Measurement of apoptosis and proliferation of bone marrow plasma cells in patients with plasma cell proliferative disorders

Authors


Dr T. E. Witzig, 920E Hilton Bldg, Mayo Clinic, Rochester, MN 55905, U.S.A.

Abstract

The proliferative rate of malignant plasma cells, as measured by the plasma cell labelling index (PCLI), is an important prognostic factor in multiple myeloma (MM); however, the PCLI alone is probably inadequate to describe tumour growth because it ignores the idea that myeloma cells may have a reduced rate of apoptosis. The aims of this study were to develop a flow cytometric method to measure the apoptosis index of fresh marrow plasma cells and develop a plasma cell growth index (PCGI) that related both proliferation and apoptosis to disease activity. Marrow aspirates were obtained from 91 patients with plasma cell disorders and the plasma cells in apoptosis were identified by either 7-amino actinomycin-D (7-AAD) or annexin V–FITC three-colour flow cytometry. The median plasma cell apoptotic index (PCAI) for patients with monoclonal gammopathy of undetermined significance (MGUS), smouldering or indolent myeloma (SMM/IMM), and new multiple myeloma (MM) was 5.2, 3.4 and 2.4, respectively (P=0.03, MGUS v MM). The median PCLI for these same patient groups was 0.0, 0.2 and 0.6, respectively (P<0.001, MGUS v MM). The paired PCLI and PCAI for each sample were used to derive the PCGI = 2 + [PCLI − (0.1)(PCAI)]. The median PCGI for patients with inactive disease (MGUS, SMM/IMM or amyloidosis) was 1.8 compared to 2.4 for those with active disease (new or relapsed MM) (P < 0.001). These results suggest that a decrease in the PCAI may be a factor in the progression from MGUS to SMM to overt MM.

In normal human bone marrow, B cells activated by antigen in the lymph nodes are stimulated to differentiate into plasma cells and secrete immunoglobulin (Ig) under the influence of interleukin-6 (IL-6). Under normal conditions these plasma cells survive for a variable period of time, but do not proliferate. The plasma cell proliferative disorders consist of monoclonal gammopathy of undetermined significance (MGUS), smouldering or indolent multiple myeloma (SMM/IMM), AL-amyloidosis (AL), Waldenström's macroglobulinaemia, and multiple myeloma (MM). All of these disorders are characterized by increased numbers of monoclonal plasma cells in the marrow that have the capacity to proliferate as measured by the plasma cell labelling index (PCLI) or flow cytometric S-phase determination (Greipp et al, 1985; San Miguel et al, 1995). In patients with MGUS or AL the plasma cells typically have a low proliferative rate; however, in many patients with MM the malignant plasma cells proliferate more rapidly, leading to the marrow failure and lytic bone lesions that are the hallmarks of the disease (Boccadoro et al, 1984; Durie et al, 1980; Ffrench et al, 1995; Greipp et al, 1988, 1993; Joshua et al, 1988; San Miguel et al, 1995; Witzig et al, 1996a).

The net growth of any tumour cell population is a function of the rates of cell proliferation and cell death. If either or both processes are altered the clone may expand. In MM the overall survival of patients varies substantially and is influenced by the proliferative characteristics of the marrow plasma cell clone. Patients with monoclonal plasma cells with a high % S-phase typically have a short survival; those with a low PCLI usually have a longer survival, but even with a very low PCLI the MM may progress (Greipp et al, 1993; San Miguel et al, 1995). This suggests that another important defect in these cells is a reduced rate of apoptosis. Indeed, IL-6 and insulin-like growth factors (IGF) I and II, both growth factors for myeloma cells (Georgii-Hemming et al, 1996; Jelinek et al, 1997), have also been demonstrated to protect cells from entering apoptosis (Lichtenstein et al, 1995; Xu et al, 1997).

