## ABSTRACT

Measurements of gill dimensions have been made for many species of fish (e.g. Byczskowska-Smyk, 1957, 1958, 1959; Gray, 1954; Hughes, 1966; Price, 1931; Saunders, 1962; Hughes & Morgan, 1969) with the purpose of estimating the total area of the respiratory surface. Of these, only Price (1931) and Hughes & Morgan (1969) have made a serious effort to determine the relation between size and area within single species. Price reported that for *Micropterus dolomieu* the total area was related to body weight by the power of 0·78. For the roach and several other species, Hughes and Morgan have found that the relationship is closer to the power of 0·9. For the large number of species ranging from ‘sluggish’ to ‘active’ that he examined, Gray (1954) reported about an order of magnitude of range in the total gill area per gram of fish, but specimen size varied considerably. I. E. Gray (personal communication to G.M.H., 1959) has pointed out that it is important to compare fish of approximately the same size. In his 1954 paper he included *Scomber scombrus* and *Gymnosarda alleterata (= Euthynnus alletteratus)* and has also reported (personal communication to G. M. H.) data on the blackfin tuna and the yellowfin tuna. Except for this preliminary work, few studies appear to have been reported on members of the genus *Thunnus*, which is of considerable interest as it includes some of the largest and most active teleosts.

In these forms it is clearly of interest to determine how the respiratory area compares with that of other fishes. It is also of interest to see whether the study of gill dimensions in these fish can contribute to the understanding of the function of the lamellar and filamentar fusions described by Muir & Kendall (1968). The present report deals with measurements of gill dimensions and estimates of the total gill surface area in *Katsuwonus pelanùs, Thunnus thynnus* and *T. albacares*.

## MATERIALS AND METHODS

Gills from skipjack *(Katsuwonus pelamis)* and yellowfin *(Thunnus albacares)* had been collected and preserved in 10% formalin by the Bureau of Commercial Fisheries in Honolulu; gills from bluefin *(T. thynnus)* were obtained through the kind assistance of the Tuna Corporation, Cambridge, Maryland. The latter fish had been frozen after they were caught, and the gills were removed when the fish were thawed prior to canning. The fish were weighed on a platform scale as soon as they had been thawed. Gills from a few specimens of these bluefin were measured shortly after thawing but most were preserved in 10 % formalin. There had been some tissue breakdown during the freezing and the lamellar fusions appear to have been destroyed. Secondary lamellae appeared to be fairly well preserved, however, and were judged to be suitable for measurements.

*2L/d′*and the surface area of an average secondary lamella

*(bl)*, their product yielding an estimate for the total gill area of the fish

*(L*is the total length (mm.) of all the gill filaments, 1/d′ is the number of secondary lamellae per millimetre on one side of a filament).

In detail, the four arches on one side were dissected out, the filaments were counted and the total filament length was determined. These values were doubled to take account of the arches of the other side. The area of individual secondary lamellae was measured by tracing the image from an A. O. Spencer microscope projection head and determining the area of the tracing with a K and E model 4236 planimeter. The area was doubled to take account of both surfaces of the secondary lamella.

In selecting the secondary lamellae to be measured it is clearly important to take a representative sample from the filaments and secondary lamellae of the whole sieve. The method adopted in the final study of this work was to mark every twentieth filament of the second arch and to remove the 20th, 60th, 100th, 140th, etc., filaments. Three lamellae were then selected at three different levels from each of these filaments. The arch was thus divided into sections of 40 filaments, the central one of each being used for the counts and measurements of the secondary lamellae. The average number of secondary lamellae per millimetre of filament and the average of the areas of the three lamellae from the central filament were taken to represent the entire section of 40 filaments. The sampling of the central filament was justified on the basis of a gradual and progressive change in the lengths of the filaments over each section of 40 filaments. This factor together with variations in the number and size of the secondary lamellae over the length of a single filament (Text-fig. 1) make the whole process of sampling difficult, but checks by various methods suggest that the one adopted here is satisfactory. More detailed controls have been made with trout gills (M. Morgan, unpublished) and have led to the adoption of a similar technique.

