Comparative Quantitative Analysis of Macrophage Populations Defined by CD68 and Carbohydrate Antigens in Normal and Pathologically Altered Human Liver Tissue

Liver macrophages, which are involved in the different types of hepatitis, may indirectly induce hepatic fibrogenesis, since they have the possibility to activate hepatic stellate cells and fibroblasts by secretion of TGF-β , TNF-α and IL-1. To evaluate variations of the number of liver macrophages and their subpopulations, a quantification was carried out in normal human liver tissue, fatty liver, fatty liver hepatitis and hepatitis B. Identification was performed by the mab PG-M1 (anti-CD68) and, comparatively, four lectins, Griffonia simplicifolia agglutinin I (GSA-I), Erythrina cristagalli agglutinin (ECA), peanut agglutinin (PNA) and soybean agglutinin (SBA). A slight decrease in the frequency of macrophages in pericentral fields was observable in fatty liver and fatty liver hepatitis as compared to normal liver tissue. On the other hand, the number of CD68 cells was significantly enhanced in hepatitis B with moderate and severe inflammatory activity. The highest incidence of macrophages was found in portal tracts of liver with fatty liver hepatitis and, particularly, hepatitis B. The fraction of cells stained by ECA, PNA or SBA did not increase significantly under pathological conditions. In contrast, the percentage of GSA-I binding macrophages was higher in liver parenchyma of hepatitis B and in portal tract macrophages in fatty liver hepatitis and also hepatitis B. In conclusion, our results indicate that GSA-I may aid in the detection of the subpopulation of activated macrophages which are assumed to play a pivotal role in liver pathology.


Introduction
Human macrophages are an important part of the functional network of the immune system and are obviously divided into subpopulations representing different functional and activation states. Macrophages can easily and most specifically be identified by the monoclonal antibody (mab) PG-M1 [4], which is directed against a well-characterized mucin-type lysosomal glycoprotein, the CD68 antigen [9]. In liver pathology, ample evidence has been provided for an indirect role of macrophages in the development of fibrosis. For example, Kupffer cells secrete factors like TGF-β, TNF-α and IL-1, which activate hepatic stellate (fat-storing) cells and fibroblasts [1,3,5,6,[14][15][16]26]. In a biliary fibrosis rat-model, a correlation of the numbers of hepatic stellate cells, Kupffer cells and the density of connective tissue was observed [7]. However, little information exists so far about the incidence of CD68 + macrophages in normal and diseased liver tissue.
In the past years, renewed interest has been focused on liver macrophages which were characterized with respect to their expression of carbohydrate antigens. Ulex europaeus agglutinin I (UEA-I) and soybean agglutinin (SBA) were found to stain Kupffer cells in a fraction of liver specimens, whereas peanut agglutinin (PNA) was non-reactive with normal Kupffer cells, but showed a binding to portal macrophages and Kupffer cells in primary biliary cirrhosis, large duct obstruction and sclerosing cholangitis [20]. Erythrina cristagalli agglutinin (ECA) was also revealed to react with liver macrophages [23]. A specific staining of Kupffer cells and of interlobular vessels and bile duct structures by PNA, SBA, BPA (Bauhinia purpurea agglutinin) and GSA-I (Griffonia simplicifolia agglutinin I) was described by another group [17]. In contrast to these results, Yamamoto et al. [27] observed a reactivity of Kupffer cells in normal liver tissue with Concanavalin A, Ricinus communis agglutinin and wheat germ agglutinin, but not with PNA, SBA, UEA-I and DBA (Dolichos biflorus agglutinin). Additionally, in viral hepatitis, PNA displayed a specific binding to Kupffer cells. In a quantitative analysis of the relative frequency (percent area) of PNA positive Kupffer cells, an increase was disclosed ranging from chronic persistent to chronic aggressive hepatitis [27]. A comparison between the binding of CD68 specific mab and lectins to liver macrophages seems interesting, since the latter may detect different macrophage populations. In a recent study [2] we were able to establish that only a subpopulation of CD68 + human bone macrophages expressed certain carbohydrate antigens detected by lectins like Griffonia simplicifolia agglutinin I isotype B4 (GSA-I-B4), soybean agglutinin (SBA) and Erythrina cristagalli agglutinin (ECA) as well as Le a blood group antigen. In this context, it is noteworthy that numerous in vitro investigations using mouse and rat macrophages indicate that these macrophage subpopulations may represent activated cells [10,12,13,18,21,24,25].
In the present study, serial sections of specimens from normal liver, fatty liver, fatty liver hepatitis and different grades of chronic hepatitis B were investigated. The total population of liver macrophages was visualized by the mab PG-M1 [4] and lectin histochemistry was performed for the identification and characterization of macrophage subpopulations.