Although there has been intense investigation into the mechanisms of apoptosis using human malignant cell lines, little is known about the apoptosis rate of malignant plasma cells from patients with plasma cell proliferative diseases. Novel methods have been developed to estimate the percentage of cells undergoing apoptosis using flow cytometry. Philpott et al (1996) used 7-amino actinomycin D (7-AAD) and found that the cell membranes of cells in apoptosis were permeable to 7-AAD and could be distinguished from cells that were alive and those that were dead. Another method uses annexin V (annexin) conjugated to fluorescein isothocyanate (annexin-FITC) to detect cells entering apoptosis. Annexin is a protein that has a high affinity for negatively charged phospholipids such as phosphatidylserine (Andree et al, 1990). As cells begin to enter apoptosis the phosphatidyl moieties on the inner leaflet of the cell membrane are externalized and are thus detected by annexin-FITC (Koopman et al, 1994). Both the 7-AAD and annexin reagents can be combined with other cell markers allowing specific cellular populations to be analysed for apoptosis.

Monoclonal plasma cells can be identified by use of three-colour flow cytometry (Rawstron et al, 1997; Schneider et al, 1997; Witzig et al, 1996b). The first aim of this study was to develop a flow cytometric method that would allow both identification of monoclonal plasma cells in fresh marrow samples and estimate the percentage of these cells that were undergoing apoptosis. We hypothesized that plasma cells from patients with MM would tend to have higher proliferative and lower apoptosis rates than cells from patients with MGUS, AL or SMM/IMM. Therefore a secondary aim of this study was to explore whether there were differences in the plasma cell apoptosis index (PCAI) between patients with various plasma cell proliferative disorders. Lastly, since a combination of both the labelling and apoptosis indices may better predict net tumour growth, we developed a plasma cell growth index that utilized both measurements.

MATERIALS AND METHODS

Patient selection

Patients were eligible for this study if they had the diagnosis of a plasma cell proliferative disorder and were undergoing a bone marrow aspirate and biopsy with a plasma cell labelling index for clinical purposes. All samples were collected and processed during 1996 and 1997. Patient records were reviewed to obtain the diagnosis and information regarding recent treatment with chemotherapy. This study was approved by the Institutional Review Board of the Mayo Clinic/Foundation.

Patients were classified as MGUS if they had a serum or urine monoclonal (M) protein, a marrow with <10% monoclonal plasma cells, normal plasma haemogobin, no lytic bone lesions, were asymptomatic, and were observed without chemotherapy or radiotherapy. Patients with SMM met the same criteria as described above for MGUS but had >10% monoclonal plasma cells in the bone marrow. Indolent MM (IMM) was diagnosed in patients who met the criteria for MM with >10% monoclonal plasma cells in the marrow, and had either anaemia or lytic bone lesions, yet were asymptomatic and did not require institution of chemotherapy. Patients with new untreated MM had a marrow with >10% plasma cells, and bone lesions or other symptoms of active disease such as anaemia, renal insufficiency, or hypercalcaemia requiring chemotherapy. Relapsed MM patients had active disease that was not responding to current chemotherapy or had recurred following cessation of effective therapy. Normal marrow controls were patients donating marrow for allogeneic bone marrow transplant. Studies of the reproducability of the PCAI were performed using the MY5 and 8226 human myeloma cell lines.

Measurement of cell proliferation: the plasma cell labelling index (PCLI)

The PCLI was performed on fresh marrow aspirate by use of a bromodeoxyuridine immunofluorescence procedure that identifies plasma cells by monoclonal cytoplasmic light chain expression and determines the percentage of cells in S-phase. This procedure has been previously described on both marrow and blood plasma cells (Greipp et al, 1985). The percentage of marrow plasma cells recorded was from the PCLI measurement.

Measurement of apoptosis: the plasma cell apoptosis index (PCAI)

The percentage of plasma cells in apoptosis, the PCAI, was performed on 0.5 ml of waste marrow aspirate obtained at the same time as the PCLI sample. The assay was completed within 6 h of the marrow procedure without knowledge of the patient's clinical status. The aspirate was washed with 15 ml of cold phosphate-buffered saline (PBS) and centrifuged at 300 g for 5 min. The supernatant was decanted and 15 ml of ACK red blood cell lysing solution added for 5 min and then centrifuged. The cell pellet was washed twice with PBS and resuspended in PBS 3% BSA to give a final concentration of 1 × 106 cells/100 μl.