Eight to 10 counts were made, over the entire length of the filament, of the number of secondary lamellae falling into the field of an 0·5 mm. eyepiece micrometer. The average of these was taken to represent the spacing for that filament and for the entire section of 40 filaments. The average area of the three selected lamellae was occasionally checked against an average area obtained by more detailed sampling (see Text-fig. 1). The average area of three lamellae was considered a reasonable representation of the actual average for the filament and, therefore, for that particular section of 40 filaments. For each section the total number of secondary lamellae was determined by multiplying the average spacing by the total length of the 40 filaments. These numbers were then summed for all sections and divided by the total filament length of the arch to get an average number of secondary lamellae per millimetre which was weighted by the filament length of each section and which pertained to the whole arch. Similarly, the sum of the total area of the secondary lamellae for all sections was divided by the total number of lamellae, giving an estimate of the average area of a secondary lamella which was weighted by the number in each section. The product of the weighted average number of secondary lamellae per millimetre and the weighted area of an average lamella was multiplied by the total filament length for the whole gill, to estimate the total area.

Two sets of data obtained during preliminary studies made use of a less detailed sampling method for the area of the secondary lamellae, utilizing only three or five filaments; values for the number of secondary lamellae per millimetre and for the lamellar area were simply averaged and these averages were multiplied by the total filament length. The analysis presented below keeps these data separate but, as will become apparent, they conform fairly closely to the main pattern shown by the more detailed method.

In determining the area of a single secondary lamella it was not possible to remove intact lamellae from the filament, so frozen sections (about 300 *μ)* were made perpendicular to the filament axis. Because of the curved face of the secondary lamellae it was difficult to obtain a true projection and some allowance had to be made for this in calculating the true area. The fusions between different parts of the gill system in the tunny also added to the difficulties in this part of the work. After inspecting the effect of the curvature, it was decided that the actual area of each secondary lamella was about 15% greater than that measured and therefore all measurements relating to the area of a secondary lamella were multiplied by 1·15.

## RESULTS

### (1) ‘Homogeneity’ of the gill sieve

At first sight the gills of fishes seem to consist of a fairly constant arrangement repeated over the whole sieve, but more detailed inspection shows that such a view can be misleading. In fact there are great variations not only in the obvious dimensions of filament length but also in more detailed aspects such as the spacing between secondary lamellae, their shape and surface area. In spite of these differences it may well be that the resistance to water flow across the sieve is fairly uniform as suggested by Hughes (1966). From the present point of view this leads to considerable difficulties because the large number of individual secondary lamellae precludes measurement of all of them in any fish. For example, a 1 kg. skipjack has about 5,000,000 secondary lamellae and hence some type of sampling method is necessary. Each worker has usually devised his own method and the variations in technique can lead to wide differences in the values obtained even for the same species.

Some attempt to determine the variations in the dimensions over the whole sieve was made in the present study and is illustrated for a single arch by reference to the results obtained for a 1677 g. skipjack. The medial surface of the second gill arch is shown in Pl. 1, fig. 1, where every 20th filament is marked by a thread. Starting with the 20th, every 40th filament was sampled, and thus for each of these sections the number of secondary lamellae per millimetre of filament and their area (average of three) was determined from the middle filament. These values are shown for each section in Table 1, together with the computations necessary for estimating the total gill area. From this study it is clear that secondary lamellae at the end of each branchial arch are small, but they are more closely spaced. Furthermore, because the filaments are short there are fewer secondary lamellae per section of 40 filaments. The computation given shows how the average area of the secondary lamellae was weighted according to their number in each section, and the average number of secondary lamellae per millimetre was weighted by the total filament length of each section. An unweighted, average lamellar area (average of column C, Table 1) is 0·332, which is quite low compared with 0·428 for the weighted value.

