Journal of Neuroscience Methods', 35 (1990) 63-73 63 Elsevier

NSM 01140

An assessment of the validity of densitometric measures of striatal tyrosine hydroxylase-positive fibers" relationship

to apomorphine-induced rotations in 6-hydroxydopamine lesioned rats

Robert E. Burke, Jean Lud Cadet, Jeffrey D. Kent, Andrew L. Karanas and Vernice Jackson-Lewis

Department of Neurology, Columbia University, 710 West 168th Street New York, N Y I O03 2 (U.S.A.)

(Received 30 November 1989) (Revised version received 25 April 1990)

(Accepted 8 June 1990)

Key words: Quantitative immunocytochemistry; Image analysis; Densitometry; Striatum; Tyrosine hy- droxylase; Rotational behavior

The power of immunohistochemical staining as a tool for the study of the neurochemical ana tomy of the brain would be greatly enhanced if quantitative measures of staining were to be developed. We have here assessed the reliability and validity of two population measures of extent of fiber innervation: percent area occupied by staining, and average optical density (AOD) of staining. We have evaluated these measures for tyrosine hydroxylase-positive staining of the str iatum in relation to apomorphine-induced rotational behavior in 6-hydroxydopamine lesioned rats. We have found that inter-operator reliability for the area measure is high (r = 0.98). Apomorphine-induced rotations were observed when the area measured was reduced to 2% or less of the control side, and when the density measure was reduced to 15% or less. These results are similar to those obtained previously for biochemical assay of TH activity, which showed rotations at reductions to 10% or less. We conclude that these density measures provide valid relative indices of extent of fiber innervation on the same section. The A O D measure appears to be more sensitive at lower levels of innervation.

Introduction

Immunohistochernistry is a powerful technique for the study of the cellular and subcellular distri- bution of antigens, either native or foreign, in tissue. Immunohistochemical demonstration of the regional distribution of endogenous antigens re- lated to neural transmission has been a major source of new information about the neurochem-

Correspondence." R.E. Burke, MD, Dept. of Neurology, Col- umbia University, 710 West 168th Street New York, NY10032, U.S.A. Tel.: (212) 305-7374.

ical anatomy of the central nervous system. Up to now, the technique has largely been used to make qualitative observations. The specificity and reso- lution of the technique have attracted efforts to enhance its power by developing quantitative mea- sures of immunohistochemical staining. These ef- forts have been abetted by the recent development of optical systems and personal computer-based software to quantitatively analyze autoradiograms generated by the in vitro binding of radiolabeled ligands to receptors in intact tissue sections (Young and Kuhar, 1979), or by the in vivo uptake of radiolabeled deoxyglucose into neural structures (Sokoloff et al., 1977). Efforts to develop quantita-

0165-0270/90/$03.50 ¢~ 1990 Elsevier Science Publishers B.V. (Biomedical Division)

64

tire measures of immunohistochemical staining have been directed at the molecular level, to quantitate the amount of antigen present per cell (Benno et al., 1982a, b; Shivers et al., 1983), at the cellular level to quantitate numbers of cells or fibers (Mize and Payne, 1987; Mize et al., 1988) and at the population level to measure mean opti- cal density per field, or area of staining per field (Cassell et al., 1982; Briski et al., 1983, Agnati et al., 1984a, b, c; Baker et al., 1985; Gross and Rothfield, 1985; Mize et al, 1988).

All of these efforts require validation. We have here undertaken to assess the inter-operator relia- bility and the validity of two population measures of innervation density, mean optical density per area and the percentage of structure area stained, for tyrosine hydroxylase (TH) staining of the striatum. We have limited our assessment to these measures on same section by comparing TH-posi- tive fiber staining of the striatum on one side to the other side in 6-hydroxydopamine (6-OHDA)- lesioned animals. Comparing these measures on the same section internally controls for variations in fixation, section thickness, degree of specific chromagen staining and background. We chose this immunostaining paradigm because staining with this antibody is robust (Pickel et al., 1975) and well-characterized (Benno et al., 1982a, b). This lesion paradigm has also been well-char- acterized in terms of the biochemical changes in striatal dopaminergic markers relative to the ap- pearance of apomorphine-induced circling behav- ior (Hef t ie t al., 1980).

