P, Pereira A, Rozman C, Aymerich M, Gine´ E, Moreno C et al. Improving 551–559. survival in patients with chronic lymphocytic leukemia (1980–2008): the Hospital 22 Perez-Andres M, Paiva B, Nieto WG, Caraux A, Schmitz A, Almeida J et al. Human Clinic of Barcelona experience. Blood 2009; 114: 2044–2050. peripheral blood B-cell compartments: a crossroad in B-cell traffic. Cytometry B 3 Moreton P, Kennedy B, Lucas G, Leach M, Rassam SMB, Haynes A et al. Eradication Clin Cytom 2010; 78(Suppl 1): S47–S60. of minimal residual disease in B-cell chronic lymphocytic leukemia after 23 Bomberger C, Singh-Jairam M, Rodey G, Guerriero A, Yeager AM, Fleming WH alemtuzumab therapy is associated with prolonged survival. J Clin Oncol 2005; 23: et al. Lymphoid reconstitution after autologous PBSC transplantation 2971–2979. with FACS-sorted CD34 þ hematopoietic progenitors. Blood 1998; 91: 4 Hallek M, Fischer K, Fingerle-Rowson G, Fink AM, Busch R, Mayer J et al. Addition 2588–2600. of rituximab to fludarabine and cyclophosphamide in patients with chronic 24 Rawstron AC, Kennedy B, Evans PA, Davies FE, Richards SJ, Haynes AP et al. lymphocytic leukaemia: a randomised, open-label, phase 3 trial. Lancet 2010; 376: Quantitation of minimal disease levels in chronic lymphocytic leukemia using a 1164–1174. sensitive flow cytometric assay improves the prediction of outcome and can be 5Bo¨ttcher S, Ritgen M, Fischer K, Stilgenbauer S, Busch RM, Fingerle-Rowson G et al. used to optimize therapy. Blood 2001; 98: 29–35. Minimal residual disease quantification is an independent predictor of progres- 25 Le Roy C, Varin-Blank N, Ajchenbaum-Cymbalista F, Letestu R. Flow cytometry sion-free and overall survival in chronic lymphocytic leukemia: a multivariate APC-tandem dyes are degraded through a cell-dependent mechanism. Cytometry analysis from the randomized GCLLSG CLL8 trial. J Clin Oncol 2012; doi:10.1200/ A 2009; 75: 882–890. JCO.2011.36.9348. 26 Hillmen P, Cohen DR, Cocks K, Pettitt A, Sayala HA, Rawstron AC et al. A rando- 6Bo¨ttcher S, Ritgen M, Pott C, Bru¨ggemann M, Raff T, Stilgenbauer S et al. Com- mized phase II trial of fludarabine, cyclophosphamide and mitoxantrone (FCM) parative analysis of minimal residual disease detection using four-color flow with or without rituximab in previously treated chronic lymphocytic leukaemia. cytometry, consensus IgH-PCR, and quantitative IgH PCR in CLL after allogeneic Br J Haematol 2011; 152: 570–578. and autologous stem cell transplantation. Leukemia 2004; 18: 1637–1645. 27 Logan AC, Gao H, Wang C, Sahaf B, Jones CD, Marshall EL et al. High-throughput 7 Moreno C, Villamor N, Colomer D, Esteve J, Gine´ E, Muntanõla A et al. Clinical VDJ sequencing for quantification of minimal residual disease in chronic significance of minimal residual disease, as assessed by different techniques, after lymphocytic leukemia and immune reconstitution assessment. Proc Natl Acad Sci stem cell transplantation for chronic lymphocytic leukemia. Blood 2006; 107: USA 2011; 108: 21194–21199. 4563–4569. 28 Pedreira CE, Costa ES, Almeida J, Fernandez C, Quijano S, Flores J et al. A 8 Rawstron AC, Villamor N, Ritgen M, Bo¨ ttcher S, Ghia P, Zehnder JL et al. Inter- probabilistic approach for the evaluation of minimal residual disease by national standardized approach for flow cytometric residual disease monitoring in multiparameter flow cytometry in leukemic B-cell chronic lymphoproliferative chronic lymphocytic leukaemia. Leukemia 2007; 21: 956–964. disorders. Cytometry A 2008; 73A: 1141–1150.

& 2013 Macmillan Publishers Limited Leukemia (2013) 142 – 149