Species were selected according to two criteria: abundances in the samples and depth habitat variety according to previous studies. Normal forms of Orbulina universa (NCBI:txid46134), Globigerinoides ruber albus (NCBI:txid2606480) sensu stricto morphotype, Globigerinoides ruber rosea (NCBI:txid2606481), Trilobatus trilobus (WoRMS:lsid1027267), Trilobatus sacculifer (WoRMS:lsid1026286), Globorotalia truncatulinoides (NCBI:txid69029), and Neogloboquadrina dutertrei (NCBI:txid51039) were picked from the 300–355 µm sieve fraction, without taking into account test coiling direction. However, the G. ruber albus morphotypes were carefully distinguished since they are known to have different habitats (Aurahs et al., 2011), with G. elongatus precipitating tests in colder surface waters than G. ruber albus sensu stricto (Steinke et al., 2005; Antonarakou et al., 2015). From every sample between 30 to 50 specimens of each species were picked and weighed in a pre-weighed aluminum carrier using a Sartorius CP2P microbalance with a precision of ±1 μg. Average sieve-based shell weights (SBW) were calculated by dividing the recorded mass by the total number of weighed specimens. Subsequently, the specimens were weighed in batches of 5 individuals in order to estimate the standard shell mass deviations. For comparisons shell weight analyses were performed on a narrow size fraction that is commonly used in paleoceanography. Performing weight analyses on a narrow foraminifera size fraction constrains the ontogenic stage of the specimens to a certain number of chambers, and thus minimizes size-related weight differences (e.g. Broecker and Clark, 2001; Barker et al., 2004; Beer et al., 2010). The analytical error, estimated by triplicate measurements of 50 random specimens, ranged from 0.42 to 0.58 μg, which is in accordance with the error of the balance.

To better interpret the changes in foraminifera test weights, for each sample a number of specimens used for the above weight measurements were analyzed using computed microtomography (μCT). More specifically, 10 specimens of the surface dwelling, spinose species G. ruber (white) and 10 specimens of the deep dwelling, non-spinose G. truncatulinoides were scanned using a GE/Phoenix v|tome|x s 240 CT scanner. In addition, 2 more species (spinose T. trilobus and non-spinose N. dutertrei) were scanned from sample 8 (RC24-10), where the lowest shell weights were recorded. The tests were fixed in a customized cylindrical container as described in Zarkogiannis et al. (2020a). Embedded in the base of the container was a calcite microcrystal, which was used to standardize the computed tomography (CT) number of the scanned specimens. A high-resolution setting (voltage of 80 kV, current 80 μA, detector array size of 1024×1024, 1501 projections/360°, 2.5 s/projection) enabled the acquisition of 3D images with an isotropic pixel size of about 1.2 μm. Phoenix datos|x 2.0 software was used to correct and reconstruct tomographic data, that uses the general principle of Feldkamp cone beam algorithm to reconstruct image cross sections from the filtered back projections. XMCT not only allows the examination of the internal structure (or internal degradation) of the calcite mass, providing a means to directly assess shell disintegration due to dissolution but it also offers valuable 3D information about test biometry and thus foraminifera physiology. This 3D information together with the test mass weight measurements allowed for the first time the determination of volume normalized shell weight or foraminifera shell bulk density (see section 2.2.2).

Foraminifera may have different means to increase their positive buoyancy (osmoregulation, lipid or metabolic gases production and storage) but biomineralization is their only mechanism to actively add negative buoyancy. Thus, the amount of heavy calcite mineral secretion determines the deepest point that foraminifera can penetrate into the water column when alive. At their optimum depth of immersion, foraminifera attain an equilibrium position in the water column. Since they are non-motile plankton, in order to occupy certain optimum depths, foraminifera should adjust their cell density to match that of the surrounding fluid at a certain point along the pycnocline. Should this not be the case then displacement occurs until the body and medium densities equilibrate. At the equilibrium point the weight force of the living foraminifer (W_foram) is equal to the weight of the liquid that it displaces (i.e. buoyancy force, B) or since both of these forces depend on the volume of the living foraminifera V_foram and the gravitational acceleration (g) the density of the whole foraminifera shell ρ _foram is equal to the density of the seawater ρ _sw, and ρ _foram is the sum of shell bulk density ρ _shell and cytoplasm density ρ _cyt. Thus:

It is possible to reconstruct the cytoplasmic density ρ _cyt if ρ _sw and ρ _shell are known. Here an attempt was made to calculate foraminifera protoplasm densities for the two different species that were X-ray scanned. Test weights of G. ruber and G. truncatulinoides were converted to shell bulk densities ρ _shell using the XMCT spatial data and ρ _sw at each species apparent living depth was calculated by fitting geochemically reconstructed calcification temperatures/depths from the literature to the Argo potential density profile data. For G. truncatulinoides ambient seawater densities were extracted at the calcification depths reported by Cléroux et al. (2013). For G. ruber ambient seawater densities were extracted from the depth horizons dictated by the calcification temperatures of Arbuszewski et al. (2010). Because the Mg/Ca analyses in Arbuszewski et al. (2010) are debated (Hertzberg and Schmidt, 2013; Dai et al., 2019), we used the δ¹⁸O-derived calcification temperatures (T_iso). For the 3 northernmost and the 3 southernmost samples from the subtropical gyres the T_iso were fitted to the average Argo profile data of the 3 warmest months of the year as suggested by Hertzberg and Schmidt (2013).

The previously published geochemical data referred to throughout this paper were reported by Arbuszewski et al. (2010) for G. ruber and by Cléroux et al. (2013) for G. truncatulinoides. In both these studies a superset of 64 core-top sediment samples spanning from 31°N to 25°S along the Mid-Atlantic Ridge were analyzed for δ¹⁸O and Mg/Ca ratios for each species. These geochemical measurements were performed on wider or larger sieve fractions than the 300-355μm that was used in the present study, but still remain reliable calcification depth indicators. In brief, 80–100 specimens of G. ruber albus (sensu stricto) were picked from each sample from the 250–355 μm size fraction and between 10 and 25 specimens G. truncatulinoides dextral from the 355–425 µm fraction, then crushed and split into aliquots for Mg/Ca and δ¹⁸O analyses. After removal of clays, and full reductive and oxidative cleaning steps, samples were analyzed for Mg/Ca ratios on an ICP-OES. Samples for δ¹⁸O analyses were also cleaned in rinses of ultra-pure water prior to analysis (Arbuszewski et al., 2010; Cléroux et al., 2013).

Mean annual and monthly ocean temperature and salinity data of the period between January 2004 and January 2020 were extracted for each core site from the International Argo Ocean Monitoring Program (Argo, 2000). The values are quality-controlled and refer to data of the first 1000 m of the water column. Instead of extrapolating single-point hydrographic data at the exact core coordinates, the surface (2.5m) temperature and salinity values of each site was extracted from a grid area of 0.1×0.1 decimal degrees (~10×10 Km) around the site location.

Ocean carbonate system data were taken from GLODAPv1.1 (Key et al., 2004). Given the late Holocene age of the core tops, we correct the modern seawater data for the acidifying influence of anthropogenic CO₂, by subtracting the GLODAPv1.1 anthropogenic DIC estimates from the total DIC values to give an estimate of pre-industrial DIC. These values, along with paired alkalinity, nutrient, and hydrographic data, were used to make pre-industrial CO₂ system determinations for GLODAP using CO2sys.m v1.1 (van Heuven et al., 2011). This dataset was then imported to Ocean Data View (Schlitzer, 2002) and carbonate system variables obtained for each site at a range of depths using Ocean Data View’s 3D estimation tool. For consistency further oceanographic parameters (salinity, temperature, density) were calculated for the same depth ranges.

The Argo temperatures and corresponding salinities of the waters overlying each sample site were plotted against each other as the temperature (T) – salinity (S) diagram of Figure 3 and some of their water mass characteristics may be inferred. In terms of static stability at the sites, stratification is stable and does not overturn locally since the density increases with the depth. In the gyres, where water mass formation takes place by sinking of surface water along isopycnal surfaces, the T-S curve is a straight line; while the curvature is an indication of diapycnal mixing of (three) water masses (Sverdrup et al., 1942). The waters masses of the north Atlantic (green curves) have distinct curves and denote North Atlantic Central Waters (Araujo et al., 2011) that are denser due to their higher salinities, while the rest of the samples were bathed in South Atlantic Central Waters.