To validate image analysis techniques at the molecular level, the determination of the density of staining must be compared to immunologic or biochemical measures of the amount of antigen present in tissue. At the cellular level, determina- tion of number of neurons must be compared to traditional and laborious, but standardized, tech- niques of neuron counting (Abercrombie, 1946; Konigsmark, 1970). Validation of methods of fiber counting, however, either at the individual or population level, is problematic, because it is not clear that there exists an absolute or even stan- dardized measure of fiber number. What is gener- ally considered to be the closest to an absolute measure of fiber number in tissue are methods

based on visualization of fibers at high magnifica- tion and then quantitation either by total fiber length in a field (Soghomonian et al, 1987: Mize et al., 1988) or by fiber intersections with a line or grid (Shivers et al., 1983; Ojima et al., 1988). While these measures are useful and have the advantage that they depend on literal visualization of the structures being quantified, they are not absolute measures of the number of fibers within a given volume of tissue. First, it is difficult to determine to what extent all "stainable" fibers have in fact been stained. Impediments to staining within a section include varying degrees of fixa- tion and fiber preservation, and varying degrees of reagent penetration. Second, fibers which are not in the plane of section are not counted or mea- sured. Attempts to correct for fiber angulation would be based on the assumption that fibers are randomly oriented, which may not be the case. Third, length measurements and intersection counts do not take fiber tortuosity into account. Highly tortuous, meandering fibers would theoret- ically be measured as "more fibers" than an equal number of relatively straight fibers by either of these methods. Thus, while these measures of fiber density are useful, in the absence of a truly ab- solute measure they are best considered as relative measures of the actual number of fibers present. We have therefore undertaken to assess validity of population-based measures of fiber density not only by inspection of the number of individual fibers at high power, but also by relating them to a functional indicator of innervation, i.e.. apomor- phine-induced rotations. We hypothesized that if these population measures are valid indices of striatal dopaminergic innervation, then it should be possible to meaningfully relate them to rota- tional behavior, a valid functional index of in- nervation.

Methods

To induce unilateral substantia nigra (SN) le- sions in male Sprague-Dawley rats (I50-200 g), they were first pre-treated with desipramine 25 m g /k g i.p., and then anesthetized with chloral hydrate 400 m g / k g i.p. They were then placed in

a Kopf stereotactic frame, and 6-OHDA (8 t~g in 4 /xl of normal saline/0.2% ascorbate) was in- jected by pressure over 4 min at AP +2.4 mm, Lat + 1.6 mm, DV - 26 mm (interaural line zero), corresponding to the left anteromedial SN. After a 1-min wait the needle was withdrawn.

Three weeks after induction of the SN lesion, the rats were injected with apomorphine (0.5 mg/kg i.p.). Each rat was then placed in a plexig- lass cylinder, and rotations were counted over the following 60 rain by a counter which was fixed to the lid of the cylinder and attached by a tether to a harness on the rat.

From 2 to 4 weeks after the determination of rotational behavior, rats were killed and processed for immunohistochemical staining of TH. Each rat was anesthetized with pentobarbital 60 mg /kg i.p., and then perfused intracardially with 0.9% saline (containing 5 units heparin/ml) at 4°C by gravity for 5 min, followed by perfusion with 4% paraformaldehyde (PF)/0 . 1 M phosphate buffer (pH 7.1) at 4°C for 15 min. The brain was then removed and a tissue block containing the stria- tum was cut in a rat brain matrix (a plexiglass block containing a well, crossed every 2 mm by slots for the insertion of razor blades in coronal planes). The coronal tissue block was taken per- pendicular to the long axis of the brain to ensure symmetrical sections. The tissue was post-fixed in 4% PF/0.1 M PB at 4°C overnight, and it was then mounted to a cryostat chuck by rapid freez- ing with powdered dry ice. Thirty /~m sections

65

were processed free-floating. They were washed twice in 0.1 M phosphate buffer (pH 7.1)/0.9% saline (PBS), and then treated with 0.1 M PBS/0.5% bovine serum albumen (BSA) for 15 min at 4°C. They were then treated with 0.1 M PBS/0.5% BSA/0.1% Triton X-100 for 15 min at 4°C. Following a rinse in PBS, sections were in- cubated with primary antibody (a gift of Dr. Tong Joh) with demonstrated monospecificity for TH (Joh et al., 1973) at 1:2500 at 4°C for 48 h. In preliminary studies we found that the densitomet- ric measure of area stained, as defined below, did not change up to a 10000:1 dilution of primary antibody (Table I). Sections were then washed in 0.1 M PBS/0.5% BSA twice, and then incubated with biotinylated protein A at 1 : 200 for 60 min at room temperature. Sections were again washed and incubated in avidin-biotinylated horseradish peroxidase (" Elite Kit", Vector Labs) per the sup- plier's instructions for 60 min at room tempera- ture.

