METHODS

Provenance and collection of materials

Most of the surveys on extant polycystine radiolarians published to date are based on samples of their skeletons preserved in the surface sediments, rather than on plankton samples. Sediment samples have some advantages over water-column materials, but also several important shortcomings (see Sedimentary vs. water-column materials).

A variety of sediment coring and grabbing devices have been used throughout the years for analyses of the polycystines from the upper centimeters of the sediments (Kennett 1982). Gravity and Kasten corers are among the simplest, permitting retrieval of up to a few meters of sediment at a time from practically any depth. Piston corers have been used widely due to their ability to recover long sedimentary sequences, up to 20-30 m in length. However, all these devices tend to disturb the sediments, especially the uppermost layer which is of particular importance for the analysis of Recent assemblages. Box corers (rectangular, shallow, ca. 1 m coring boxes which ensure complete closure of water-flow passages after sampling and before leaving the seabed, thus minimizing sample washout during ascent) are preferable for retrieval of the top layer of the sediments. However, the fact that box core samples usually lack the thin uppermost phytodetrital film characteristic of most sediments (Billett et al. 1983) suggested that the bow-wave of the device is strong enough as to wash away any mobile particles before hitting the bottom. Multicorers, an arrangement of several short coring tubes mounted on a rigid frame seem to overcome this problem successfully as they have been shown to collect phytodetritus, as well as significantly higher numbers of macrobenthic specimens than box corers (Bett et al. 1994).

Plankton samples for radiolarian studies are usually collected with nets. However, this group, as well as a few other microzooplanktonic taxa, pose serious methodological difficulties. Indeed, they are too small (around 20-30 to 300 µm) to collect effectively with standard zooplanktonic nets (100 to 300 µm in pore size), yet too scarce in most areas to yield adequate catches with water-bottles or low-powered pumps. Thus, fine-meshed nets have to be employed, which significantly complicates not only the concentration of the radiolarians (due to the concomitant retrieval of other organisms, some of which, like the diatoms, cannot be fractioned out later; see Swanberg and Eide 1992), but also because net clogging jeopardizes subsequent estimations of the volume of water filtered (Tranter and Smith 1968; Boltovskoy 1981b). In order to avoid clogging by smaller particles, thus ensuring better estimates of the volume of water filtered and larger sample-sizes, meshes ranging between 60 and up to 100 µm are traditionally used for polycystine studies in the water-column. It should be stressed, however, that both absolute quantitative estimates of radiolarian abundance, and the proportions of at least some species and developmental stages may be seriously biased in these collections: Boltovskoy et al. (1993a) reported that in sediment trap materials from the tropical Atlantic shells below 40-60 µm represent roughly 50% of the overall polycystine fauna.

Estimates of radiolarian abundances in the water-column must be performed with flow-metered nets; clogging of the meshes, in particular of those with small pores, makes assessment of the volume of water filtered based on distance towed and mouth diameter extremely unreliable (Tranter and Smith 1968; Boltovskoy 1981a, b). Thus, whenever unflowmetered nets are employed, such as those derived from Tucker's (1951) opening-closing mechanism (e.g., the Multinet, based on Bé's 1962, design; the MOCNESS, Wiebe et al. 1976; the RMT 1+8, Baker et al. 1973), it is strongly recommended that evaluation of radiolarian concentrations be avoided (species proportions, on the other hand, are in principle unaffected in these samples).

For assessment of the delicate colonial forms, as well as for studies of feeding, growth, metabolism, etc. of live individuals, specimens are collected by divers (e.g., Swanberg 1979), or by means of very short and slow plankton tows, thus ensuring a better preservation of the protists (Matsuoka 1992).

Sediment trap techniques have undergone major improvements in the last years, thus constituting a very useful tool for the collection of polycystine materials (US GOFS 1989; Lange and Boltovskoy 1995). Simple sediment traps consist of a concentrating cone or funnel which tapers into a collecting jar; the array, which can have either one or several traps, is moored to the bottom or drifts with the current suspended from a buoy at the surface. Time-series models are deployed at different oceanic locations for periods up to a year or more, and are provided with a mechanism which replaces the collecting cup at predetermined intervals thus yielding a detailed record of the changes in the amount and type of flux throughout several seasons (Honjo and Doherty 1988; Lange and Boltovskoy 1995).

