Quantification of vascular graft seeding by use of computer-assisted image analysis and genetically modified endothelial cells K u r t D. N e w m a n , M D , Nga Nguyen, BS, and David A. Dichek, M D , Bethesda, Md. Current methods for the evaluation of retention of endothelial cells seeded on vascular grafts are limited by the inability to specifically identify and quantitate seeded cells on a long-term basis. To address this problem we developed a method of quantification of graft surface coverage using genetic labeling of endothelial cells combined with computerassisted image analysis. Rabbit aortic endothelial cells were transduced with a marker gene (lac-Z) and seeded on polytetrafluoroethylenegrafts. After histochemical staining in which the genetically labeled cells turn blue, computer-assisted image analysis was used to measure the percentage of graft surface covered by the seeded cells. The utility of the method was evaluated by using it to assess the effect on graft coverage of seeded cell density and by precoating with fibronectin. Quantification of surface area coverage was automated and reproducible both between scans and between observers. Use of this method allowed the determination of a linear correlation between cell density in the seeding suspension and graft coverage (re -- 0.93, p < 0.0001). The method also permitted confirmation of the positive contribution of fibronectin coating to graft coverage by seeded cells: 73% coverage coated versus 8% coverage uncoated (p < 0.0001). The ability of this method to specifically identify genetically marked endothelial cells and their progeny makes it attractive for use in studies targeted at optimization of graft coverage in vivo. (J Vase St~c 1991;14:140-6.)

Endothelial cell seeding of small diameter vascular grafts to enhance long-term graft patency has been a goal of substantial research effort. 1-3 In animal studies seeding of grafts is associated with an improvement in patency. 4°5 H u m a n clinical trials, although preliminary, have also been encouraging. 6"7 Two hypotheses are implicit in all graft seeding studies: (1) A graft that is covered with seeded cells is more likely to remain patent than an incompletely covered graft. (2) Seeded cells remain on the graft after implantation, and consequent increases in graft patency are a direct result of the continued presence of these cells in vivo. In accordance with these hypotheses several quantitative methods have been developed and are currently used as tools in the optimization of cell seeding and seeded cell retention. From the Molecular Hematology Branch, National Heart, Lung, and Blood Institute, Bethesda, Md. Dr. Newman is a member of the Department of Surgery, Children's National Medical Center, Washington, D.C. Presented at the American Heart Association Scientific Sessions, Dallas, Texas, November 12, 1990. Reprint requests: David A. Dichek, MD, Molecular Hematology Branch, National Heart Lm~g and Blood Institute, Building 10, Room 7D-18, Bethesda, MD, 20892. 24/1/29421 140

These methods include radiolabeling of cells, planimetry, assessment of thrombus-free graft surface area, evaluation of platelet morphology, and quantification of deoxyribonucleic acid (DNA) extracted from seeded cells. 8-1S Although each of these methods has its advantages, none permits the specific identification and quantitation of seeded (versus ingrowing) cells on a long-term basis. Therefore none of these methods is able to specifically address the question of whether long-term increases in graft patency are correlated with the persistence of large numbers of seeded cells. Recently Zwiebel et al. 16 using retroviral vectors and Etchberger et al. 17 using plasmid D N A demonstrated that foreign genes can be introduced into endothelial cells. 16"17These inserted genes may serve as markers to identify specific cells and their progeny. In particular, insertion of the Escherichia coli lac-Z gene allows identification of transduced cells with a simple histochemical stain, in which cells expressing the lac-Z gene turn blue in the presence of an added substrate, and cells lacking the gene remain colorless. 18"~9 Coupling this genetic marking technique with an automated scanning system capable of specifically identifying and quantitating the marked

Journal of VASCULAR SURGERY

142 Newman, Nguyen, and Dichek

+4

B ,¢.~-p ++7!+, • 4+ ~ i + . +¢++,j + + + ~ +~++...+'7 ++ ~+++~ + + ++.+,,'++

++++,+++ + *+.~++:+++++ ~'~

O,

D

G

H ~

+ iiii~i "'i~ ~, i+i~i¸+~+¸¸

~ ~!