Our approach to optimizing conditions of di- aminobenzidine and H202 incubation was based on two considerations~ (1) we wished to optimize the quality of staining of anatomical structures for purposes of quantification; and (2) we acknowl- edged that all sections from all animals could not be identical in terms of fixation, thickness or other factors which affect specific and background staining. Therefore, rather than using a prede- termined protocol for all material, we used fixed concentrations of DAB and H202, but varied the

TABLE 1

THE EFFECT OF D I L U T I O N OF P R IM AR Y ANT IB ODY ON STRIATAL TH DENSITY A N D AREA MEASURES

Sections from the same rat were incubated at the primary antibody dilutions shown. Sections treated with higher dilutions were incubated for longer times in DAB, to optimize specific staining (see Methods). While the mean striatal OD decreased with greater dilution, the percent of the striatum stained positive (as defined in Methods) did not vary greatly, up to 1 : 10000. At 1 : 20000 specific staining was so faint that OD B >> 25% specific max imum OD (ODMax), making background values fall within the window of density defined as specific. For this reason area of specific staining could not be determined (ND)

Background OD Lesioned side Control side ODB/ODM~ ~

(OD~) Mean OD %Striatum Mean OD %Striatum ratio (%) (corpus callosum) str iatum stained str iatum stained

TH 1 : 2500 0.04 0.04 0 0.11 99 20 TH 1 : 5000 0.05 0.04 0 0.09 95 29 TH 1 : 10000 0.04 0.04 0 0.08 97 25 TH 1 : 20000 0.04 0.03 0 0.06 N D 36

66

time of incubation according to visualization of the ongoing reaction under a dissecting micro- scope. Gross and Rothfield (1985) quantitatively demonstrated what many have observed visually, that within a certain period of incubation, specific staining densities saturate, and beyond that point there is a progressive increase in background staining. We would stop the DAB reaction when specific staining appeared optimal, and back- ground began to increase; this was at 9-12 min for these sections. Sections were incubated in di- aminobenzidine tetrahydrochloride (DAB) (Al- drich) 50 mg/100 ml 0.1 M Tris (pH 7.6) in the presence of glucose oxidase (Sigma) 0.9 mg/100 ml, D-( + )-glucose 200 mg/100 ml and ammonium chloride 40 mg/100 ml to generate H202 (Lundquist and Josefsson, 1971). This concentra- tion of DAB was chosen because Benno et al. (1982) had previously shown that, using this anti- body, the peroxidase reaction saturates between 1.0 and 2.5 mM, and because Gross and Rothfield (1985) had demonstrated, in another system, that DAB 0.05% was an optimal concentration for varying concentrations of H202. The latter authors had shown that non-specific background increases rapidly at higher concentrations. The glucose- oxidase method for H202 production was chosen primarily because this method (performed with aliquoted, frozen reagents) provides a reproduci- ble, fresh, continuous supply of H 202. H 202 stored as a liquid has variable rates of decomposition depending on storage and frequency of use. The concentration of H202 generated using these re- agents was determined by HPLC with electro- chemical detection using a platinum electrode, and found to be 0.001% after 30 min. After the DAB reaction, sections were mounted to gelatin-coated slides, dried, dehydrated in ascending concentra- tions of ethanol, cleared in xylene, and cover- slipped in Permount.

Densitometric analysis of stained sections was performed on an Amersham RAS-3000 image analysis system using Loats Associates software (RA-3000, Version 2.1). Prior to application of population-based densitometric analysis of fiber innervation to these sections, all were examined at 1000 x to insure that fibers were individually de- fined at a microscopic level. The RAS system