For the 7-AAD method, 100 μl of cell suspension was added to two tubes and stained with monoclonal antibodies. Plasma cells were identified by use of 5 μl anti-CD38 conjugated to phycoerythrin (CD38-PE, Becton Dickinson, Mountain View, Calif.) and either 10 μl anti-syndecan-1 conjugated to fluorescein isothiocyanate (BB4-FITC, Serotec, Raleigh, N.C.) or 5 μl CD45 fluorescein isothiocyanate (CD45-FITC, Becton Dickinson). The monoclonal antibodies were incubated for 20 min at 4°C; then washed in PBS, spun down, resuspended in 500 μl PBS. 7-AAD (Calbiochem, San Diego, Calif.) at a final concentration of 20 μg/ml was added to one of these samples. 7AAD was also added to third tube containing cells alone (no antibody) that was used to set compensation on the flow cytometer. The tubes were washed once in PBS, resuspended in 500 μl of PBS and analysed within 30 min on a FacScan flow cytometer (Becton Dickinson).

For the annexin method, 100 μl of the cell suspension, together with 100 μl of the Hepes-Ca++ buffer were added to each of two tubes; 100 μl of cell suspension only was added to a third tube and stained with 5 μl anti-CD38-PE alone to set instrument compensation. Annexin V conjugated to FITC (Caltag, Burlingame, Calif.) was added to cells in one tube and was also used for instrument compensation. In the remaining tube the plasma cells in the suspension were identified by staining with 5 μl of anti-CD38 PE, 5 μl CD45 PERCP (Becton Dickinson), and 5 μl of the annexin FITC. Samples were incubated for 20 min at 4°C, washed once with 50% Hepes-Ca++ buffer/50% PBS, spun down and resuspended in 500 μl of Hepes-Ca++ buffer for analysis by flow cytometry within 20 min. Propidium iodide was not used in this three-colour flow cytometric assay because our aim was to specifically determine the apoptosis rate of malignant plasma cells which required the use of CD38 PE and CD45 PERCP. Therefore, with the annexin method we were unable to distinguish dead from apoptotic annexin-positive cells. The gate used to determine annexin-positive cells was set using the non-plasma marrow fraction as an internal control.

The Cell Quest software program (Becton Dickinson) was used to analyse the results. Plasma cells are usually easily identified by the unique dual expression of bright CD38 and BB4 or bright CD38 and CD45−/dim+ on the cell surface membrane (Rawstron et al, 1997; Witzig et al, 1996b). These cells were gated on and the 7-AAD or annexin expression of these cells determined (Fig 1). For the 7-AAD method, cells that were 7-AAD negative were alive and effectively excluded the dye; cells that were bright 7-AAD positive were dead and readily permeable to 7-AAD; cells undergoing apoptosis had 7-AAD staining intermediate between these two gates. The gate between the 7-AAD-negative and apoptotic cells was set using the control tube containing CD38 and CD45 (or BB4) without 7-AAD. The gate between apoptotic and dead cells was set below the cluster of low scatter, 7-AAD bright cells consistently seen in each sample (Fig 1D). The percentages of live, apoptotic and dead cells were calculated by the program. For annexin, the cells that stained positive were the cells undergoing apoptosis or which were dead.

Figure 1.

. Determination of the percentage of plasma cells in apoptosis by 7-AAD or annexin-V methods on the same patient marrow sample. In (A) the expression of CD45 and CD38 indicates 28% plasma cells in the marrow sample; in (B) 6% of these gated plasma cells are found to be annexin-V positive. In (C) 28% of the cells in the marrow sample are CD38+ BB4+ plasma cells, and the 7-AAD results (D) demonstrate 2% of the gated plasma cells to be in apoptosis and 1.4% are dead.

Statistical methods

Data were analysed and the disease groups compared by use of StatView 4.1 software (Abacus Concepts Inc., Berkeley, Calif.). All 91 patients had an analysis of the PCAI by either the 7-AAD or annexin method, 43 had both techniques performed on the same marrow sample, and in 12 patients only the annexin method was used. The correlation of the 7-AAD and annexin values in the 43 patients was measured by use of the Pearson correlation coefficient. In order to include all patients in the comparisons with disease activity, we used PCAI 7AAD = 0.34 (PCAI annexin) − 0.24 determined from linear regression analysis, to obtain the 7-AAD values from the 12 cases with only annexin data.

It was already known that a high PCLI indicates progressive myeloma. It was our hypothesis that a lower PCAI would also suggest active disease; therefore the PCLI and PCAI were used along with disease activity to develop the plasma cell growth index that includes both values. The labelling, apoptosis and growth indices for the active, inactive and newly treated disease groups were compared. Global comparisons were performed with the Kruskal-Wallis test; when significant differences were found the groups were further analysed by use of the rank-sum test.