+

++~

~++

"~

Fig. 1. Microscopic (original magnification × 25) and macroscopic views of polytetrafluoroethylene vascular graft segments seeded at the indicated cell density (cells/cm2 of graft surface area): A and B, 2 x 105; C and D, 7 x 104; E and F, 2 x 10% G and H, 7 × 10~. +

C

D

Fig. 2. Microscopic (original magnification x25)and macroscopic views of polytetrafluoroethylene graft segments incubated either with or without fibronectin then seeded at a density of 2 × 10 s cells/cm2. A and C, with fibronectin coating; B and D, without fibronectin coating. and by a different observer using a slightly modified sampling protocol (fields sampled along lateral rather than longitudinal axes) for the selection o f the 15 measured fields. Further experiments were performed in which grafts were seeded with mixtures of BAG transduced and untransduced cells. Pure populations as well as defined mixtures (i.e., 1 : 3, 1 : 1, and 3 : 1 ratios of

BAG transduced to untransduced cells) were seeded onto ePTFE grafts at a total seeding density of 2 x 105 cells/cm2. These experiments were performed to determine (1) whether genetically labeled cells could be identified and quantitated when present within a population of unmarked cells and (2) whether mixtures of genetically labeled and unlabeled cells could be used to set up a standard curve

Volume ?.4 Number 2 August 1991

cells would seem to offer an ideal method for the quantitation of graft surface coverage by seeded cells in vitro and eventually in ex vivo specimens. With this goal of quantitation of graft surface coverage by seeded cells in view, we developed a method of computer-assisted image analysis of grafts seeded with genetically marked endothelial cells. We tested the image analysis method for reproducibility when repeated by the same observer and when performed by a different observer. In addition, we assessed the ability of the method to provide an accurate measurement of seeded cell coverage by comparing the results we obtained by computer scanning to those obtained by macroscopic evaluation of the stained graft surfaces. Finally we used the method to perform quantitative studies aimed at optimization of graft coverage in vitro. MATERIAL A N D METHODS Cell culture and transduction Rabbit aortic endothelial cells (RAECs) were initially cultured on fibronectin and were subsequently maintained on a gelatin substrate in Ryan Red medium. 2° Rabbit aortic endothelial cells were transduced with the BAG retroviral vector by use of methods described previously.16'18'Ig"2~The cells were transduced at passage 9 and used for seeding studies at passages 28 through 34. Untransduced endothelial cells were used at passage 29. The BAG vector contains both the E. coli lac-Z gene (coding for the bacterial enzyme 13-galactosidase) and the neomycin resistance gene. Transduced cells are therefore resistant to G418 (a neomycin analog) and stain blue with 5-bromo-4-chloro-3-indolyl-~3-D-galactopyranoside (X~gal), a chromogenic substrate for ~-galactosidase. After transduction the cells were selected in G418 for at least 2 weeks to eliminate untransduced cells.

Graft seeding and staining Four or 6 cm long segments of 6 mm internal diameter expanded polytetrafluoroethylene (ePTFE) vascular grafts (W. L. Gore & Associates, Elkton, Md.), were seeded with a single cell suspension of transduced RAECs. Grafts were first incubated in phosphate-buffered saline (PBS) for 30 minutes at 37 ° C then, except as noted below, were incubated with 3 i~g/cm 2 of human fibronectin in PBS with calcium and magnesium (Biofluids, Rockville, Md.) for 1 hour. After rinsing the grafts with PBS, and sealing one end with a plastic plug, approximately 1.0 to 1.5 ml of a suspension of RAECs in Ryan Red medium with 100 units/ml penicillin, 0.25 ~xg/ml amphotericin B, and 100 ~tg/ml streptomycin added