utilizes a Dage-MTI series 68 Newvicon video camera. The camera and the digitizing circuitry support 256 gray levels image acquisition for a 256 x 256 pixel area. A black level (zero light) reference built into the camera permits continual adjustment of camera electronics to control signal drift as a function of temperature. At the begin- ning of each session, camera gain was set to make the illumination source equal to 256 transmission units (OD = 0.0). All images were digitized as an average of 16 scans using the Loats Average Sub- routine, to control for transient fluctuations in lighting and camera response. The illumination source image was digitized, and correction for non-uniformity of the source was made by Loats software on subsequent images. Although the sys- tem was allowed to warm-up 30 rain prior to use, the illumination source was always re-digitized prior to analyzing a new section to ensure that the correction for non-uniformity was based on the brightness of the source at that point in time. In addition, camera gain was re-checked. After set- ting camera gain and non-uniformity correction, an experimental section was digitized. On these sections, we then read background density of staining over the external capsule right and left. We proceeded with analysis only if right and left background were almost the same; we had noted that occasionally sections fold in the mid-line dur- ing free-floating, and a significant difference in background staining between the two sides may result. In addition, we proceeded with a section only if OD over white matter (background) was 25% or less of OD over specifically stained struc- tures. In pilot studies, we had demonstrated in controls without primary antibody that back- ground over the striatum was the same on experi- mental and control sides. Following digitization of the section image, we measured the maximum OD of TH staining in the section. Maximum OD for the digitized image was determined by Loats software, but in all cases we ascertained that this reading was derived from a region of specific TH staining. If it was not (e.g., if it was an edge artifact), then maximum OD was measured by hand. We then defined " T H staining" as a window of density values between maximum OD and 50% of that value. A value of 50% was chosen so that

all values defined as stained would be at least two-fold above background, thus ensuring that false positive staining would be minimal. Using Loats software, the staining with OD values be- tween maximum and 50% of maximum could be assigned a single color. Using a hand-held cursor we then outlined the striatum on the experimental and control sides. It was then possible to de- termine the percent of total striatal area occupied by " T H staining". For evaluation of inter-rater reliability, these percent areas for single striatal planes were compared. To compare whole-stria- turn percent area to rotational behavior, we mea- sured the percent area on sections representative of 4 striatal planes from each brain (Paxinos and Watson 10.2, 9.7, 9.2, 8.2), expressed experimental as percent of control for each plane, and then took an average of these 4 measurements to derive a single measure of T H innervation density for each brain. The variability in the percent area measure among the 4 striatal planes was small (mean standard error 3.4%) in comparison to the dif- ference between control and experimental sides (mean 65.2%; see Results).

We also measured the average optical density (AOD) over the striatum on experimental and control sides as a measure of TH-positive innerva- tion density. For each section AOD over each striatum was determined by outlining the striatum with a cursor: AOD for the striatum was then determined by Loats software. We then subtracted background OD from each striatal AOD to obtain "specific striatal AOD". Background was defined as the OD over the corpus callosum, where no TH-positive fibers were apparent. For each sec- tion, the experimental side was expressed as a percent of control. For each brain, this determina- tion was performed in 4 striatal planes, and then an average taken to derive a single percent of control AOD measure of TH innervation for each brain. As for the area measure, the variability in specific striatal AOD among the 4 striatal planes was small in comparison to the difference between Control and Experimental sides.

We did not manipulate our primary digitized image of the brain sections, i.e., we did not, for example, perform image smoothing, erosion, dila- tation, or skeletonization; nor did we remove

67

artifacts from the image. In general, we chose sections which were free of tears, large dust par- ticles over the striatum, folds, and so on, which would complicate straight forward image analysis.

Results

Figs. 1 and 2 demonstrate in pseudocolor " T H staining", as we have defined it for two 6-OHDA- lesioned rats, one with rotations (Fig. 1) and one without (Fig. 2). For both rats, Paxinos and Wat- son plane 9.2 is shown. In this plane, rat 12 (with rotations) had 1.5% of the area of the lesioned striatal plane occupied by TH staining. Rat 1 (without rotations) had 10% of the area of the lesioned striatal plane occupied by TH staining. The percent of area of the whole striatum on the lesioned side, normalized as a percent of innerva- tion of the control side, and averaged for 4 planes (Paxinos and Watson 10.2, 9.7, 9.2, 8.2), was 1.4 and 12.2% respectively.

Regions of the lesioned striatum in rat 1 are shown at the individual fiber level in Fig. 3. Fig. 3A shows a ventromedial portion of the striatum which has been defined as positive for fiber stain- ing in Fig. 2. It can be seen that many fibers occupy this region. Fig. 3B shows a lateral portion of the lesioned striatum, which has been defined as negative for fiber staining. Although a few individual fibers are visible in the micrograph, there are not enough to sufficiently stain this region above background levels.