While for most of the sites the T-S curves have a strong curvature in the upper ~100m, indicating distinct surface water hydrography, the water of the three northern- and the three southern-most sites form straight lines, that are indicative of more active mixing of surface and subsurface waters. This is particularly true for most of the water column of the southern sites, while in the northern-most sites the slope of the curves changes at about 100m, indicating a diversification in the water masses. Sites 6-11 are below low salinity surface waters, and within the equatorial upwelling belt. The shallowest thermocline (widest temperature range for the first 100m) is found over sites 8 and 9, where the equatorial upwelling is strongest along the western margin of the African continent. Site 5 (VM22-26) is particularly affected by the river Amazon discharge exhibiting the lowest salinities for the first 30 m.

The weight of the shells of foraminifera are found to be a function of their depth habitat ( Figure 4 ). Spinose surface-dwelling species are found to be lighter than the deep dwelling non-spinose ones. O. universa was found to be the lightest species followed by G. aequilateralis and G. ruber albus. G. ruber rosea produce in general heavier shells than G. ruber albus but they are both lighter than the Trilobatus group. N. dutertrei produces the heaviest specimens in this working sieve fraction, while in some locations it is competing in being the heaviest with G. truncatulinoides, which in most cases is slightly lighter. Within the same sieve fraction, there is a 150% difference in the average shell masses between the shells of O. universa (avg. 13.6 ± 2.6 μg) and those of N. dutertrei (avg. 33.9 ± 3.4 μg). The shell weight variability for each species between the samples was on average 9%. O. universa was found to have the greatest shell weight variability (19%) along the transect. G. ruber rosea with 7% variation ( Table S1 ) is the most consistent surface dweller in weight, while the deep dwelling G. truncatulinoides varied only 6% in the central Atlantic. Due to close similarity in weights, data from T. trilobus and T. sacculifer were averaged together and are referred to as Trilobatus group. However, for a better agreement between test weight and species-specific depth habitat the size of the shell should be accounted for and thus (SBW) shell mass measurements should be normalized to cell volume (see section 3.3).

Fragmentation of specimens shows a peak in tropical waters but remains relatively mild even in the samples with the highest value [33% in sample 10 (RC 24-16)], implying relatively good preservation. By averaging the shell weights of all species for each sample ( Table 1 ) we find the lowest shell weights in sample 8 (RC24-10), while shell fragmentation was average. Shells decrease in weight by approximately 34% from the extratropical regions of the high salinity gyres to the fresher equatorial zone, when comparing the maximum recorded weights at the extratropics to the minimum in the equatorial regions. The highest contrast in test weight measurements between these regions was recorded for O. universa (82%) and the lowest for G. truncatulinoides (27%). The contrast for G. ruber albus and T. trilobus is 40% and 45% respectively, while for N. dutertrei and T. sacculifer it is 41% and 49%. G. ruber rosea and G. aequilateralis shells varied only by 7% between the equatorial and extra-equatorial sites. When shell weight measurements of all species for each site are pooled together and are divided by the number of the weighed species, an average shell weight for the foraminifera population can be determined ( Table 1 ). The largest difference in the average population weight of 34% is between the equatorial sample 8 and that of the S. Atlantic gyre (sample 16). The population average SBW has almost equally moderate correlation with sea surface density (R² = 0.27, p<0.05) and salinity (R² = 0.26, p<0.05), though not with sea surface [ CO 3 2 − ], and a hind of an anticorrelation with temperature ( Supplementary Figure 1 ). This negative relation to temperature is due to the lighter shell weights in the warmer tropical waters and is in line with observations of a negative temperature impact of temperature on foraminifera calcification (Titelboim et al., 2021). In contrast, comparing data averaged over the range of habitat depths of most species (0-100 m) shows a strong correlation with salinity (R² = 0.54, p<0.01) a relatively strong correlation with CO 3 2 − (R² = 0.36, p<0.05) and no correlation with seawater density or temperature ( Supplementary Figure 2 ).

Since foraminifera are unicellular organisms with an exoskeleton, the extent of their test primarily defines their cell size. The determination using XMCT of the total volume that their cells occupy allowed the normalization of the measured test weights to the space over which the test is distributed. This allows the determination of shell bulk density or shell density in short ( Figure 2C ), which is measure of skeletal concentration, and is expressed in μg/μm³ or more practically in g/cm³. The calculated shell densities of G. ruber albus and G. truncatulinoides along the study transect are shown in Figure 5 . Shell (bulk) densities are refinements of sieve-based shell weights and thus follow the shell weight trends of Figure 4 . G. ruber shells are denser in the higher latitudes, while their density decreases at the tropical sites. G. truncatulinoides shells on the other hand although they are consistently dense in extratropics, their density varies considerably in the tropics. Vertically, shell densities increase from G. ruber albus to G. truncatulinoides to fit the depth habitat preferences of each species. This is better manifested by sample 8 (RC24-10) from which the shell densities of 4 different species were calculated. Shell bulk densities increase from G. ruber albus to T. trilobus to N. dutertrei to G. truncatulinoides, following the known calcification depths of these species in the Atlantic (Anand et al., 2003; Steph et al., 2009; Cléroux et al., 2013; Rebotim et al., 2016; Lessa et al., 2019; Rebotim et al., 2019).