Computer analysisofgrafl seeding 141 (Biofluids) was introduced into the graft lmnens tilting them completely. The other end of the grafts was sealed with a second plastic plug, and the grafts were slowly rotated (1.5 rpm) for 15 hours at 37 ° C. The plugs were then removed and the grafts transferred to fresh medium in stationary culture tubes in a 5% CO 2 37 ° C incubator for 72 hours. Experiments were carried out to examine the effects of seeded cell concentration and fibronectin precoating on graft coverage. Fibronectin precoated grafts were seeded with suspensions of the following cell concentrations (cells/cm2 of graft surface area): 2 x 10 ~ ( n = 7 ) , 7 x 104 ( n = 4 ) , 2 x l04 7 x 103, ( n = 3), 2 x 103 ( n = 3). Three additional grafts were seeded with 2 x 10 s cells/cm2 without fibronectin precoating. After the 72-hour incubation, grafts were fixed with 2% formaldehyde and 0.2% glutaraldehyde in PBS, then stained with X-gal as described elsewhere. 22After staining, grafts were rinsed in PBS, cut longitudinally, and mounted between glass slides in 90% glycerol and 10% PBS. Computer-assisted image analysis The seeded, stained grafts were examined at x 10 on an inverted microscope (Olympus, Tokyo, Japan). Fields were imaged on a model ITC-50 video camera (Ikegami, Ma}weood, N.J.), and data were transferred into an Optomax V (Analytical Instruments Ltd., Essex, U.K.) image analyzer scanning system. 23 Blue cells were recognized by the Optomax as "features." For each microscopic feld the percentage area covered with features was measured. The system was programmed to measure the area of the field covered by the blue cells, and to express this area as a fraction of the total surface area in each field. A standardized sampling protocol was used to estimate the fractional surface coverage of individual graft segments. Only the central area of each graft segment was seeded, because the ends of the graft segments had been occluded with plugs before seeding (Figs. 1 and 2). From within the central area of each graft segment, 15 fields were selected for microscopic examination. The fields were selected by use of a grid with three lines (5 mm apart) with five evenly spaced points (2 mm apart) along each line. From the 15 individual measurements, a mean and standard deviation of fractional surface coverage for the entire graft were calculated. To assess the ability . of this sampling technique to give a reproducible result, the protocol was repeated three times for each graft by the same observer. In addition, each graft was scanned by the same observer on different days

Volume 14 Number 2 August 1991

according to which the coverage of experimental grafts with unknown numbers of transduced cells might eventually be determined. Grafts in these experiments were incubated, stained, and scanned as described above. Statistical analysis Data were analyzed with Student's t test and nnweighted least squares linear regression by use of a computer program (Statistix 3.1, St. Paul, Minn.). RESULTS Quantitation o f surface area coverage Seven grafts that were seeded with a cell concentration of 2 × 10 s cells/cm 2 appeared to be covered with a confluent layer of RAECs both by microscopic and macroscopic observations (Figs. 1 A and B). Analysis of fractional surface area coverage with the Optomax scanning system revealed a coverage of 73% + 10% (mean + SD) of the grafts (Table I). Reproducibility of fractional surface area determination Separate determinations of mean fractional surface coverage of grafts by the same observer typically varied by less than 10% (data not shown). When determinations of surface area coverage of the same grafts were made by two different observers the calculated mean coverage did not differ significantly (Table II). Effect o f seeded cell concentration on cell graft coverage A linear correlation was found between the concentration of cells in the seeding suspension and the fractional surface area coverage of the grafts (r ~ = 0.93,p < 0.0001)o This correlation applied to seeding concentrations of 7 x 103 c e l l s / c m 2 t o 2 x l0 s cells/cm2. When the seeding ,density was less than 7 x l0 s, there was insignificant graft coverage (Table I). The correlation of seeding density with graft surface coverage obtained with the Optomax corresponded to microscopic and macroscopic observations of the graft surfaces (Fig. 1). Effect o f fibronectin preincubation on seeded cell density Three grafts were seeded with a call concentration of 2 × 105 cells/cm~ without fibronectin preincubation. These three grafts had a fractional surface area coverage of 7.96% + 2.6% (mean + SD). When

