Method and Apparatus for Time-Resolved Measurements of Lymphocyte Function and Aggregate Structure Using Computer-Automated Microscopy

A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.

The invention relates to a method and apparatus for time-resolved measurements of lymphocyte function and cell aggregate structure using computer-automated microscopy.

Presently accepted methods for quantifying adhesion mechanics among cell aggregates are insensitive to the fine details of molecular pathways and structures which govern the cell aggregation process. Current methods of data presentation loose much information, and many physical characteristics of the aggregated cells which are important to the understanding of cell adhesion mechanics are never considered.

Known methods for quantifying cell aggregation often blur the distinction between aggregate shapes by use of rather simplistic, subjective scoring methods. For example, one known method for quantifying aggregate structure is through the use of a relative number scale presented below.

1+=<10% of cells in small aggregates (5-10 cells/clump)

2+=10-40% of cells aggregated (size varies)

3+=40-90% of cells aggregated (size varies)

4+=>90% of cells aggregated, usually in medium to large clumps

5+=approximately 100% of cells aggregated in large clusters

In order to eliminate the effects of personal bias, assays using this number scale are usually scored double-blind. While this has proved somewhat helpful in eliminating errors resulting from the subjectivity of the observer, the scoring method is still subject to human error and to variability between the perception of different observers. More importantly, this number scale does not reflect the potential differences in the physical dynamics or morphology that may exist between various adhesion events.

For example, using this number scoring method, adhesion induced in a human lymphoblastoid T-cell line, JURKAT, by anti-VLA-4 antibody was 5+, adhesion induced by the lectin PHA (phytohemagglutinin) was 5+, while adhesion induced by the phorbolester PMA (phorbol 12-myristate 13-acetate) was scored a 4+. These similar number scores occur despite the fact anti-VLA-4 induced aggregates are large round and compact, with very few free non-aggregated cells, aggregates induced by PHA are long-chained structures, and those induced by PMA are smaller, more loosely structured and contain far more non-aggregated cells. Stated otherwise, application of the known number scoring method to these three aggregate structures results in very similar numerical scores despite a markedly different appearance of the aggregates.

Another limitation of the simplistic numerical scoring method is the inability to establish rates of aggregate formation. Again, this results in a substantial loss of important information. For example, two antibodies that induce lymphocyte adhesion have recently been discovered; L25 (anti-VLA-4) and IC9 (ligand presently unknown). Both antibodies induce 5+ adhesion, but it has been observed that the rate of IC9-induced aggregation is considerably faster than that of L25-induced adhesion. A complete understanding of the different adhesion mechanisms induced by these two antibodies requires assessment of the relative rates of aggregation, an assessment which is not possible using presently available methodologies.

The present invention avoids the above-noted drawbacks of the prior art by quantifying morphological features of cell aggregates with several new structural measurements obtained by digital image analysis. The invention involves obtaining magnified images of cell aggregates, digitizing and filtering the images, segmentation of the processed images to identify the aggregates, and analysis of the segmented images to determine size, perimeter, shape and texture of each aggregate within the images.

The new structural measurements of the present invention include aggregate size distribution, aggregate size frequency distribution, aggregate shape factor distribution, the coverage of the image area by cell aggregates, and the integral optical density of the digitized image. These new structural measurements offer information regarding the size distribution, texture, shape and optical density of the cell aggregates, and also provide an indication of the percentage of the total image area that is covered by aggregates. These new parameters provide additional information which is useful in classifying subtle differences in the character of aggregation in various cell and induction systems, and may provide new insight regarding the physical and molecular bases of each.

The apparatus of the present invention includes a television camera for producing a video image of a magnified number of cell aggregates, a digitizer for digitizing the video image, and a computer for analyzing the digitized images, and for calculating the above structural parameters.

In addition, in order to assess the time rate of cell aggregation, a number of video images can be taken at spaced apart time intervals, and the structural parameters can be calculated for each image in the sequence. This will provide cell aggregation rate information, also valuable in morphological studies.

FIG. 1 is the apparatus of the present invention used to practice the method of the present invention.

FIG. 2 illustrates the image acquisition of the present invention.

FIG. 3 is a flow chart of the operation of the apparatus of FIG. 1.

FIG. 4 is a flow chart of the calculation of the morphological indices of the present invention.

FIGS. 5A-D are computer-generated images of different aggregated cell structures used to illustrate the present invention.

FIGS. 6A-D are graphs of the aggregate frequency distribution parameter of the present invention calculated for the cell aggregations of FIGS. 5A-D.