Our protocol for defining positively stained striatal area was quite reproducible for an individ- ual operator, performing the determinations days apart. Intra-operator area determinations corre- lated at an r = 0.99 ( P < 0.01) for n = 36 striatal measurements (data not shown). Inter-operator reliability was also high. Fig. 4 demonstrates the correlation between percent striatal TH-positive area determinations for 2 operators performed on 20 striata. Most determinations correlated well except in a single instance operator 1 found 58.5% of a striatal plane stained, and operator 2 found 24.5% stained. This particular section contained an edge artifact which stained more densely than the TH immunostained regions. Thus the maximal

68

Fig. 1. Pseudocolor image of TH-positive fiber staining in a rat (R12) with a unilateral 6 -OHDA lesion of the substatltia nigra, with rotations following apomorphine injection. TH-positive fiber staining has been defined as all staining between OD values of 0.24

(maximal TH staining) to 0.12 (50% of 0.24). TH staining is colored red; it occupies 1.5% of the area of the lesioned striatum.

Fig. 2. Pseudocolor image of TH-positive staining in a rat (R1) with a 6 -OHDA lesion, without apomorphine-induced rotations. TH-positive staining has been defined as in Fig. 1 ; in this particular section the window for defining TH staining was similar to that for the section in Fig. l (maximum fiber OD = 0.23. min imum fiber OD = 0.I1). Fiber staining occupies 10% of the inJured striatum.

69

Fig. 3. Darkfield photomicrograph of TH-positive fibers in the lesioned striatum of rat 1, shown in Fig. 2. A: ventromedial striatum defined as positive for staining by image analysis. Numerous fibers are observed; 4 have been marked by sets of arrows. The reticulated haze is due to fibers out of the plane of focus. B: lateral striatum defined as negative for staining by image analysis.

Although occasional fibers are observed, as shown, there are not enough to stain this region above background levels. Bar = 10/tin.

O D for the section (which is automatical ly de- termined by Loats software) could not be consid- ered as the maximum O D of specific staining. Each operator had to measure maximal specific T H staining from the section. This process led the operators to define slightly different windows for fiber staining (a difference of 0.01 O D unit in the min imum density), resulting in different estimates of the percent of the striatum stained. This occur- rence indicates that analysis may be enhanced by editing out of the image densely stained artifacts.

The relationship between percent of whole striatal area occupied by T H staining and rota- tions is shown in Fig. 5. It can be seen that the relationship between rotations and % striatal TH- positive area appears to be discontinuous. At area values f rom 12 to 90%, no rotations were ob- served. At low values ( < 2%), rotat ions occur in all animals. Thus there appears to be a min imum

area stained threshold between 2 and 12%, less than which rotat ions appear. A similar threshold relationship was observed using A O D over the striatum, as shown in Fig. 6. Rotat ions were not observed at percent A O D values greater than 25%. Between 0 and 15% A O D values, rotat ions were in all animals, but there was no clear relationship between percent reduct ion A O D and number of rotations. These results are similar to those ob- tained by Hefti and co-workers for T H activity measured by radiochemical assay (Fig. 5 of Hefti et al., 1980); apomorphine- induced rotations were not observed until experimental T H activity fell below a threshold of 10% of control. Below that value, there was no inverse relationship between percent reduct ion and the number of rotations.

Fig. 7 shows the relationship between the area and density measures. Once a threshold for the area measurement is achieved, there is a close

70

o D._

100

90

80

70

60

50

40

30

20 10

i

r = 0.98 • p < 01

2~0 3~0 4~0 ,50 6J0 7J0 8L0 9LO 1(30 Per Cent TH-PosJtive Area

(Operator 1 ) Fig. 4. Relationship between percent of striatal area occupied by TH-positive staining measured by 2 operators. There is a good correlation ( r = 0.98, P < 0.01) between the 2 sets of results, indicating that inter-operator reliability is good. A single outlying point lying below the others is discussed in the

text.

linear relationship between the two measures. We interpret the y-intercept of 16 (% Density) on this graph to mean that below a certain level of stain- ing (50% of maximum), the area measure defines staining as absent and the area of striatal staining

200 Contralateral 150

100! 501

Rotations 0 t o • i i 7~c~ ~ i 10 20 30 40 510 6J0 ,~ 80 9t0 100 50

Ipsilateral 1 O0

t 50 200

% Striatal TH-Positive Area (Ipsilateral / Contralateral)

Fig. 5. Relationship between percent of striatal area occupied by TH staining and apomorphine-induced rotations. The ab- scissa represents percent of whole striatal area that is T H positive, i.e. an average of % ipsilateral/% contralateral for sections representative of 4 planes (Paxinos and Watson 10.2, 9.7, 9.2 and 8.2). Rotations were observed when percent area

fell below 2% on the lesioned side.