Sediment-trap materials have some important advantages over planktonic collections. Sample-size is usually much larger in sediment traps than in plankton nets, with fluxes as high as 200,000 shells/m²/day having been recorded in the equatorial Atlantic (Boltovskoy et al. 1996; see also table 3 in Boltovskoy et al. 1993a). Seasonal plankton collections are composed of a sequence of snapshots which represent but an insignificant proportion of the total time elapsed between tows, and may therefore not only under- or overestimate mean protist abundances (e.g., Bé et al. 1985), but also yield "atypical" specific assemblages. Time-series sediment trap samples, on the other hand, integrate over preselected depth and time ranges, thus averaging the overlying plankton over restricted periods which yield adequate chronological resolution to allow pinpointing the relative importance of limited offsets of the yearly cycle. Furthermore, since seasonal variations in total mass flux are usually closely coupled with primary production in the upper mixed layer (Honjo et al. 1982, 1988; Deuser et al. 1983, 1990; Wefer 1989), comparison of total flux vs. radiolarian numbers and specific makeup can furnish first hand information on indicators (and paleoindicators) of the biological productivity of the associated water masses.

Sediment trap materials, however, also have some shortcomings. Because of limitations associated with the hydrodynamic properties of particle accumulation in the traps, these devices are most effective when deployed at depths in excess of 500-700 m (US GOFS 1989; Lange and Boltovskoy 1995). As a result, they integrate the flux from several biologically dissimilar layers (e.g., Kling 1979; Kling and Boltovskoy 1995). Furthermore, sinking skeletons intercepted at these depths may not adequately reflect their standing stocks at the surface, nor their specific composition. Boltovskoy and Alder (1992) concluded that, in the Weddell Sea, over 90% of the polycystines that inhabit the upper 400 m are destroyed (probably due to fragmentation by grazing) before reaching 400-900 m of depth. Subsurface advection of shells produced at higher latitudes and integration of low protist abundances over large depth intervals may be responsible for the fact that, in the eastern equatorial Atlantic, polycystine assemblage compositions recorded in plankton samples at 0-300 m are totally different from those recovered in traps at 800-2000 m (Boltovskoy et al. 1995; Figure 4A, 4B).

It should be borne in mind that the yields of sediment trap samples are not amenable to direct comparisons with those of plankton samples: while the former are an expression of the downward flux, which in turn is associated with productivity and preservation, quantitative plankton samples give information on standing stock only. Hence, compositional differences may not only reflect advection, destruction by grazing, etc., but also biological traits of the species considered. Thus, a scarce species with high reproduction, mortality and output rates may be rare in the plankton but abundant in the underlying sediment trap (Kling and Boltovskoy 1995; Boltovskoy et al. 1995).

As with other zooplanktonic groups, analyses of radiolarian vertical distribution patterns are usually performed with the aid of vertically stratified plankton tows (e.g., Renz 1976; Kling 1979; Dworetzky and Morley 1987; Kling and Boltovskoy 1995; Abelmann and Gowing 1997). However, because their identification is based on the siliceous skeleton which preserves after the death of the cell, in order to discriminate live vs. dead protists in the subsurface layers the cytoplasm is often stained with Rose Bengal, Sudan black B, or eosin (Petrushevskaya 1971b; Swanberg and Bjørklund 1986; Abelmann and Gowing 1997). Although this technique can furnish some clues on the living depth ranges of the species, it does not provide unequivocal information because of uncertainties associated with the speed of decomposition of the protists' cytoplasm. Boltovskoy and Lena (1970), for example, concluded that specimens of several planktonic foraminifera still contained protoplasm in their shell 98 days after death. Bernhard (1988) compared estimates of the proportions of presumably live benthic Foraminifera as indicated by Rose Bengal and Sudan black B staining and by ATP assay, concluding that stained protoplasm was present in individuals up to four weeks after actual death of the cell. These lapses are significantly longer than the time it takes a radiolarian shell to reach the sea-floor (Takahashi and Honjo 1983).