Patients
Tissues from liver biopsies were derived from the files including 15 specimens from normal livers. The following non-viral liver diseases were entered into this study: fatty liver (n = 15), fatty liver hepatitis due to drug or alcohol toxicity (n = 15). Specimens from 28 patients known to be infected with hepatitis B virus (HBV) were scored histologically according to the classification of the International Association for the Study of the Liver (IASL): (1) no or minimal histological alteration (n = 7); (2) mild or moderate (score 2 or 3) inflammatory activity, fibrosis grade 2 or 3 (n = 7); (3) moderate inflammatory activity (score 2 or 3), cirrhosis (n = 7); (4) severe inflammatory activity (score 4), cirrhosis (n = 7).

Immunohistochemistry
Tissue specimens were fixed in 5% PBS-buffered formalin and embedded in paraffin. Five-micrometer thick sections were deparaffinized according to routine histological technique. It was established in preliminary experiments that proteolytic pretreatment of sections increased the reactivity of certain lectins, especially GSA-I and PNA [11]. Therefore, incubation with pronase E from Streptomyces griseus (1 mg/ml; Serva, Heidelberg, Germany) was performed for 30 min at 37 • C. Monoclonal antibody PG-M1 as well as lectins were incubated overnight in a humid chamber at 4 • C. The biotin-streptavidin-alkaline-phosphatase method was applied as described in detail elsewhere [2].

Immunohistochemical double-labelling experiments
Since it cannot be generally excluded that other cells stained by lectins may not be distinguished morphologically, we performed a double-labelling immunohistochemistry in order to determine the percentage of lectin-binding cells which are not PG-M1 * . The following incubation steps were performed: (1) biotinylated GSA-I (100 µg/ml), 60 min; All incubation steps were carried out at room temperature. After every step -with the exception of step (4) -three-fold washing with TBS was performed. Vector kits were applied according to the manufacturer's description. As a control, steps (1) and (2) were replaced by steps (5)-(9) and vice versa, using PG-M1 or GSA-I, respectively, in both staining steps.

Morphometry and statistical evaluation
Generally, macrophages were counted per square millimetre (mm 2 ) at a magnification of 400× using a grid ocular. Pericentral and portal fields were evaluated separately. Macrophages in portal fields from normal livers, fatty livers and HBV-infected livers without inflammatory activity or cirrhosis were counted per square millimeter using a manual optic planimeter (MPO-A-MO1-Kontron) involving a standard program set (Kontron software), as described earlier [2]. Only cells containing a nucleus stained by haematoxylin were considered. Endothelial cells as well as bile duct cells reactive with some lectins were excluded. Counting was performed without knowledge of the diagnosis of the specimen by two observers independently. The difference of their results was never greater than 10%, and the mean value was used. The statistical significance of differences between two groups of data was calculated using the Mann-Whitney U -test.