According to the estimates from sample 8, T. trilobus shell density is ~8% higher, while that of N. dutertrei and G. truncatulinoides are increased by 14% and 28% respectively. The average shell density of G. ruber albus specimens is 1.251 ± 0.198 g/cm³, while for G. truncatulinoides 1.743 ± 0.147 g/cm³, and both are higher than Sea Surface Density (SSD), which along the transect is 1.023 ± 0.001 g/cm³. However, for sample 9 (RC24-11) G. ruber albus shell density was found as low as 0.933 g/cm³, which is lower than the SSD. This deviation of shell density from sea water may possibly allow the quantification of dissolution. This low-density reconstruction is due to the low average shell weights of sample 9 (16.4 μg). Such low weights may be the result of dissolution and if shell bulk densities of sample 9 were between that of sample 8 and 10 then we estimate that ~14% or 2.7 μg of test calcite has been dissolved.

Another useful application of the determination of foraminifera shell bulk densities is that it allows estimates of total protoplasm mass densities. Foraminifera protoplasm densities necessary for free floating (neutral buoyancy) at calcification depth densities were calculated from Eq. 4 and are shown in Figure 5 with a dashed line. G. ruber albus protoplasm was estimated to be on average 249 mg/mL, which is s similar to human cells (Kim and Guck, 2020) but showed considerable variation (74%). In some samples the protoplasm densities were estimated as low as 9 mg/mL for the equatorial upwelling samples. We concluded that these values are artificially low, and they might be the result apparent buoyancy gained by the spines and/or the upwelling water velocities that force foraminifera to non-equilibrium density horizons within the mixed layer and are not plotted here (see discussion section 4.4). For the non-spinose species G. truncatulinoides protoplasm densities were reconstructed to be on average 715 mg/mL and less variable (21%) that agree with previous rough estimates (Marszalek, 1982).

The results of the X-ray attenuation analyses of G. ruber albus specimens are shown in Figure 6 . The relative densities (mean CT number) of G. ruber albus tests are variable along the transect but they are overall increased, especially for specimens from the southern sites. The lowest densities are found at the tropical sites (samples 9-10; Figure 6A ). The statistical analyses show that G. ruber albus relative test densities do not follow sieve-based shell weights but show some correlation with (volume normalized weights) ρ _shell (R² = 0.27, p<0.05). Furthermore, although for G. truncatulinoides mean test weights did not correlate with relative test densities, test bulk density showed some moderate correlation (R² = 0.32, p<0.05) to relative test density. Since foraminifera undergo ontogenetic vertical migration to greater depths (Meilland et al., 2021), the recorded mean CT values (relative densities) of G. ruber (~960) according to the study of Ofstad et al. (2021) point to adult specimens that have lived deeper in the water column, which is line with the geochemically reconstructed calcification depths considered in the present study and denote specimens in their late ontogenic stage.

The percentage of low attenuation voxels within G. ruber albus test volumes (CTX) is relatively low along the transect ( Figure 6B ) denoting good preservation of the sedimentary calcite foraminifera tests. In contrast to the northern equatorial samples, CTX within specimens is increased in the 4 southern equatorial samples. This is in line with the decreased mean CT numbers of these samples. The generally low X-ray attenuation suggests low relative test densities that may be the result of poor preservation. However, all of the specimens retain the higher magnesium juvenile chambers (Sadekov et al., 2005) that are most susceptible to dissolution and hence they are only moderately dissolved (Johnstone et al., 2010). The highest dissolution of up to ~18% CTX by volume is recorded at sample 9 (RC24-11). Based on the CT images (Supplementary material) nearly half of the specimens from this sample show slight internal disintegration of shell walls, which is a sign of dissolution. In most samples the percentage of low X-ray attenuation regions is around 10%, while the variability of preservation between specimens is higher in the South Atlantic. The samples below the North Atlantic subtropical gyre show the best preservation. Specimen fragmentation (F.I) appears to be more related to test relative density (p <0.05) than test dissolution-CTX (p < 0.10). Furthermore, although mean shell weights were found to be only moderately correlated (p <0.05) to CTX, bulk shell densities (i.e. volume normalized shell weights) were found to be significantly (p <0.01) controlled by dissolution CTX.