The greatest source of error in the method lies in the selection of the three lamellae to be sampled from each of the selected filaments. The variation in size encountered, and the potential error in sampling the secondary lamellae, is demonstrated in Text-fig. 1. A 30 mm. filament (selected from the 201−240 section) was cut into six pieces, each of 5 mm., and the spacing of the secondary lamellae was counted along the whole length of this filament. The area of the secondary lamellae at the centre of each of the pieces was then determined and results are given in Text-fig. 1. The top panel shows the filament marked at 5 mm. intervals and the arrows indicate the positions from which the three lamellae were selected in the routine analysis. The centre panel shows the area of one surface of the middle secondary lamella of each 5 mm. piece and the lower panel the number of secondary lamellae per millimetre. The sudden decrease in secondary lamellae per millimetre at the tip of the filament was found quite often though not in all cases.

Multiplying the area of the middle lamella of each piece by the number of secondary lamellae in that piece, summing and dividing by the total lamellae on the filament (181·06/960) yields 0·189 mm.^{2} as the weighted average area of a secondary lamella. In comparison, averaging the areas obtained at the levels of the three arrows gives 0·182 mm.^{2}. This magnitude of error is acceptable but the positions from which the three secondary lamellae are sampled is obviously critical. When sampled from the three positions shown here, the results are fairly satisfactory.

For the skipjack, and for the other two species, the average thickness of a secondary lamella is taken to be 10 *μ*. No measurable variation was detected in the value from different locations on a single arch nor from fish of different sizes. This is to be expected, of course, for the thickness of a secondary lamella must be related to the dimensions of the red blood cells. The dimensions of ‘relaxed’ nucleated red blood cells are approximately 11 × 6·6 × 2·5 *μ* but they appear to be quite distensible and to undergo large distortions as they course through the lamellar channels.

### (2) Relationship between body weight and total gill area

Measurements and final calculations for four skipjack, two yellowfin and five bluefin which were determined by the procedure illustrated in Table 1 are given in Table 2. Also shown are the results from eight bluefin from which only three filaments were sampled and six bluefin from which five were sampled. In all these species the measurements and dimensions show trends increasing or decreasing with fish size. Of primary concern is the effect of fish size on the total area of the secondary lamellae, and examination of this trend indicated that a logarithmic transformation would give the best approximation to a straight line. The total gill areas for all fish examined are plotted against weight on log/log co-ordinates (Text-fig. 2). Lines were fitted by least-squares regression to the points, in one case for the four skipjack (equation (1)), and a second line for the two yellowfin and five weighted bluefin (equation (3)). A further line is given which summarizes the results obtained for all the bluefin measurements. The fact that the regression line obtained from all the bluefin data does not differ widely from that of the other two suggests that it might be taken as a good indication of the validity of the mode of analysis. However, because it contains the results obtained by several different methods, it is considered preferable to take the two lines obtained from the results of the final calculations, using weighted averages, to be the most significant for comparison with data from other species.

*A =*total secondary lamellar area in mm.

^{2},

*a*= total secondary lamellar area for a 1 g. fish in mm.

^{2},

*W*= weight of fish in g.,

*b*= regression coefficient (slope). The equations obtained, with the 95 % confidence intervals for the regression coefficients, are:

*n —*2). Examination of Text-fig. 2 shows that, but for one high point for a bluefin, the yellowfin-bluefin line could nearly parallel the line for skipjack. If it were parallel, we would have an estimate of about 4025 mm.

^{2}for the area of a 1 g. fish which may be a better estimate than 3151 from equation (3).

As both the yellowfin and bluefin are members of the same genus *(Thunnus)*, it is reasonable to suppose that they might have approximately the same total gill area for fish of the same size. If this is true, then the similarity in these results may be taken to indicate that the dimensions of the gills were no more seriously altered by freezing of the bluefin than by quick fixation in formalin of the yellowfin. The skipjack were treated in a similar way to the yellowfin, so it seems reasonable to assume that the data for all three species are comparable in spite of differences in their treatment. For a 10 kg. fish the skipjack had a total gill area about 30% greater than that of the other two. If the regression line for skipjack on the one hand and yellowfin and bluefin on the other were parallel, this difference would be the same for all sizes. Based on the regression lines (equations (1) and (3)), however, the difference increases for smaller fish and would be about 70% for a 1 g. fish.