Quantification of CD68 positive macrophages
In normal liver, Kupffer cells were distinctively stained by PG-M1 (155 ± 24 cells/mm 2 ) ( Fig. 1(a)), but also showed a certain reactivity with all lectins under study. In specimens of fatty liver (Table 2), the number of macrophages was decreased (129 ± 27), possibly due to the enhanced volume of fatstoring hepatocytes. On the opposite, the number of macrophages stained by PG-M1 in the pericentral fields of fatty liver hepatitis was generally higher than that in fatty liver, but did not exceed the value of the normal liver. For the details of morphometric analysis, see Table 2. Pericentral fields of biopsies derived from patients suffering from chronic HBV infection but without accompanying florid hepatitis revealed a number of PG-M1 stained macrophages (159 ± 29/mm 2 ) which was similar to normal liver tissue. In comparison, HBV-infected livers showing moderate inflammatory activity not accompanied by cirrhosis (Table 2) exhibited a significantly (p < 0.005) elevated number (285 ± 63/mm 2 ) of PG-M1 + macrophages in the pericentral fields. However, the frequency was lower in pericentral fields of livers with moderate inflammatory activity and cirrhosis. Incidence increased again significantly (p < 0.01) in cases with severe inflammation (Fig. 1(d); Table 2). In the portal tracts of fatty liver hepatitis (Table 3), the lymphohistiocytic infiltration was associated with a high number of macrophages compared to normal and fatty livers as well as HBV-infected livers without inflammatory activity. A significantly (p < 0.05) higher amount of reactive macrophages per mm 2 (1173 ± 373), however, was observed in the portal tracts of chronic hepatitis B exhibiting moderate inflammatory activity without cirrhosis (Table 3). Accompanying the development of cirrhosis, the number of macrophages in the portal tracts decreased to slightly lower values which was possibly due to the space occupated by connective tissue.

Quantification of lectin-binding cells
The number of lectin-binding cells was correlated with the number of PG-M1 + macrophages (Table 4). Our results show that this relation is not constant in normal and pathologically altered liver tissue. However, the number of cells stained by the various lectins never exceeded the number of PG-M1 + cells. The relative number of ECA + , PNA + and SBA + cells in fatty liver and fatty liver hepatitis did not change remarkably compared to normal tissue. In HBV-infected livers, ECA, PNA and SBA identified an equal or even higher percentage of macrophages in the pericentral fields without or with minimal histological changes compared to those exhibiting moderate or severe inflammation. Moreover, the percentage of ECA and PNA binding cells in the portal tracts in fatty liver hepatitis is lower than in the pericentral fields. Altogether, the portion of ECA + and PNA + cells in relation to CD68 + macrophages was not increased in hepatitis B specimens containing inflammation and/or cirrhosis compared to normal livers, fatty liver and also fatty liver hepatitis.

Percentage of GSA-I positive macrophages
In comparison with normal liver tissue, the portion of GSA-I + cells in the pericentral fields increased significantly (p < 0.05) during the evolution of fatty liver. An even higher value was encountered in Table 4 Means of numbers of lectin-positive cells expressed as percentage of the mean number of CD68 + macrophages   ECA  GSA-I  PNA  SBA   PC  POR  PC  POR  PC  POR  PC  POR   Normal liver  83  64  21  29  69  61  70  76  Fatty liver  80  73  29  54  63  60  72  66  Fatty liver hepatitis  71  63  32  36  56  44  68  65  HBV: no inflammation, no cirrhosis  82  72  33  56  53  57  68  the pericentral fields and portal tracts of specimens with fatty liver hepatitis (Table 3). Contrasting PG-M1 and ECA, which stained nearly equal numbers of cells in normal and HBV-infected livers without relevant pathological changes, GSA-I reacted with 56 ± 16 cells per mm 2 in the pericentral fields of the latter specimens compared to 30 ± 11 cells in normal liver biopsies (p < 0.05). In the portal tracts of fatty liver, fatty liver hepatitis and HBV-infected liver, the number of GSA-I + macrophages was increased (Table 3). In summary, the portion of GSA-I + cells showed an increase in all forms of inflammatory liver disease as compared to normal liver tissue (Table 4).