Contralateral

Rotations

Ipsilateral

200 150 t00 50

o 1'0 50

100

150

20O

2'0 %' 4'0 30 sb 6'0 7'0"+'0 9'0 40

% Reduction in Stdatal AOD (Ipsilateral I Contralateral)

Fig. 6. Relationship between density of striatal TH-positive staining and apomorphine-induced rotations. The abscissa rep- resents the A O D on the lesioned s i d e / A O D on the control side, averaged for 4 planes, and expressed as a percentage. Rotations were observed when A O D on the experimental side

fell below 15% of the A O D on the control side.

as zero. The AOD measure, however, does not impose a defined cut-off for staining and remains a sensitive measure until the number of fibers is so few as to actually reach background levels. Thus

10090 • •

80

70 o c)

.=>'_~ 60

~ 5o ._ ~ n=6

~ ~ 4O ×

~ 3O

201

10

1'0 210 310 4'0 -50 610 ¢0 8'0 910 180 % Area

(Experimental/Control)

Fig. 7. Relationship between percent of striatal TH-positive area and percent AOD (experimental/control) . Above a threshold of detection by the area measure, the two measures correlate closely (r = 0.998 for n = 6 points on the line). How- ever, A O D values 16% or less do not meet the min imum threshold necessary for definition as ' T H positive' on the area measure, so all of these low-staining striata are zero on the area

measure.

AOD, as a continuous measure, is more sensitive to lower levels of staining.

Discussion

Both area of staining and AOD measures of fiber number have been employed at the popula- tion level by other investigators (Cassel et al., 1982; Briski et al., 1983; Agnati et al., 1984c), but they have not been extensively evaluated in terms of their reliability, validity or relation to one another. We have found that the intra- and inter- operator reliability of the area of innervation mea- sure is good. We have assessed the validity of these two measures by comparing results obtained with them to a functional index of striatal dopaminergic innervation, i.e., apomorphine-in- duced rotations. This index has, in turn, been studied in relationship to a biochemical index of striatal innervation, i.e., TH activity. We have found that both the area measure and the density measure relate to rotational behavior like the bio- chemical measure; rotations are not observed until they are reduced to thresholds of at least 2 and 15% of control levels, respectively. Below the threshold values, there was not a quantitative rela- tionship between rotations and either morphologic measure. This was also true of reductions in TH activity (Hefti et al., 1980).

We have found that the percent area measure, as a discontinuous measure, is not sensitive to changes at low levels of fiber innervation, i.e., levels below the defined threshold for positive fiber staining. Beyond the threshold of detection, however, the area measure correlated closely with the density measure.

Based on our results, we anticipate the AOD measure to be more widely applicable. It is more sensitive at lower levels of innervation, and, in addition, it would be sensitive to experimental manipulations which increase density over control. The area of innervation measure, however, achieves a maximum when the striatum becomes fully oc- cupied by fibers staining above the defined threshold level, i.e., 100% area. Experiments which would produce such a situation could include those which induce striatal shrinkage, with an increase

71

in the density of stained fibers. In some investiga- tions, however, the area measure may be more closely related to the aspect of interest. In a trans- plant study, for example, the area of a structure re-innervated by the transplant may be of greater interest than optical density.

As a practical issue, although we have used here the Loats software to automatically de- termine maximal OD, we recognize that in many instances it will not be possible to do so, either because of dense artifacts on the section, or be- cause the investigator chooses to focus on a region of specific staining where the maximum is less than the specific staining in another region of the section. In these instances, the operator must manually determine maximum OD. When this needs to be done, there must be a well-defined procedure for performance of the maximum OD measurement, or significant intra-operator varia- bility may result, as we observed on one of our sections.

The population measure approach has several advantages. Large regions of the structure of inter- est are represented in the assessment. Individual fiber counting methods must be performed at high power, and a small volume is visualized per analy- sis field. For example, the Nikon CFN Plan Apochromat lens (100 x , oil, numerical aperture 1.25) examines a field containing a 100-t~m 2 square, with a 0.3-~m photomicrographic depth of field, giving a 3 x 10-6 mm 3 volume of tissue in focus, which is 1 x 10-s % of the volume of adult rat striatum. Data from such small volumes is subject to sampling errors, particularly in anatom- ically heterogeneous structures, or following ex- perimental manipulation (e.g., lesions). In ad- dition, population measures can be applied to thick, unembedded sections which contain too many fibers to count individually (Burke and Karanas, 1990). The disadvantage of the popula- tion approach, using a segmented field analysis, is that because fibers are not resolved at an individ- ual level, there is the possibility of including false-positive staining in the assessment. For ex- ample, edge artifacts on blood vessel walls may meet the threshold criteria for "specific staining" and add to both area of staining and AOD mea- surements. In addition, extracellular localization

72

of antigen, positively stained, may also affect either measurement. When measured differences are large, these effects may be minor, but caution must be exercised when small differences are studied.