### (3) Relationship between body weight and gill dimensions

The effect of size on a number of dimensions of the gills can be determined from the data summarized in Table 2. We will consider only three components which were used to estimate the total area, i.e. average secondary lamellar area, number of secondary lamellae per millimetre and total filament length. Logarithmic transformations again gave best approximation to a straight line and wet weight, as used above, is a convenient expression of body size.

#### (a) Area of a single secondary lamella

Weighted average lamellar area is plotted against weight on log/log co-ordinates for the eleven fish studied in detail (Text-fig. 3). The values represent the surface area of both sides of the secondary lamellae, being twice that determined from the tracings. The two lines shown were obtained by regression for skipjack (equation (5)) and for yellowfin and bluefin omitting the one high point for bluefin (equation (7)).

(b) *Spacing of the secondary lamellae*

The average number of secondary lamellae per millimetre on one side of a gill filament has been plotted (log/log) against weight of fish for the eleven specimens (Text-fig. 4). The line shown was obtained by re-gression for all the eleven points and therefore represents all three species (equation (11)).

(10) suggest that the number is the same for a 1 g. fish of all three species, but that it decreases more rapidly with size in bluefin and yellowfin than in skipjack. The higher intercept and lower slope for equation (11) is a result of these apparent differences.

#### (c) Total filament length

The total length of all filaments are plotted (log/log) against weight of fish in Text-fig. 5 for all the twenty-six fish given in Table 2. In a few cases the ends of the arches were inadvertently cut off when cannery personnel removed the gills, and the number missing had to be estimated. This accounts for some of the variations seen in the points for bluefin. The lines shown were obtained by regression for skipjack (equation (12)) and for yellowfin and the weighted bluefin (equation (13)).

## DISCUSSION

In spite of differences in the source and treatment of the gills used in this study and in the way in which the areas have been measured, it seems that all data are fairly comparable. This is particularly important because of the small numbers of fish available and also because of difficulties in obtaining exact figures for the gill areas arising from the sampling problems mentioned earlier. The adoption of a statistical method of analysis, together with a small amount of non-statistical interpretation, allows some useful conclusions to be drawn regarding the gill dimensions. Bluefin and yellowfin appear to be quite similar in all of the parameters investigated. Skipjack may have a slightly larger area for an average secondary lamella and they may be more closely spaced. This species certainly seems to have a greater total gill area for a given weight. With the present material and the preservation methods used, it was thought that the figures obtained here might be relatively low, but examination of the gills from a freshly caught bluefin indicated that the measurements had not been detectably altered by preservation.

The total area of the secondary lamellae for each fish was obtained by multiplying the total filament length by the number of secondary lamellae per millimetre and the product by the area of an average lamella. Each of these four parameters has been examined with respect to weight by logarithmic transformations. If such transformation is correct, then the regression coefficients of the three components should sum to the coefficient obtained for the total area. The values obtained are given below:

From equation (1), and interpretation of equation (3), the regression coefficient for the total area is about 0·85. In view of the variation encountered and the small number of specimens used, this good agreement supports the use of the method of logarithmic transformations in this way.

It can be seen that the total gill area increases with body weight to the power of about 0·85. Price (1931) reported a value of 0·78 for the bass and Ursin (1967), analysing 19 ‘intermediate’ species from Gray’s (1954) study, obtained 0·82 for that group. For the roach and some other species, values near to 0·9 have been found (G. M. Hughes, unpublished). In the present study the regression coefficient for all the bluefin data also exceeded 0·9.