The results of the X-ray attenuation analyses of G. truncatulinoides specimens are shown in Figure 7 . The relative test densities of G. truncatulinoides are similar to that of G. ruber but less variable along the transect. They are overall high and like G. ruber the calcite is densest at the southernmost sites ( Figure 7A ). The lowest densities are recorded in sample 2 (VM19-308) from the north subtropical gyre for which G. ruber relative test densities are low as well. G. truncatulinoides test relative density was not found to be a function of mean shell weight, bulk shell density or bottom water [ CO 3 2 − ] (p >0.10). Furthermore, like G. ruber the calculated mean CT numbers denote specimens in their late ontogenic stage (Ofstad et al., 2021).

G. truncatulinoides specimens are found to be well preserved according to the %Low-CT volume dissolution index ( Figure 7B ). The very good preservation of the samples from the tropical regions is consistent both within and between the samples, while CTX variability is greater for the specimens from the subtropical gyres. This consistency in CTX (at approximately 7%) suggests that variations in relative test densities ( Figure 7A ) are not a result of test dissolution. Furthermore, the tomographic images show that G. truncatulinoides specimens have retained the higher magnesium juvenile chambers that are most susceptible to dissolution, providing further evidence of the excellent specimen preservation (section 4.1). Similar to G. ruber albus, G. truncatulinoides shell weights show a moderate correlation (p<0.05) to the CT dissolution index, but their bulk shell densities did not. The planktonic foraminifera F.I. was found not to be related to the relative density nor to the CTX of G. truncatulinoides tests.

The segmentation and the subsequent spatial analysis of the high-resolution CT images revealed some further information of the foraminifera tests. By dividing the test volume to the test surface area, mean shell thicknesses were calculated (section 2.2.2) for the scanned species. With an average thickness of 6.3 ± 0.7 μm, G. ruber albus tests are the thinnest ( Figure 8A ), while overall the test occupies on average 34% of its cell volume ( Figure 8B ). Test thicknesses for G. ruber vary up to ~50% along the transect, with the highest values in the subtropical gyres and lower values near the equator. G. truncatulinoides has an average test thickness of 9.3 ± 0.8 μm; test thicknesses vary up to ~40%, while its test comprises on average 47% of the cell. At location 8. RC24-10 for which two additional species were scanned, the thickest shells were calculated for N. dutertrei, whose tests however occupy less cell space in comparison to G. truncatulinoides since its cell is more voluminous (see also section 3.3). T. trilobus had an intermediate test thickness and test percentage in line with its living depth. Furthermore, G. ruber albus test thickness correlated moderately (p < 0.05) with the degree of fragmentation (F.I).

Test thicknesses of the surface and deep dwelling species appear to overlap at the subtropical gyres, while the signal disentangles at the tropical areas where G. ruber albus starts thinning. Test properties overlapping may be related to habitat compression at the downwelling gyre regions. G. ruber albus tests become up to 46% thinner at the equatorial sites and this is also reflected in the solid test percentage within its cell. G. truncatulinoides shows no distinct trend along the transect but its test thickness variability is higher at the gyres that may be the result of a mixture of crusted and non-crusted individuals. We note that for the equatorial samples 6 (RC13-189) to 12 (VM22-175) where G. truncatulinoides test thickness shows the least variability the weighed specimens were counted to be 100% dextral ( Supplementary Figure 3 ). The better test thickness consistency in this region may indicate non-encrusted G. truncatulinoides specimens. However, although test thickness variability may be distinct between the two species, the variability in test secretion is almost the same for both species (avg. 3-4%; Figure 8B ).