One must also recognize the limits of the type of data we have shown here. The raw density measures we have performed provide at most rela- tive data about the extent of innervation on the same section. It would not be meaningful to inter- pret these measures as interval or ratio data which can be compared among sections from different animals; a,specific density reading of 0.10 units on one section cannot, for example, be interpreted as twice as much fiber staining as a reading of 0.05 on another section. Thus, we do not assume there is a linear relationship between density and the number of fibers, in contradistinction to the ap- proach of Benno et al. (1982) who at tempted to develop a measure of T H molecules which could be compared between sections, and which there- fore required a strictly linear relationship between density and TH content. We assume only that there is a monotonic relationship between density and fiber number. We consider such an assump- tion justified,based on empirical comparisons be- tween individual fiber number at the microscopic level (as shown in Fig. 3) and relative density measures. We recognize that it is possible that in some experimental settings our assumption may not be justified. For example, density of staining in a particular brain region may change not be- cause of a change in fiber number, but rather a change in caliber of fibers or amount of antigen present per fiber. Nevertheless, in the particular paradigm used here, our assumption seems to have permitted implementation of a meaningful mea- sure of striatal TH-posit ive fiber innervation. The only data which we have used here to compare among animals is data that has been normalized (for O t h e r area or density) in reference to the control side of each brain.

Ultimately, the validity of these densitometric approaches lies in the fact that they capture what the eye can see. Their power, however, lies in their ability to assign an ordinal numerical value to what is seen, Such an assignment provides objec- tivity and documentation, Furthermore, when such

data is normalized for an internal control, numeri- cal comparisons can be made among sections and animals. It then becomes possible to submit these comparisons to quantitative and statistical analy- sis, which is not possible with images observed by eye alone.

Acknowledgements

R.E.B. is supported by N I N D S TIDA No. 1 K07 NS00746, R29 NS26836, the Dystonia Medi- cal Research Foundat ion and the Parkinson's Dis- ease Foundation. J D K was supported by a Re- search Fellowship under an Institutional Research Service Award to Columbia University ( N I H grant No. 5P35 NL07616-09). We are grateful to Dr. Andrew Towle of Eugenetech International, and Mr. John Kinney of Analytical Systems Consult- ing, Inc., for their constructive criticism of the manuscript. We are also grateful to Ms, Monica Fey Hat ten for typing the manuscript.

References

Abercrombie, M. (1946) Estimation of nuclear population from microtome sections. Anat. Rec., 94: 239-247.

Agnati, L.F., Fuxe, K., Benfenati, F., Zini, I., Zoli, M., Fabbri, L. and Harfstrand, A. (1984a) Computer-assisted mor- phometry and microdensitometry of transmitter-identified neurons with special reference to the mesostriatal dopamine pathway. I. Methodological aspects. Acta Physiol. Scand. (Suppl. 532), 122: 5-36.

Agnati, L.F., Fuxe, K., Goldstein, L.C.M., Toffano, G., Giardino, L. and Zoli, M. (1984b) Computer-assisted mor- phometry and microdensitometry of transmitter-identified neurons with special reference to the mesostriatal dopamine pathway. II. Further studies on the effects of the GM 1 ganglioside on the degenerative and regenerative features of mesostriatal dopamine neurons. Acta Physiol. Sand. (Suppl. 532), 122: 37-44.

Agnati, L.F., Fuxe, K., Benfenati, F., Toffano, G., Cimino, M., Battistini, N., Calza, L. and Pich, E.M. (1984c) Computer- assisted morphometry and microdensitometry of trans- mitter-identif'u~d neurons with special reference to the mesostriatai dopamine pathway. III. Studies on the aging process. Acta Physiol. Scand. (Suppl 532), 122: 45-66.

Baker, H., Sved, A.F., Tucker, L.W., Alden, S.M. and Reis, D.J. (1985) Strain differences in pituitary prolactin content: relationship to number of hypothalamic dopamine neurons. Brain Res., 358: 16-26.