FIGS. 7A-D are graphs of the aggregate size distribution parameter of the present invention calculated for the cell aggregate structures of FIGS. 5A-D.

FIGS. 8A-D are graphs of the shape factor parameter of the present invention calculated for the cell aggregate structures of FIGS. 5A-D.

FIG. 9 is a graph of the fraction of image area covered by aggregates for the cell aggregates structures of FIGS. 5A-D.

FIGS. 10A-C are graphs of the integral optical density parameter of the present invention.

Referring to FIG. 1, an apparatus according to the present invention which operates according to the method of the present invention is disclosed. Connected to inverted microscope 11 is video camera 12 which produces on line 13 NTSC video signals of the magnified image produced by microscope 11.

Cell aggregates under study are preferably contained within Terasaki plastic plates with inverted frusto-conical tapered wells having a 1.3 mm diameter bottom and a 3.9 mm diameter top. According to the preferred method of the present invention, a single drop of 1.5 microliters from a cell suspension is placed with a pipette on the bottom of each well in the Terasaki plate. It has been determined that volumes of cell suspension larger than 1.5 microliters lead to uneven distribution of the cells on the bottom of the well, with the majority of cells near the periphery of the well. This non-uniformity leads to erroneous results.

An appropriate reagent which induces or inhibits cell aggregation is placed in each of the wells, and the Terasaki plate is incubated at an appropriate temperature. The Terasaki plate is then placed on the stage of microscope 11.

Microscope 11 preferably has a 10.times. objective which provides a 1.00.times.0.75 millimeter field of view covering almost the entire cross section of an individual well in the Terasaki plate.

Referring also to FIG. 2, four images 16, 17, 18 and 19 of the well are taken in order to provide complete imaging of the cell aggregates on bottom 20 of each individual well of the Terasaki plate. The images are later combined during data processing to form a single image that contains all cells in a particular well.

Images are taken at various times and analyzed, or can be recorded by, for example, video cassette recorder 15. Video monitor 14 can be used to confirm the images produced by video camera 12. Video switch 21 is used to direct video images from camera 12 to recorder 15 through line 22, or for direct application to computer 23 through line 24.

Computer 23 can be, for example, a VAXstation II/GPX computer, available from Digital Equipment Corporation. Computer 23 is equipped with a video image digitizer board, for example a FG100 image digitizer available from Imaging Technologies, Inc., which digitizes the analog images provided by camera 12 into 640.times.480 8-bit pixels thereby providing 256 grey levels for each pixel. Preferably a total of 6 wells are imaged in order to obtain good population statistics.

The combined four images (FIG. 2) are then analyzed in computer 23 according to the procedures described in the flow charts of FIGS. 3 and 4, to calculate the morphological indices of the present invention in order to assess aggregate structure and lymphocyte function. The morphological indices can be displayed in histogram form on video display terminal 25.

Referring to FIG. 3, the operation of computer 23 according to the present invention is disclosed in flow chart form. Most of the procedures shown in FIG. 3 are preferably embodied in a commercially available image processing software package, Icoemtrics, available from Everest Technologies. Calculation of the specific aggregate structural parameters according to the present invention is shown in block 42 of FIG. 3, in slightly more detail in flow chart form in FIG. 4, and in the form of a source code computer program at the end of this specification.

Referring to FIG. 3, beginning in block 26, a calibration routine is performed in order to automatically accommodate variations in the magnification power of microscope 11. Control is then transferred to block 27 where the image digitizer within computer 23 is activated to digitize the video image. Then, in block 28, the digitized video image is clipped to remove black lines that occasionally appear around the boarders of the digitized image. Control then passes to decision block 29 where it is determined whether the video image is derived from the wells of a Terasaki plate. If not, control passes to block 30 where the digitized image is filtered with an averaging filter using a 3.times.3 kernel, and by a gradient filter with a 3.times.3 neighborhood. Control then passes to block 31 where image enhancement is performed by multiplying the image by itself using double precision arithmetic and then compressing again to 8-bits per pixel. Control then passes to block 32 where the grey levels of the image are segmented. In block 33, the image is eroded and dilated twice to smooth aggregates and to eliminate small one- or two-pixel objects that are due to pixelization errors and video noise. Control then passes to block 34 where the image is processed to remove holes appearing in the middle of cell aggregates. Control then passes to block 35 where images from several different cell cultures are combined to provide good population statistics.

On the other hand, if decision block 29 determines that Terasaki plates are used to produce the image, control passes to blocks 36, 37 and 38 which perform functions identical to blocks 31, 32 and 33, described above. Control then passes to block 39 where the portion of the image corresponding to the circular edge of the Terasaki plate is removed. Holes are then removed from within the image of the cell aggregates in block 46, using an identical procedure as that of block 34.

