Study design
Three techniques were assessed for their ability to identify AM. First, the standard FC techniques for identifying AM,
i.e. intrinsic forward scatter and autofluorescence. Second, the ability of the LSC to capture and contour around AM nuclei, stained with the nucleic acid dye propidium iodide (PI). Thirdly, capture of whole AM when the cytoplasm was stained with a fluorescently conjugated monoclonal antibody to CD68, which reacts selectively with a specific 110-kDa cytoplasmic glycoprotein present in mononuclear phagocytes. Slides were processed as outlined below and where data was obtainable for any of the above techniques the "sensitivity" and "positive predictive value" of the technique was tested as well as the repeatability of the analysis.
Methods
[size=-1]
Bronchoalveolar lavage[/size]
All children received a general anaesthetic. At induction, either an inhalational agent or propofol was used according to the preference of the anaesthetist. Immediately after intubation and within 5 min of induction of anaesthesia, children underwent nonbronchoscopic BAL as previously described
16. Briefly, the child's head was turned to the left and a suction catheter (six French Gauge (FG) <1 yrs, eight FG 1–5 yrs and 10 FG >5 yrs) was wedged in the lower airway. A total 1 mL·kg body weight–1 (maximum 20 mL) of sterile saline at room temperature was instilled and immediately aspirated into a suction trap using gentle negative pressure (100–200 mmHg). The procedure was repeated twice. Pooled bronchoalveolar lavage fluid (BALF) was stored on ice prior to processing.
[size=-1]
Processing of samples[/size]
BALF was immediately centrifuged at 250[font=arial,helvetica]x[/font]
g for 5 min at 4°C. After removal of the supernatant the cell pellet was resuspended in phosphate-buffered saline (PBS) (Sigma-Aldrich Company, Dorset, UK). This procedure was repeated twice and the cells resuspended in PBS at a final concentration of 1[font=arial,helvetica]x[/font]105 cells·mL–1. A total 1[font=arial,helvetica]x[/font]104 unfixed cells were cytocentrifuged at 800 rpm for 5 min onto several glass microscope slides (Shandon Scientific, Runcorn, Cheshire, UK). Slides were then air-dried for 1 h before fixing in chloroform:acetone mix at a ratio of 1:1 for 10 min. After fixation, slides were air dried for a further hour prior to storing for up to 18 months at –20°C until immunoassayed. The BALF cell differential count was determined prior to fixation by staining a slide with Diff-Quik (Dade Behring, Deerfield, IL, USA) and counting >300 leukocytes. Slides were brought up to room temperature and washed in PBS for 10 min. Slides were processed for each child in one of three ways.
[size=-1]
Capture using autofluorescence and forward scatter[/size]
Slides were left unstained in order to assess contouring around forward scatter and autofluorescent properties of the cell.
[size=-1]
Capture using nuclear staining[/size]
To one slide, PI (Sigma-Aldrich) was added at a concentration of 20 µg·mL–1 with Ribonuclease H (Sigma-Aldrich) at a concentration of 15 µg·mL–1 for 30 min at room temperature in the dark. Preliminary work using concentrations of PI ranging from 20–100 µg·mL–1 achieved the same intensity of staining.
[size=-1]
Capture using a pan-macrophage monoclonal antibody[/size]
Two slides from each child were processed using standard immunostaining techniques after blocking with goat serum for 25 min. One slide was incubated with a mouse monoclonal antibody to human CD68 (BD Pharmingen, San Diego, CA, USA) in goat serum at a final concentration of 10 µg·mL–1, and one with an approriate isotypic control (mouse immunoglobulin (Ig)G2b) (Serotec Ltd, Oxford, UK) in goat serum for 1 h. Both slides were then washed twice in PBS before incubation with the secondary antibody Alexa Fluor 633 goat anti-mouse IgG2b (Molecular Probes, Leiden, the Netherlands) in goat serum at a final concentration of 13 µg·mL–1 at room temperature for 30 min in the dark.
[size=-1]
Laser scanning cytometry[/size]
All slides were mounted in Immunomount (Shandon Scientific) and processed on the LSC within 48 h. Slides were scanned on the LSC (CompuCyte, Cambridge, MA, USA) and cells detected by contouring around one of the following: intrinsic cellular autofluorescence, forward scatter, nucleic acid PI and cytoplasmic CD68. For each slide at the start of scanning the threshold photomultiplier tube (PMT) sensitivity was adjusted such that the maximum detection of stained areas was achieved with a minimum detection of background staining. Once adjusted, the slide was scanned and data collected. A minimum of 1,000 cells per slide were analysed by the LSC using the [font=arial,helvetica]x[/font]20 objective.
During scanning, images of the area being scanned were randomly selected and displayed on the data display window and inspected visually. This window shows pixelated areas and the presence of software-generated contours. This enables an assessment of whether contours have appropriately identified single AM, or have inappropriately characterised debris or clumps of cells as individual AM with a single contour, or even failed to detect and hence contour around AM. For each child a minimum of three data display windows were examined. AM were identified by the operator according to either the nuclear or cell size and contour shape and the following parameters were calculated: 1) the proportion of pixelated clusters that were morphologically single AM and were contoured around (sensitivity), and 2) the proportion of contours associated with pixelated clusters that were morphologically single AM (positive predictive value).
For AM stained with CD68, a qualitative assessment of specificity and negative predictive value for the sample was made by comparing the amount of pixelated clusters contoured around on the isotypic control slide compared to the CD68 stained slide. The variation in repeated measurements on the same subject (repeatability) was determined in AM from three randomly selected children. Two sets of slides obtained from a single sampling procedure and stored in an identical manner were immunostained with CD68 on two separate occasions and then scanned on the same LSC by the same observer each within 48 h of immunostaining.
After imaging, a presumed population of single AM were selected according to homogeneity and contour area from the software produced graphs of contour area
versus pixel intensity. These "cells" were then displayed using the "relocation" feature on the software. For each sample, once scanning was recommenced, the first 36 contoured pixelated clusters that were within the selected area were visualised and the positive predictive value of the test for selecting single AM was calculated. This determined whether selecting a homogeneous population of cells with similar contour size would exclude debris or clumps of cells that had been inappropriately contoured around and increase the positive predictive value of the test for selecting single cells.
To determine whether it was possible to obtain fluorescent data within the contours, the integrated fluorescence was determined for each of the fluorochromes used. The integrated fluorescence for the active sensor is the background-corrected sum of the pixel values within the data contour (equivalent to FC-integrated fluorescence). The integrated fluorescence was corrected for contour area.
Confocal microscopy
To independently assess the specificity of CD68 fluorescent staining of AM and the degree of background staining, CD68-immunostained AM were imaged using the laser scanning confocal microscope (Bio-Rad Radiance 2000, Hertfordshire, UK). Positively stained slides were compared to the isotypic control and a qualitative assessment of the specificity and negative predictive value of CD68 immunostaining was estimated.
Statistics
The data are presented as medians and interquartile ranges (IQR). Sensitivities and positive predictive values were calculated according to the methods by Altman and Bland
17,
18. In order to assess the repeatability of the test, the Bland-Altman repeatability coefficient was calculated. This defines the range in which 95% of measurements for sensitivity and positive predictive value would be expected to fall if the test was repeated on the same subject under the same conditions of measurement and is the probability of technical error. It was calculated by plotting the difference between the test and retest results for the three children studied and then calculating the mean and sd of the differences. The repeatability coefficient was calculated as twice the sd of the difference of the repeated measures
19.