Computer analysisofgraft seeding 143 Table I. Relationship of seeded cell density to surface coverage of polytettafluoroethylene grafts Seeded cell concentration (cells/cm2) 2 × 7 x 2 × 7× 2 ×

lO s 104 104 lO s 10 a

No. of grafts 7 4 3 3 3

Percent surface area coverage* 73 ± 35 ± 9 ± 0.6± 0.8 ±

10 6 4 0.2 0.14

*Data presented are the mean ± SD o f determinations o f surface area coverage of the indicated n u m b e r o f grafts. Each determination was performed by the same observer and is based on the sampfing of 15 microscopic fields per graft.

compared to the seven grafts seeded with 2 × 105 cells/cm2 and fibronectin precoating (Table I), a significant difference was found in fractional surface area coverage (p < 0.0001). This difference correlated with microscopic and macroscopic observations of the grafts, as seen in Fig. 2. Detection of genetically labeled calls within a mixed cellular population Grafts were seeded with mixtures of transduced and untransduced cells (as described in "Methods" section), stained with X-gal, and scanned with the Optomax system. Eight seeded graft segments were used. The results (Table III) demonstrate a linear correlation (r ~ = 0.88, p < 0.0003) between the percentage of BAG transduced cells seeded onto the graft and the percentage of surface area covered by transduced cells ("features" according to the Optomax). The scanning protocol was therefore able to quantitate genetically labeled cells within a heterogeneous cellular population. A second, identically designed experiment produced similar statistics correlating the percentage of transduced cells in the seeding inoculum with the percentage graft surface coverage by transduced cells (r 2 -- 0.70,p < 0.006). In this latter experiment, the absolute percentage coverage of the grafts, (i.e., "features", as measured by the Optomax) was far lower than in the first experiment (data not shown). Based on the appearance of the grafts after' staining, the variability in absolute percent coverage was most likely due to differences in the intensity of the X-gal staining reaction on different days. Although in general, day-to-day differences in staining intensity were quite small, this particular experiment suggests that a standard seeded graft may need to be stained along

rouriaaI of VASCULAR SURGERY

144 Newman,Nguyen, and Dichek

Table II. Determination of surface area coverage of seeded polytetrafluoroethylene grafts by two independent observers

Graft no.

Seeding density (cells~era2)

Observer 1

Observer 2

1 2 3 4 5

2 x 10 s 2 x 10 s 7 x 104 7 X 104 2xlO 4

64-+ 9 58_+ 12 26 --2- 10 30 -----13 7--+5

57-+ 12 56+- 14 16_+ 8 26 --+ 12 4_+3

Table III. Detection of genetically labeled cells within a mixed population of labeled and unlabeled cells No. of grafts

BAG transduced cells (% of total)

Percent surface area coverage s

1 2

100 75

2

50

2

25

1

0

42 _+ 9 43 _+ 25 4 4 _+ 17 27 _+ 17 31 _+ 15 6_+4 11_+9 4-+3

Percent area coverage s

*Data presented are the mean _+ SD of 15 microscopic fields per observer.

with experimental grafts to more accurately compare grafts that are stained on different days. DISCUSSION An accurate and easily reproducible method that can specifically identify and quantitate seeded cells over extended time periods is a prerequisite for studies aimed at the optimization of both graft seeding and seeded cell retention in vivo. Although many of the methods currently in use for the quantitation of graft seeding have advantages, none fulfills all of the above requirements. None therefore has yet achieved the status of a clearly superior "gold standard" in the practice of vascular graft seeding. A survey of the most commonly used quantitative methods reveals their strengths and weaknesses. Perhaps the most widely used quantitative technique involves the radiolabeling of endothelial cells with indium 111 (lnln) followed by gamma counting of seeded grafts. This method is both quantitative and reproducible. 8°9The natural decay of 11~In, which has a half-life of 2.8 days, limits the usefulness of this technique for long-term evaluation of seeded cell retention. In addition, a radionuclide scan cannot differentiate between a labeled endothelial cell and radioactive cellular debris within other cells, such as macrophages. Finally questions of safety both for the labeled cells and the whole organism complicate the use of ~HIn in vivo. Direct evaluation of graft surface coverage by means of planimetry has the advantage of directly imaging the luminal graft surface) °'~ However planimettT is tedious and is subject both to sampling error and to observer bias. It is most important to note that this method does not discriminate between seeded cells and host cells that have grown onto the graft. Measurement of thrombus-free graft surface area or platelet deposition are indirect methods of measuring the extent of graft seeding.Z2'~sa5As such, neither of these methods can discriminate between