RESULTS

The 7-AAD flow cytometry procedure to measure per cent apoptosis has recently been described (Philpott et al, 1996). Their report demonstrated that the cells in the apoptosis gate by this method were in apoptosis when these cells were sorted and analysed for apoptosis by other techniques such as DNA fragmentation. We were interested in measuring the per cent apoptosis of a single cellular population, the plasma cells, in a marrow sample that contains multiple cell types. We have previously shown that malignant marrow and blood plasma cells can be identified by three-colour flow cytometry using CD38++CD45−/dim+ monoclonal cytoplasmic kappa or lambda light chain or two-colour flow cytometry with CD38++ BB4+ (Witzig et al, 1996b). We modified these techniques to enable the simultaneous identification of the plasma cells and calculation of the percentage of these cells that were in spontaneous apoptosis using 7-AAD or annexin-FITC. As depicted in Fig 1, we found that three-colour flow cytometry using either 7-AAD or annexin was able to identify marrow plasma cells and determine the percentage of these gated cells that were undergoing apoptosis or were dead.

The reproducibility of the 7-AAD assay results was tested with the MY5 and 8226 human myeloma cell lines. The cells were harvested after 60 h of growth in nutrient media, split into three aliquots and assayed for unstimulated apoptosis by the 7-AAD method. The percentage of cells in apoptosis or dead in each of the three samples was 13.3%, 14.2% and 13.2%, respectively, for 8226 and 11.7%, 11.7% and 11.1%, respectively, for the MY5 cells. The same analysis was repeated for both cell lines >10 times and similar reproducibility was obtained. Normal marrows from five allogeneic donors were obtained and studied with both the 7-AAD and annexin methods. For 7-AAD the median per cent plasma cells in apoptosis was 3.5 (mean 7.1; range 2.7–17.3); for annexin the median per cent cells in apoptosis or dead was 12 (mean 15.4; range 10.2–25.5).

These flow cytometry apoptosis assays were utilized to study fresh marrow aspirate from 91 patients with plasma cell proliferative disorders. 48 patients had inactive disease (nine MGUS, 27 SMM/IMM, 12 AL), 29 were active MM (23 new untreated MM, six relapsed MM) and 14 had been previously treated with chemotherapy. In 43 patients the PCAI was estimated by both 7-AAD and annexin methods and the annexin values tended to be higher. This finding was expected, because annexin detects earlier events in the apoptosis cascade than the 7-AAD method, and the three-colour method used in this study to detect annexin staining does not distinguish apoptotic from dead cells. When the percentage of plasma cells staining with annexin was compared with the percentage of plasma cells in apoptosis or dead by 7-AAD the correlation was 0.65 (Fig 2). In 12 patients the PCAI value was obtained only by the annexin technique; therefore these values were converted to 7-AAD values as described in Methods for the purposes of comparison with disease activity.

Figure 2.

. Graph of per cent plasma cells in apoptosis or dead by 7-AAD v annexin-V as determined on marrow samples from 43 cases. The correlation coefficient was 0.65 (P<0.01).

The median PCAI for the 91 patients with MGUS, SMM/IMM and new MM was 5.2, 3.4 and 2.4, respectively (P=0.059 by Kruskal-Wallis) (Table I, Fig 3). The difference between MGUS and new MM was significant (P=0.03); however, the differences between MGUS and SMM/IMM (P=0.19) and SMM/IMM v MM (P=0.11) were not. As found in previous studies, the PCLI results showed low values for patients with MGUS, SMM/IMM or AL and higher values for patients with new or relapsed MM. The median PCLI of the marrow plasma cells for patients with MGUS, SMM/IMM and new MM was 0.0, 0.2 and 0.6, respectively (P<0.001 by Kruskal-Wallis). The difference in PCLI in each of the two-group comparisons was significant (P<0.05).

Table 1. Table I. The % S-phase (labelling index) and percentage of marrow plasma cells in apoptosis by disease activity. Abbreviations: MGUS, monoclonal gammopathy of undetermined significance; SMM/IMM, smouldering or indolent multiple myeloma; MM, multiple myeloma; PC, plasma cell; PCLI, plasma cell labelling index.Thumbnail image of
Figure 3.