Benno, R.H., Tucker, L.W., Joh, T.H. and Reis, D.J. (1982a) Quantitative immunocytochemistry of tyrosine hydroxylase in rat brain. I. Development of a computer assisted method using the peroxidase-antiperoxidase technique. Brain Res, 246: 255-236.

Benno, R.H., Tucker, L.W., Joh, T.H. and Reis, D.J. (1982b) Quantitative immunocytochemistry of tyrosine hydroxylase in rat brain. II. Variations in the amount of tyrosine hydroxylase among individual neurons of the locus coeruleus in relationship to neuronal morphology and topography. Brain Res., 246: 237-247.

Briski, K.P., Baker, B.L. and Christensen, A.K. (1983) Effect of ovariectomy on the hypothalamic content of immunoreac- tive gonadotropin-releasing hormone in the female mouse as revealed by quantitative immunocytochemistry and ra- dioimmunoassay, Am. J. Anat., 166: 187-208.

Burke, R.E. and Karanas, A.L. (1990) Demonstration of a medial to lateral gradient in the density of cholinergic neuropil in the rat striatum. Neurosci. Lett., 108:58 64.

Cassell, M.D., Mankovich, N.J., Gray, T.S. and Williams, T.H. (1982) Computer-assisted image analysis of the distribu- tions of peptidergic terminals in the central nucleus of the amygdala: a preliminary study. Peptides, 3: 283-290.

Gross, D.S. and Rothfeld, J.M. (1985) Quantitative im- munocytochemistry of hypothalamic and pituitary hor- mones: validation of an automated, computerized image analysis system. J. Histochem. Cytochem., 33: 11-20.

Hefti, F., Melamed, E. and Wurtman, R.J. (1980) Partial lesions of the dopaminergic nigrostriatal system in rat brain: biochemical characterization. Brain Res., 195:123 137.

Joh, T.H., Geghman, C. and Reis, D. (1974) Immunochemical demonstration of increased accumulation of tyrosine hy- droxylase protein in sympathetic ganglia and adrenal medulla elicited by reserpine. Proc. Natl. Acad. Sci. USA, 70:2767 2771.

Konigsmark, B.W. (1970) Methods for the counting of neu-

73

rons. In: W.J.H. Nauta and S.O.E. Ebbeson (Eds.), Con- temporary Research Methods in Neuroanatomy, Springer- Verlag, New York, pp. 315-380.

Lundquist, I. and Josefsson, J. (1971) Sensitive method for determination of peroxidase activity in tissue by means of coupled oxidation reaction. Anal. Biochem., 41: 567-577.

Mize, R.R. and Payne, M.P. (1987) The innervation density of serotonergic (5-HT) fibers varies in different subdivisions of the cat lateral geniculate nucleus complex. Neurosci. Lett., 82:133 139.

Mize, R.R., Holdefer, R.N. and Nabors, L.B, (1988) Quantita- tive immunocytochemistry using an image analyzer. I. Hardware evaluation, image processing, and data analysis. J. Neurosci. Methods, 26: 1-24.

Ojima, H., Yamasaki, T., Kojima, H. and Akashi, A. (1988) Quantitative analysis of cholinergic fibers in the rat cerebral cortex. Abstr. Soc. Neurosci., 14: 679.

Pickel, V.M., Joh, T.H. and Reis, D.J. (1975) Ultrastructural localization of tyrosine hydroxylase in noradrenergic neu- rons of brain. Proc. Natl. Acad. Sci. USA, 72: 659-663.

Soghomonian. J.J., Doucet, G. and Descarries, U (1987) Serotonin innervation in adult rat neostriatum. 1, Quanti- fied regional distribution. Brain Res., 425: 85-100.

Sokoloff, L., Reivich, M., Kennedy, C., Des Rosiers, M.H., Patlak, C.S., Pettigrew, K.D., Sakurada, O. and Shinohara, M. (1977) The [lac]deoxyglucose method for the measure- ment of local cerebral glucose utilization: theory, proce- dure, and normal values in the conscious and anesthetized albino rat. J. Neurochem., 28:897 916.

Shivers, B.D., Harlan, R.E., Morrell, J.l. and Pfaff, D.W. (1983) lmmunocytochemical localization of luteinizing hormone-releasing hormone in male and female rat brains, Neuroendocrinology, 36:1 12.

Young, W.S. and Kuhar, M.J. (1979) A new method for receptor autoradiography: [3H]opioid receptors in rat brain. Brain Res., 179:255 270.