*Data presented are the mean +_ SD of determinations of surface area coverage, based on the sampling of 15 separate microscopic fields per graft, as described in Methods section.

seeded cells and host cells, and neither allows a direct test of the hypothesis that a decrease in thrombogenicity is a direct result of the continued presence of large numbers of seeded cells. The protocol described in the present study provides solutions to the problems inherent in the aforementioned methods. The protocol can be performed quickly and is standardized to eliminate observer bias and minimize sampling error. No radioactivity is used, and the seeded cells are counted directly rather than having their presence inferred, for example from a lack of thrombus. An additional advantage is that by counting cells in multiple fields of each graft, quantitative information is obtained both on the distribution of cells within a graft (variance among the fields) and the difference in coverage between grafts (comparison between the means of the field counts of separate grafts). The most powerful advantage of the current protocol is its ability to specifically identify seeded versus host cells through the use of a stably integrated genetic marker, the lac-Z gene. For as long as expression of this inserted gene persists, seeded cells or their progeny should be specifically identified by a histochemical stain. Although the work of Wilson et al.z~ indicates that in vivo expression of the lac-Z gene continues for 5 weeks, the stability of expression of the recombinant gene at this and later time points has not yet been unambiguously determined. Therefore although this method may be ideally suited to long-term in vivo studies of graft seeding, further studies in intact animals will be required to confirm this. We tested the abifity of the protocol to quantitate the contributions of seeded cell density and fibronectin precoating to graft surface coverage. The positive correlations we found are not surprising,

Volume 14 Number 2 August 1991

having been described previously by others. 25'26 Nevertheless, we performed these experiments to demonstrate the ease with which our protocol can produce quantitative, statistically meaningful data. O f note, even when graft coverage is complete by both macroscopic and microscopic observation (Fig. 1, A ) , the O p t o m a x scanner reports a value o f 73% (not 100%) surface coverage. The reason for this apparent discrepancy is seen clearly in Fig. 1, B, wherein it is evident that the cells stain centrally, not peripherally. The same central staining is also seen when confluent monolayers o f cells are stained on tissue culture plastic. 19 The O p t o m a x measures only the centrally stained area, not the er~tire cell surface area. For each graft material and cell type it may therefore be necessary to scale the O p t o m a x readings rather than simply use the raw surface area coverage figure. In this manner scales from 0% to 100% coverage (corresponding in this case to O p t o m a x readings from 0% to 73%) can be established. It is also important to emphasize that the present data (Table I, Fig. 1) do not specifically address whether low density seeding might lead to a lower attachment rate, however we d o u b t that this is the case. We should emphasize that the present study is limited to in vitro specimens. The true ability of this quantitative protocol to measure seeded cell retention in grafts that have been placed in vivo remains to be established. Wilson et al.24 demonstrated that lac-Z marked endothelial cells that had been seeded onto Dacron grafts could be visualized as blue spots on the luminal surface o f e x vivo grafts, a result that suggests that our protocol will be applicable to the quantitation of seeded cell retention in ex vivo specimens. Care will be necessary in analyzing ex vivo specimens, however, to minimize false-positive results by taking into account that untransduced cells such as macrophages may, under certain conditions, stain blue with the X-gal staining reagents. 2r A final limitation o f the present study is the high cost o f the O p t o m a x scanning system and associated software. Recent developments in image analysis technology, however, indicate that lower priced systems may be equally effective. 28 In summary, we developed a protocol o f genetic labeling o f seeded endothelial cells combined with computerized scanning and image analysis o f graft surfaces. Although this protocol does require the use of relatively sophisticated gene transfer and imaging technology, it nevertheless posscsses several advantages over currently used quantitative methods and may be uniquely capable o f exploring the long-term relationship between seeded cell retention and graft performance in vivo.