. Dot-plot display of the per cent apoptosis results by diagnosis as measured with three-colour flow cytometry. The horizontal bars in each group are median values. Values displayed as open circles were determined by the annexin-V method and converted to 7-AAD values (see text for formula).

The finding that the monoclonal plasma cells from patients with new untreated MM had a higher % S-phase and a lower apoptosis rate than plasma cells from patients with MGUS or SMM/IMM led us to explore combining these two measurements into a formula, the plasma cell growth index (PCGI). The PCGI = 2 + [PCLI − (0.1)(PCAI)]. When the derived PCGI for each patient was compared to disease activity there was a progressive increase in the PCGI when the MGUS, SMM, new MM and relapsed MM groups were compared (Table I). When the PCGI from inactive and active disease groups were compared, we found a median PCGI of 1.8 for inactive disease (MGUS, SMM/IMM and AL) compared to 2.4 for active disease (new and relapsed MM) (P<0.001) (Fig 4).

Figure 4.

. Plasma cell growth index (PCGI) values for patients with monoclonal gammopathy of undetermined significance, smouldering or indolent multiple myeloma or AL-amyloid (inactive disease) versus patients with new or relapsed multiple myeloma (active disease). The values displayed as open circles were obtained with the annexin-V method and converted to 7-AAD values. The median PCGI for the inactive group was 1.8 v 2.4 for the patients with active disease (P<0.001).

Two patients had a second marrow study after 1 year of observation without treatment. The first patient had an initial marrow sample that contained 13% monoclonal plasma cells with a PCLI of 1.2% (high), a PCAI of 9.1% (by 7-AAD) and a PCGI of 2.39. His radiographic bone survey showed a few small asymptomatic lytic lesions, the serum M-protein was 29 g/l, and the plasma haemoglobin was 15.2 g/dl. He was treated with monthly pamidronate and 1 year later he remained clinically asymptomatic. Repeat marrow showed 41% monoclonal plasma cells, the PCLI had fallen to 0.2% (low), PCAI had decreased to 2.8%, and the PCGI had decreased to 1.92. A radiographic bone survey revealed the lytic bone lesions were unchanged, the haemoglobin was 14.8 g/dl, and the serum M-protein was 30 g/l. The second patient had a serum monoclonal protein of 29 g/l, 43% monoclonal plasma cells, PCLI of 0.4% (low), a PCAI of 3.7%, and a PCGI of 2.03. The haemoglobin was 12.6 g/l and the radiographic bone survey was negative for lytic lesions. The patient was diagnosed as SMM and observed without chemotherapy. 10 months later the serum M-protein had risen to 71g/l and the marrow contained 77% plasma cells. The PCLI had increased substantially to 2.0% and the PCAI had decreased to 2.0%, producing a marked increase in the PCGI to 3.9. The plasma haemoglobin had decreased to 10 g/l and the radiographic bone survey was negative. Chemotherapy was subsequently initiated.

DISCUSSION

Previous studies of marrow plasma cells in patients with MGUS and MM have focused on measuring cell proliferation (Boccadoro et al, 1984; Greipp et al, 1993; San Miguel et al, 1995). This study describes a three-colour flow cytometry assay that can be applied to fresh marrow samples to measure the level of apoptosis of the monoclonal plasma cell population. This technique was subsequently applied to 91 patient samples and the results demonstrated that the rate of apoptosis in marrow plasma cells can vary with disease activity. There was a trend toward decreasing plasma cell apoptosis and increasing proliferative rate when the patients with MGUS v SMM/IMM v new MM were compared. This confirmed our hypothesis that increased disease activity is characterized by either an increase in malignant plasma cell proliferation, a decrease in apoptosis, or changes in both processes.

When the PCGI, derived by use of both the PCLI and the PCAI, was examined with respect to disease activity the groups with active disease had higher values; however, there was overlap between the groups (Fig 4). One possible cause of this overlap is that the disease activity of each patient was assigned using standard clinical criteria without the benefit of long-term follow-up. It will be important to re-evaluate the PCGI after long-term follow-up clarifies which of the MGUS and SMM/IMM were actually active myeloma cases compared to those patients who follow a course consistent with MGUS or SMM/IMM. Although serial studies have been performed in only two patients, these anecdotes suggest that the malignant plasma cells can change their rates of proliferation, apoptosis or PCGI, and that these changes may be clinically meaningful.