Whether these differences are real or not cannot be determined from so few species. It is significant that the relationship of area to weight is similar to that between oxygen consumption and body weight. For oxygen consumption Winberg (1956) derived 0·81 as an average value for a large number of species. Paloheimo & Dickie (1965) suggested that 0·8o characterized most studies. There are well-documented departures from 0·80 (see Brett, 1965) and we might expect species variation in the gill area relationship as well.

Because the total secondary lamellar area increases with weight by a fractional power, the values given for a standard 1 g. fish are larger than the area per gram value often cited by workers in this field. The former statistic is more meaningful when comparing different species, especially when specimen sizes are substantially different. In deriving the total area for such a standard 1 g. fish in large species such as tunny, however, the extrapolation of the regression line is very long and correspondingly involves greater errors. In equation (3), for example, the estimate of the gill area in a 1 g. fish was 3151 but the 95 % confidence limits are approximately 10,000 and 1000. This extrapolation of course also assumes a common growth constant at all sizes during the development of these fish. However, accepting the risk of these long extrapolations, our estimates are compared with those from other species for three sizes in Table 3. The largest figure for the 1 g. size is slightly more than thirteen times the smallest. The regression coefficients are not the same, however, and the values for standard 100 g. and 1000 g. sizes are also shown in Table 3. The ratio of the largest to the smallest at the 1000 g. size is about fourteen. Both Gray (1954) and Hughes (1966) encountered species with gill areas smaller than that of bass at the same weight, so the range for all teleosts studied is even greater than this.

Equations (12)-(14) indicate that the total filament length is about 5600 mm. for a i g. tunny. This is more than seven times the total filament length found in a 1 g. bass by Price. The equation, derived graphically, from data given by Price is approximately log *L* = log 760+0·47 log *W*.

The spacing of the secondary lamellae varies widely in different species and some consideration has been given to its ecological significance (Hughes, 1966; Hughes & Shelton, 1962). For the three species discussed here this analysis suggests that there would be 60 secondary lamellae/mm. on each side of the filament for the 1 g. fish, the number being related to body weight by a power of about –0·08. Examination of the filaments of the first arch of a skipjack which probably weighed 1 g. or less yielded an average of about 55 secondary lamellae per millimetre. At the tip of the filament there were as many as 120/mm. Graphical analysis of Price’s bass data shows that for this species a i g. fish has about 40/mm. and that the number decreases with weight by the power of about –0·1. At the 1000 g. size the tunny has about 34/mm. (weighted average) and bass would have only 18/mm. Judging from Hughes’s (1966) counts, and his publication of Gray’s (1954) counts, most species are distributed above and below the line for bass (305 g. toadfish had 11/mm.) but the tunny appears to have the largest number of all fishes so far investigated for a given weight.

Price (1931) gave no detailed information about the average size of individual secondary lamellae, but an estimate of this can be obtained by dividing his total gill area by the total number of secondary lamellae. The figures obtained suggest that in bass the area of an average secondary lamella is greater than that for a tunny of the same weight. Examination of Hughes’s (1966) measurements for specimens of several species of various weights further indicates that the weighted average size of the secondary lamella in the tunny is small for fish of a given size relative to those of other species. This fits in with the general conclusions (Hughes, 1966) that more active species appear to have smaller secondary lamellae which are closely spaced. The specimens examined here were very large so that the actual areas reported in Table 2 are relatively large also.

The tunny is one of the most active fish and probably represents one extreme of the range of gill dimensions. The smallmouth bass studied by Price (1931) is a smaller fish which is far less active, and the comparison has been informative. Most species fall between the two, with respect to both activity and gill dimensions, but there are others that lie even below the bass in both of these features. With respect to gill dimensions, the tunny (and the bass) follow a growth pattern that may be fairly general among teleosts. For a given size of fish the tunny packs more secondary lamellae of a substantial (though smaller) area into a very large total filament length. Considering the great size attained by the tunny, and the effect of that size on gill dimensions, it is not surprising that special modifications have evolved in the system which appear to strengthen the entire gill meshwork (Muir & Kendall, 1968). However, it does not seem to be entirely a question of size of the sieve as a whole, for these fusions appear to be present at all post-larval stages. This might be taken to suggest that they are functionally associated with size as such but may be necessary during early development from the morphogenetic point of view. As the number of secondary lamellae per millimetre is greatest in the smallest specimens, its importance in keeping the interlamellar channels open may be greater at these stages.