Unless special cytological studies are required (e.g., Petrushevskaya 1986), plankton and sediment trap samples can be preserved in 4-5% formaldehyde; the addition of picric acid to the solution enhances the preservation of the colonies, yet acidification should be avoided if the calcareous plankton is to be saved from dissolution.

Sample preparation and analysis

The following section offers some general comments on the preparation of whole samples for routine counting and identification procedures. It does not review the methods involved in special cytological and ultrastructural studies (see Anderson 1983a, for a review of these topics), as well as those used for detailed taxonomic work, which can involve thin-sectioning, etching and polishing, etc. (Riedel and Sanfilippo 1977; Boltovskoy et al. 1983; Petrushevskaya 1986).

Pelagic surface sediments are usually clean enough as to require little treatment before preparation of the slides. Elimination of the organic matter and disaggregation of the materials is achieved by boiling the sample (5-10 g) for a few minutes in a beaker with water to which hydrogen peroxyde (10%, 300 ml per liter) and tetrasodium pyrophosphate (10 g per liter) have been added. Disaggregation, cleaning and removal of clay coatings and infilling particles can be aided by treating the sample in a gentle ultrasonic bath. For further disaggregation of heavily indurated sediments various products, such as kerosene, paint thinner, or ammonia can be helpful (the sediment is dried, soaked in the solvent, and then immersed in water, upon which disaggregation usually occurs rapidly). If calcareous material is abundant it can be removed with a few drops of hydrochloric acid (after eliminating the hydrogen peroxyde by wet-sieving). The resulting clean material is then sieved with abundant water in order to eliminate the reagents and smaller particles. The mesh size used depends on the aims of the study; most surveys routinely employ 40-60 µm-meshes, yet these, as described above, miss many of the smaller species, as well as most developing forms. If precise abundance estimates are sought, mesh openings around 15 to 20 µm should be employed, although these will retain large numbers of unidentifiable skeletal fragments, as well as non-radiolarian material (especially diatoms), which can make subsequent observation more laborious. The clean residue in the sieve is pipetted onto glass microscope slides, dried, and soaked with a few drops of xylene; before the xylene has evaporated the mounting medium is added and covered with a cover glass. Canada Balsam is most often used for these preparations, although it takes longer to harden than some other synthetic materials, commercially known as Norland, Pleurax, Hyrax or Depex.

Moore (1973) proposed a convenient method which allows quantification of the number of radiolarian shells per unit weight of sediment. Before processing as described above, the sample is dried and weighed. This weighed sediment is then cleaned and sieved, and all the resulting residue is poured into a large (e.g., 5 l) beaker full of distilled water, on the bottom of which one or two cover gasses have been positioned. The water with the sediment in the beaker is then thoroughly stirred (avoiding rotational motion, which will result in centrifugal fractionation) for achieving a random distribution of the particles, and the sediment is allowed to settle. With the aid of a siphon all but 30-50 mm of water are removed, and the remainder is evaporated with an overhead infrared lamp. When the surface of the cover glasses is dry they are removed from the beaker and mounted as described above. The slide thus prepared will contain a fraction of the radiolarian shells present in the original sample, this fraction being equivalent to the proportion that the surface of the cover glass makes of that of the surface of the bottom of the beaker.

Preparation of plankton and sediment trap samples is somewhat more labororious due to the large amounts of organic material they contain. When both absolute radiolarian concentrations and specific inventories are sought, it is recommended that counting be performed separately from the identifications. Polycystines can be counted (although not identified) in whole, unprocessed samples in counting chambers under the inverted microscope (Hasle 1978; Boltovskoy 1981c; Villafañe and Reid 1995). Subsequently, either the entire sample or a subsample can be treated in order to eliminate all organic matter leaving the clean siliceous skeletons that will be mounted as described above for sedimentary materials. It should be born in mind, however, that radiolarian cells are often very difficult to recognize in preserved, unprocessed plankton samples. The siliceous skeleton, usually the most conspicuous distinguishing feature, is obscured by the cytoplasm to such an extent that radiolarians are easily confused with other planktonic protists, fecal pellets, eggs, various organic aggregates, debris, etc. Adding a few drops of hydrogen peroxide and/or hydrochloric acid, which slowly digest the organic matter, and comparing the dubious particles before and after treatment can greatly help to pinpoint radiolarian cells (Alder, personal commun. 1997).