There are several methods to measure tumour cell apoptosis (McGahon et al, 1995). The 7-AAD flow cytometry method we have described is well suited for the clinical laboratory, uses a small amount of marrow aspirate, does not require density-gradient centrifugation, and is easy to add to the PCLI. We postulate that the PCGI, which includes measurement of both the PCAI and the PCLI, may provide a better estimate of net tumour growth. The PCGI will need to be studied in larger groups of patients over longer periods of time to discover if the apoptosis rate or the PCGI predicts survival better than the PCLI alone. These studies are best performed on patients entering randomized clinical trials where the treatment and follow-up is conducted in a standardized manner.

Other studies using different methods to detect apoptosis have been performed on various tumours. Philpott et al (1995) found a higher percentage of the CD34+ cells to be in apoptosis in the marrow of patients with aplastic anaemia compared to normal controls. In a small study of paraffin-embedded gastric cancers, Kasagi et al (1994) found a mean apoptosis index of 10.9% in well-differentiated cancers versus 4.0% in poorly differentiated cancers. They found no correlation between the apoptosis and proliferation rates on the same tumour sample. Steck et al (1996) found a mean apoptosis index of 8.2% in 55 paraffin-embedded breast cancers. The apoptosis index directly correlated with the cell proliferation (%S+%G2M) results. King et al (1996) determined that the apoptosis index of 54 samples of transitional cell carcinomas of the bladder was higher than normal transitional epithelium (range 0.6–1.8%); however, the apoptosis indiex did not correlate with disease stage. Lu & Tanigawa (1997) found a median apoptosis index of 0.7 in 101 cases of gastric carcinoma. There was no correlation between the apoptosis and proliferation rates. A low apoptotic index was associated with increased tumour vascularity; however, intratumour microvessel density did not correlate with tumour cell proliferation. In our study the PCAI decreased with advancing disease and there was an inverse relationship between myeloma cell proliferation and apoptosis. Therefore, in general, these studies do not show consistent results as to the correlation between apoptosis and proliferation and between apoptosis and disease stage. It is probable that these relationships may differ between tumour systems.

Our results support the hypothesis that there are two defects in plasma cell disorders: proliferative and apoptotic. Both of these defects can result from the action of IL-6. Multiple previous studies have demonstrated that IL-6 can cause plasma cell proliferation (Jelinek et al, 1997; Klein et al, 1995; Lichtenstein et al, 1989; Ogata et al, 1997; Spets et al, 1997; Urashima et al, 1997; Westendorf et al, 1994, 1996). It has also been demonstrated that IL-6 and IGFs can protect myeloma cells from dexamethasone-induced apoptosis (Lichtenstein et al, 1995; Xu et al, 1997). Schwarze & Hawley (1995) demonstrated that IL-6 can suppress apoptosis by up-regulating cellular bcl-x mRNA and Bcl-xL protein. Therefore IL-6 and IGFs are both proliferation and survival factors for myeloma cells. The studies by Krajewski et al (1997) also provide support for a decrease in apoptosis as a defect in malignant plasma cells. They used immunohistochemistry to stain normal and malignant plasma cells for the apoptosis-effector protease CPP32 (Caspase-3). They found normal plasma cells to be CPP32 positive whereas 33% (4/12) plasmacytomas were CPP32 negative. Teoh et al (1997) studied murine double minute 2 (MDM2) expression and function in myeloma cells and found that inhibition of MDM2 using antisense oligodeoxyribonucleotides induced MM cell apoptosis. They also demonstrated binding of MDM2 to E2F-1 and postulated that MDM2 may contribute to both growth and survival of myeloma cells.

The finding of defects in myeloma cell apoptosis have potential therapeutic implications. The proliferative phases of myeloma such as newly diagnosed and relapsed disease may require high-dose therapy, whereas the nonproliferative phases such as MGUS, SMM and plateau-phase disease may be more amenable to immunologic approaches or non-myelosuppressive therapy. Further research into the mechanism(s) of plasma cell apoptosis and the effects of different agents on apoptosis is needed to target drugs for each of these phases of myeloma cell growth.

Acknowledgements

This work was supported by grant number CA62242 from the National Cancer Institute (National Institutes of Health).

Ancillary