Control: Immunohistochemical double-labelling
Double-labelling control experiments using PG-M1 and GSA-I were performed staining 3 specimens representing each diagnosis included in the study. The percentage of red cells, i.e., cells without double-staining by PG-M1, which resulted in a blue-red mixture or pure blue staining, did not exceed 7%. The control sections did not show a blue staining reaction, indicating that cross-reactivities of Vector blue with antigens bound in step (1), did not contribute to the Vector red staining reaction.

Macrophages in liver diseases
In the present study, the number of macrophages as detected by the CD68 specific mab PG-M1 in normal liver parenchyma and various liver diseases was evaluated. Consequently, our findings are in agreement with recently published data on other liver cell populations like α-SMA + Ito cells [19]. An increase of PG-M1 + macrophages could be particularly observed in HBV-induced hepatitis with moderate and severe inflammatory activity. The highest frequencies of macrophages, however, were obtained in the portal tracts of livers exhibiting fatty liver hepatitis or HBV infection. Using an experimental rat model, a correlation between the numbers of hepatic stellate as well as Kupffer cells with evolution of fibrosis following bile duct ligation has been observed [7,8].

Characterization of liver macrophages by CD68 antigen and lectin-binding sites
Comparatively, four lectins (GSA-I, ECA, PNA, SBA) reported previously [17,20,23,27] to identify liver macrophages were investigated. However, these stained varying numbers of cells, which were always lower than the number of PG-M1 + macrophages. Therefore, it is reasonable to assume that all of them detect only subpopulations of liver macrophages. On the other hand, it cannot be definitively ruled out by light microscopical immunohistochemistry that all cells stained by these lectins are really macrophages. Whereas endothelial and bile duct cells may easily be excluded from counting. It may be diffcult to discriminate between myofibroblastic or hepatic stellate cells. For this reason, lectins are a less reliable tool for the identification of liver macrophages compared to a CD68-specific mab. Furthermore, our results are not in agreement with the hypothesis of Yamamoto et al. [27], who regarded PNA to be a marker of activated Kupffer cells, since it did not stain Kupffer cells in normal liver parenchyma. However, in their study, the number of PNA + cells was not counted but quantified by image analysis as percent area of the parenchyma. Inherent to this method, enlargement of cells and variations of staining intensity may significantly influence the results and therefore a number of caveats have to be taken into account when interpretating the findings of these authors.

GSA-I binding sites on activated macrophages
Ample evidence has shown that the mab PG-M1 directed against the CD68 antigen can be regarded as the most specific tool for the visualization of liver macrophages. The use of three lectins in this study (ECA, PNA, SBA) does not result in relevant information additional to that obtained by PG-M1 staining. However, the reactivity pattern of GSA-I lectin suggests that not only the number of macrophages, but also the percentage of the activated subfraction may increase in the course of inflammatory activity as well as with periportal fibrogenesis. As reported in several experimental studies, GSA-I characterizes a carbohydrate antigen associated with macrophage activation [10,13,18,21,24,25]. By double-labelling immunohistochemistry could be ruled out that a significant portion of GSA-I + cells in human liver do not belong to the CD68 + macrophage lineage. The percentage of GSA-I + cells related to the number of PG-M1 + macrophages was significantly increased in the pericentral fields in inflammatory liver disease as compared to normal livers. A high portion of GSA-I + cells was also observed in the portal tracts under these conditions, which may also be regarded as activated macrophages. Interestingly, in vivo studies [22] have shown that binding of GSA-I-B4 to peritoneal mice macrophages induces the secretion of TNF-α, a mediator which is involved in hepatic fibrogenesis like TGF-β and IL-1 [1,3,5,6,[14][15][16]26]. Considering these findings and the results of the present study it is tempting to speculate that not only the absolute number of Kupffer cells, but their expression of α-D-galactose containing glycoproteins are part of the complex functional cellular network inducing hepatic fibrogenesis. For this reason, our results are in keeping with the assumption that the GSA-I + activated macrophage subpopulation plays a key role in liver pathology.