Several different methods have been used for eliminating organic material from water-column samples, including high- and low-temperature ashing, oxidizing with hydrogen peroxide and/or ultraviolet light, etc. (see review in Boltovskoy et al. 1983). One of the most widespread, however, is that proposed by Simonsen (1974) for cleaning diatom frustules. The plankton sample is rinsed with abundant fresh water (wet-sieving), and placed in a beaker to which an equal volume of saturated KMnO₄ is added; it is then left for 24 hs. A volume of concentrated HCl equivalent to that already contained in the beaker is subsequently added to the sample; the dark brown liquid is gently heated until it becomes transparent or light yellow. Once the sample has cooled, it is sieved again thoroughly with fresh water and rinsed with distilled water. The residue is pipetted onto microscope glass slides as described above.

Analysis of the specimens is best performed in mounted slides, which by transparency permits observing the internal structures (such as medullary shells, spiral structures, etc.), and the wall-thickness. In addition, slight variations in the depth of field allow one to determine whether a shell of circular outline is a disc (in which case most of the surface is in focus simultaneously), or a sphere (either the central part or the periphery are in focus). Photographs taken in the light microscope have the advantage of being readily comparable to mounted specimens. The scanning electron microscope (SEM), on the other hand, is especially suitable for analyzing the surface morphology, but only in specimens with large openings in the outermost shell, or in those partially broken, can internal structures be observed. SEM photographs produce very appealing results, but their comparison with routine collections mounted in slides is tricky (see Figure 5A, 5B, 5C). Ideally, both techniques should complement each other (Boltovskoy et al. 1983, described a method which allows performing light and SEM observations and photographs of the same radiolarian specimens).

Assessment of radiolarian species-specific absolute and relative abundances are based on identifications and counts. Since any given slide often contains thousands of polycystine shells, the researcher is forced to decide how many specimens should be identified and counted in order to achieve an adequate estimate of overall numbers and species proportions. Several methods have been proposed for the assessment of bias in sample-based particle counts (see reviews in Venrick 1978, 1995; Frontier 1981), and in the appraisal of species proportions (Patterson and Fishbein 1989; Buzas 1990). Patterson and Fishbein (1989) concluded that for species representing >50% of the overall taxocoenosis at least 50 specimens should be counted in order to achieve reliable percentage data, 300 counts for species which comprise approximately 10% of a sample, 500-1000 counts for species that make up 5%, and counts of several thousands for those that comprise 1%. Unfortunately, in the case of the polycystines these efforts are unrealistic because in any given sample containing 100-150 species only one-three are above 10%, and 70-90 occur at levels below 1% (see "Geographic and vertical distribution"). In terms of the amount of information attained, it is more profitable to analyze more samples at a lower resolution, than to examine fewer sites at these statistically more reliable levels. Thus, in practice proportions are estimated in bulk, regardless of the individual species abundances, usually scanning 300-600 specimens per sample. It is common practice to identify the first 300-600 individuals on the slide, and then check the rest of the slide or slides for the given sample in order to account for the rarer taxa. The relative abundances of the latter are estimated approximately, and they are usually excluded from subsequent general numerical analyses (e.g., multivariate techniques, such as cluster and factor analysis) because of the uncertainties associated with their assumed absences. It should be stressed, however, that the counting effort necessary for reliable estimates of the fractional abundance of the rare species is inversely proportional to the equitability of the assemblage. Thus, when the sample is strongly dominated by a single or only a few taxa, such as in polar areas, chances of recording the rare polycystines in random sequential counts are low because the observer repeatedly hits the dominant species. On the contrary, as equitability increases so does the probability of logging a so far unrecorded species with every new specimen scanned.