The existence of such an enormous gill area exposed to the water clearly presents other problems from a physiological point of view. For the gills provide a surface for the exchange of both ions and water in addition to that of the respiratory gases. It suggests that the problem of osmoregulation might be difficult for a fish with such a gill system if it were subjected to substantial osmotic stresses, and it is perhaps for this reason that the tunny is limited to a marine environment. Clearly involved in this situation is the whole nature and permeability properties of membranes separating the blood from the water and the nature of the vascularization of the secondary lamellae.

## SUMMARY

Estimates have been made of the total area of the secondary lamellae in the gills of skipjack tuna

*(Katsuwonus pelamis)*, yellowfin tuna*(Thunnus albacares)*, and bluefin tuna*(T. thynnus)*. A sampling method is described which takes into account the variation in size and spacing of the secondary lamellae in different portions of the sieve.Twenty-six specimens in the weight range 1-40 kg. were examined and analysed by logarithmic plots of different gill dimensions against body weight. A good fit was found to the general equation

*A = aW*^{b}.The slope

*(b)*of the regression line for the total area*(A)*against body weight*(W)*was found to be about 0·85 for all three species. This relationship is similar to that (0·85) between oxygen consumption and body weight for a large number of species of teleost fish.The corresponding regression coefficients for the relationships between body size and average area of a secondary lamella, number of secondary lamellae per millimetre and total filament length were +0·53, − 0·08 and +0·38 respectively.

A comparison is made between the three species of tunny and the limited data available for size ranges of other teleosts. On the basis of values obtained by extrapolating the regression lines, it is concluded that the tunny has a larger gill area per unit of body weight than any other fish so far investigated. This is mainly due to the large total length of the gill filaments and the very close spacing (up to 120 per mm. have been measured) of relatively small secondary lamellae.

It is concluded that the extensive gill area of the tunny is related to its very active mode of life.

## ACKNOWLEDGEMENTS

We wish to thank Clinton E. Brown for his encouragement throughout this work and Dr D. C. Wiggert for his help with the computer analysis. This work was made possible as a result of grant NOOO 14−68−C−0326 from the Office of Naval Research to Hydronautics Incorporated.

## REFERENCES

*(Oncorhynchus nerka)*

*J. Fish. Res. Bd Can*

*Pleuronectes platessa*, and perch,

*Perea fluvia tills*

*Zoológica Pol*

*(Anguilla anguilla*L.), the loach

*(Misgumus fossilis*L.) and the perchpike

*(Lucioperca lucioperca*L

*Acta biol. cracov*

*(Tinca tinca*L.), the silver bream

*(Blicca bjoerkna*L.) and the chondrostoma

*(Chondrostoma natus*L

*Acta biol. cracov*

*Biol. Bull. mar. biol. Lab., Woods Hole*

*J. exp. Biol*

*Biol. Rev*. (in preparation)

*Adv. comp. PhysioL Biochem*

*Copeia*

*Proceedings of the Governor’t Conference on Central Pacific Fishing Resources, Honolulu*

*J. Fish. Res. Bd Can*

*Micropterus dolo-mieu*, Lacépède

*Ohio St. Univ. Stud*

*Can. J. Zool*

*J. Fish. Res. Bd Can*

*Rate of Metabolism and Food Requirements of Fishes. (Nauchnye Trudy Belo-, russkogo Gosudarvennogo Universiteta imeni V. I. Lenina, Minsk*

*(Fish. Res. Bd Can*., Trans. Ser. no. 194).

## EXPLANATION OF PLATE

Photograph showing the second gill arch of the left side of a skipjack. Cotton threads have been inserted at the base of every 20th filament as discussed in the text.