After images are combined in block 35 to improve cell population statistics, control passes to block 41 where objects within the image which are attributable to partially visible aggregates at the edges of the image are removed. Thus, the image contains only complete aggregates found in the field of view, thereby preventing erroneous morphological analysis by including artificially chopped aggregate images.

Finally, control passes to block 42 for calculation of the structural parameters of the present invention (described in more detail referring to FIG. 4), and the parameters are then formatted for display in block 43.

Referring to FIG. 4, the calculation of the specific structural parameters of the present invention is disclosed. The source code computer program appearing at the end of this specification presents the details of these procedures. In block 44, the frequency distribution function, n(A), for aggregate profiles is calculated, where A is the size, in .mu.m.sup.2, of the aggregate profiles or unconjugated cells identified and measured during digital image analysis. The frequency distribution function n(A) is defined as:

Then, in block 45 the cell aggregate size distribution function, s(A) is calculated according to the following equation:

Where A.sub.t is the sum of the area of all aggregates in the image. Next, in block 46, the shape factor, .tau., for each cell aggregate in the image is calculated according to the equation: ##EQU1## Where: P is the perimeter of the aggregate and A is its area. This descriptor takes its minimum value of one for a circle. Then, also in block 46, the shape factor distribution function s(.tau.) is calculated according to the equation:

Next, in block 47, the coverage, C, of the image area is calculated according to the equation:

Where: A.sub.t is the area in the image covered by aggregates or unconjugated cells, and A.sub.img is the total image area.

Finally, in block 48, the integral optical density descriptor, IOD, is calculated according to the equation: ##EQU2## Where m(d) is the number of pixels having an optical density, or grey level, d, and where m.sub.t is the total number of pixels in the area of interest. Low values of optical density, which correspond to dark internal areas, indicate the formation of multi-layer aggregates.

Control is then returned to the routine of FIG. 3 for display of the results.

FIGS. 5-9 illustrate a specific application of the present invention. The computer-generated images of FIGS. 5A-D correspond, respectively, to the aggregates of JY cells using different induction protocols. Specifically, FIG. 5A is the control and illustrates spontaneous aggregation of JY cells. FIG. 5B illustrates aggregation induced by L25 (anti-VLA-4). FIG. 5C is aggregation induced by IC9 (ligand presently unknown), and FIG. 5D is PMA (phorbol 12-myristate 13-acetate) induced aggregation. All four images were obtained four hours after the start of the experiment. By visual observation, it can be seen that the aggregates of FIGS. 5A-D exhibit a wide sized distribution, and the that the different aggregate induction protocols produce aggregates with widely different morphologies.

In the control, FIG. 5A, aggregates are chain or chain-bead structures and almost all appear to consist of a single layer. A large fraction of the cells do not form aggregates.

In the L25-induced adhesion of FIG. 5B, aggregates are large beads with very few short-chained structures. Many aggregates are dark in the middle with a characteristic bright single-cell ring around their periphery. Such aggregations are most likely piles of cells including several layers of cells, but this 3-dimensional structure cannot be confirmed by using standard light transmission microscopy.

The IC9-induced adhesion of FIG. 5C illustrates aggregates that are almost exclusively large multi-layered clusters, with very few unconjugated cells or single-layer clusters.

The PMA-induce adhesion aggregates shown in FIG. 5D include aggregates which are circular and rather small, consisting of only a single, or possibly two, layers of cells.

By application of the present invention, the graphs shown in FIGS. 6-10 result.

FIGS. 6A-D present the frequency distribution functions, n(A)dA, respectively computed for the four adhesion protocols shown in FIGS. 5A-D. All four distributions exhibit a maximum in the range of 200-400 .mu.m.sup.2. This corresponds to unconjugated JY cells. Also, all plots show a relatively small fraction of conjugates with sizes greater than 2000 .mu.m.sup.2 with the possible exception of PMA-induced aggregation, which exhibits a maximum of around 3000 .mu.m.sup.2. The plots of FIG. 6 are notable for their lack of a distinguishing characteristics. The similarities between the plots obtained for spontaneous adhesion and for L25-induced adhesion are particularly striking given the totally different structures of the aggregates in those cases (FIGS. 5A and B).