Computeranalysisofgraft seeding

145

The authors thank Dr. W. French Anderson for support and helpful advice. We are grateful to Dr. Una Ryan for endothelial cells, ro Dr. Constance Cepko for the BAG vector, to Dr. Allan Callow for advice concerning cell seeding, and to Dr. Derrick Grant for assistance with the Optomax scanner. REFERENCES

1. Graham LM, BurkelWE, Ford JW, Vintner DW, Kahn RH, Stanley JC. Immediate seeding of enzymatically derived endothelium in Dacron vasculargrafts. Arch Surg 1980;I 15: 289-94. 2. Shindo S, Takagi A, Whittemore AD. Improved patency of collagen-impregnated grafts after in vitro autogenous endothelial cell seeding, l VASCSURGi987;6:325-32. 3. Herring MB, Gardner AL, GloverJ. A single staged technique for seeding vascular grafts with autogenous endothelium. Surgery 1978;84:498-502. 4. Campbell JB, Glover JL, Herring B. The influence of endothelial seeding and platelet inhibition on the patency of ePTFE grafts used to replace small arteries-~an experimental study. Eur I Vase Surg 1988;2:365-70. 5. Boyd KL, Schmidt S, Pippert TR, Hire SA, Sharp WV. The effects of pore size and endothelial cell seeding upon the performance of smaU-diametere-PTFE vascular grafts under controlled flow conditions. I Biomed Mater Res 1988;22: 163-77. 6. Herring MB, Compton RS, Legrand DR, Gardner AL, Madison DL, Glover JL. Endothelial seeding of polytetrafluoroethylene popliteal bypasses-apreliminary report. J VASC SURGI987;6:114-8. 7. Zilla P, Fasol R, Deutsch M, et al. Endothelial seeding of polytetr-,cfluoroethylenevasculargrafts in humans: a preliminary report. J VAse SUV,G I987;6:535-4I. 8. Patterson R_B, Keller JD, Silberstein EB, Kempczinski. A comparison between fibronectin and Matrigel pretreated ePTFE vascular grafts. Ann Vase Surg 1989;3:160-6. 9. GreislerI-IP,Enden EP, KlosakJI, et al. Hemodynamiceffects on endothelial cell monolayer detachment from vascular prostheses. Arch Surg I989;I24:429-33. 10. Herring M, EvansD, BaughmanS, GloverJ. The quantitation of cultured cellular surface coverage: applications for transparent and opaque surfaces. I Biomed Mater Res i984;18: 567-75. I1. Lalka SG, Oelker LM, Malone JM, et al. Acellular vascular matrix: a natural endothelial cell substrate. Ann Vase Surg 1989;3:108-17. i2. McGee GS, Shuman TA, Atldnson JB, Weaver FA, Edwards WI-I. Long-term assessment of a damp-stored, albumincoated, knitted vascular graft. Am Surg I989;55:174-6. 13. Sterpetti AV, Hmlter WI, Schultz RD, et al. Seeding with endothelial cells derived from the microvesselsof the omenturn and from the jugular vein: a comparativestudy. J Vase Sutm 1988;7:677-84. 14. Fasol R, Zilla P, FischMn T, Laufer G, Deutsch M. Surface morphology of circulatingplatelets: a suggestedparameter for the monitoring of endothelial seeded grafts. }"CardiovascSurg 1989;30:398-401. 15. SchneiderPA, Hanson SR, Price TM, Harker LA. Preformed confluent endothelial cell monolayers prevent early platelet deposition on vascular prostheses in baboons. I VAsc SUI