The size distribution functions, s(A)dA, shown respectively in FIGS. 7A-D, on the other hand, clearly indicate significant morphological differences among the aggregates obtained with the four different protocols. Referring to FIG. 7A, with spontaneous adhesion, several distinct peaks are identified. The first peak occurs in the 200-400 .mu.m.sup.2 range and corresponds to unconjugated cells. After a plateau, indicating the presence of the small clusters with 5-10 cells each, a second peak appears in the 4000-10,000 .mu.m.sup.2 range. The third peak is actually due to several smaller clumps or chains that have joined together to form chain-bead and highly tortuous structures.

In the L25-induced adhesion of FIG. 7B, one major peak in the 10,000-40,000 .mu.m.sup.2 range corresponds to large aggregates. A substantial amount of unconjugated and small clusters is indicated by a broad smaller peak extending from 200-3,000 .mu.m.sup.2.

FIG. 7C, IC9-induced adhesion, exhibits a peak corresponding to large aggregates that is higher, narrower and slightly shifted toward larger areas. This indicates very few unconjugated cells and clumps.

Finally, PMA-induced adhesion, shown in FIG. 7D, is similar to the control of FIG. 7A and exhibits three peaks. The first is mostly attributed to unconjugated cells. The major peak appears in the 8,000-10,000 .mu.m.sup.2 range and corresponds to medium-sized clumps. A much smaller peak at 30,000 .mu.m.sup.2 indicates the presence of a few large clusters.

Turning to the graphs of the shape factor distribution function, s(.tau.)d.tau., for the four different protocols shown FIG. 8A-D, IC9-induced adhesion (FIG. 5C) leads to aggregates with almost circular profiles (over 90% of the aggregates having shape factors between 1 and 2), while L25-induced adhesion (FIG. 5B) has a wider distribution, and the aggregates produced by the control (FIG. 5A) and by PMA-induced adhesion (FIG. 5D) are rather loose structures having shape factors exceeding the value of 10.

FIG. 9 shows the coverage, C, computed for the four cases of adhesion. The smallest coverage is observed for IC9-induced adhesion, indicating a tight-packing of cells in the aggregates, possibly in a 3-dimensional multi-layered arrangement.

FIGS. 10A-C is the frequency distribution function of optical density for unconjugated, small aggregates, and large aggregates, respectively, showing aggregates of both IC9-induced and L25-induced adhesion (FIGS. 5C and B, respectively).

The time rate of cell aggregation also provides valuable information regarding the physical and molecular events underlying the adhesion process. The morphological indices of the present invention can be calculated for each image of a time-sequence of images of a cell suspension undergoing aggregation, thereby providing objective cell aggregation information as a function of time.

The present invention which quantifies lymphocyte adhesion is insensitive to possible differences in the physical and molecular events which govern the adhesion process. Consequently, the assay can be used to classify subtle differences in the character of lymphocyte aggregation in various cell and induction systems. The results provide new insight regarding the physical and molecular bases of each, and allow investigation of several issues.

For example, evaluation of digital images of aggregation induced with and without the presence of monocolonal antibodies to various surface molecules provides an increased chance to observe any aggregation inhibition. Any change in the structural parameters measured by the present invention due to the addition of a particular monocolonal antibody would indicate that the corresponding antigen was involved.

In addition, it is presently hypothesized that magnetism, and not calcium, is required for LFA-1 dependent adhesion to occur. The increased sensitivity of the image analysis of the present invention will provide a more rigorous test for this hypothesis.

Also, the techniques of the present invention can be used to investigate synergism between different modes of adhesion induction. For example, if PMA-induced adhesion is a completely committed process, the addition of L25 or IC9 will not alter the character of the adhesion. In other words, once cells are committed to a PMA-induced adhesion process, other adhesion inducers would have no affect, and the structural parameters of the present invention would indicate that the adhesion event occurred like normal PMA-induced adhesion. 0n the other hand, if PMA- and IC9-induced adhesion are complementary, lymphocytes exposed to both induction methods should demonstrate more complete aggregation than either induction protocol alone. However, if IC9- and PMA-induction mediated adhesion through identical mechanisms, then lymphocytes exposed to both adhesion inducers will exhibit adhesion characteristics for the most efficient process, in this case IC9.

These examples of specific applications of the present invention to assess adhesion mechanics will be understood by those of skill in the art to be merely exemplary, and not limiting of the present invention. In addition, one of skill in this art will appreciate that changes, additions and modifications can be made to the apparatus of the present invention, and that different apparatus can be used to practice the method of the present invention, without departing from the spirit and scope of this invention.

The following is a source code listing of a computer program for computer 23 of FIG. 1, and details the procedures shown in flow-chart form in FIG. 4. ##SPC1##