The spatial XMCT data may also indirectly offer information on carbonate “rain ratio” (Archer and Maier-Reimer, 1994) changes in the central Atlantic surface waters from volume measurements of the internal shell space that are meant to accommodate the protoplasm. Protoplasm space for G. ruber albus and G. truncatulinoides were calculated by subtracting the calcite test volume from the total cell volume and the results are given in Table 2 together with the amount of (sediment) contamination segmented for each sample. It appears that most (66%) of G. ruber albus cell space may be filled with protoplasm, while for G. truncatulinoides this space is reduced to 52%, so the deep dwelling globorotalid cell is almost half calcite and half biomass. Although the “rain ratio” decreases towards the equator from a surface-dweller perspective, a generalization to population level is not supported by heavier calcified deeper dwelling species ( Figure 8A ). Lastly, G. truncatulinoides cells are on average 12% larger than that of G. ruber but they can accommodate 12% less protoplasm ( Table 2 ).

The cells of G. ruber albus become more voluminous at low latitude ( Figure 8A ) and since their tests are also thinner in this region ( Supplementary Figure 4 ) the protoplasm voids are increased by ~20% in the south equatorial sites. Although the trend is not clear, almost the opposite is true for G. truncatulinoides for which the protoplasm voids are ~18% larger at the gyres than in most equatorial sites. There is also no clear trend for the total G. truncatulinoides cell volume along the transect. Thus overall G. ruber albus thickness and weight decrease as the protoplasm voids increase, while for G. truncatulinoides this is true only partly for its shell thickness ( Supplementary Figure 4 ). The above suggests that surface plankton in the Atlantic equatorial upwelling regions have the potential to produce more organic to inorganic carbon compared to higher latitudes and that the overall efficiency may also be related to the faunal composition.

Contamination of specimens with sediment infill varies both between species and samples ( Table 2 ). G. ruber specimens are infilled on average by 15% with detritus material and most contaminated are tests from the subtropical sediments. G. truncatulinoides tests however are far less contaminated (7%), possibly due to their smaller apertural openings. The degree of contamination of the G. ruber albus tests seems to have affected the shell weight measurements, since ~50% (R² = 0.53) of the shell weight variation may be explained by contamination. However, the good correlation between its test XMCT thickness and the independent shell weight measurements (R² = 0.63) restores credibility to the reported shell weight trends of G. ruber albus.

The carbonate preservation of the studied samples was mainly assessed by XMCT scanning of two different planktonic foraminifera species of different solubility ranking. G. ruber albus is known as one of the most susceptible species of planktonic foraminifera to dissolution, while G. truncatulinoides is amongst the species most resistant to dissolution (Berger, 1968; Parker and Berger, 1971; Thunell and Honjo, 1981). In accordance with previous studies (Hertzberg and Schmidt, 2013), G. ruber albus specimens were found to be well preserved in most of the core locations with some signs of dissolution at the sites below the equatorial upwelling. The sample for which the highest dissolution was inferred from the CT analysis also showed deviation from the expected shell bulk densities. This deviation was translated to a ~14% G. ruber calcite loss to dissolution at this location and this is assumed to be the highest amount of dissolution in our samples. In line with the dissolution susceptibility ranking of these species (Berger, 1968; Petró et al., 2018), G. truncatulinoides tests were found consistently well preserved along the transect ( Figure 7B ). Consequently, we interpret the overall variation in foraminifera shell weights along the transect to reflect mainly a biotic response, rather than changes of dissolution states.

The X-ray analyses were performed on tests of planktonic foraminifera species that are known to have different dissolution susceptibilities. The difference in their susceptibility is most probably a result of the mineralogical composition of the tests rather than their morphological characteristics (Petró et al., 2018). Their preservation potential is most probably related to known Mg-banded lamellar calcite layering (Johnstone et al., 2011) as the different species display large variations in the Mg/Ca composition of individual tests. The pattern of Mg/Ca variation is notably different between symbiont-bearing and symbiont-free species. In symbiotic species like G. ruber, a cyclic Mg/Ca compositional banding occurs intercalated between broader low Mg/Ca bands (Fehrenbacher et al., 2017), while symbiont-free species, such as G. truncatulinoides, typically have fewer and broader compositional bands that may reflect more closely changes in calcification temperature as they migrate within the water column (Sadekov et al., 2005). This change in calcite mineralogy/purity can be traced in the CT number histograms of the different species as a shift in the distribution of the voxels within the high X-ray attenuation area towards higher CT numbers in the lower Mg-calcite of the deeper dwelling G. truncatulinoides ( Figure 9 ).

Instead of the mean CT number, which as explained above might primarily reflect foraminifera test relative density, the ratio of the low attenuation voxels to the total test voxels was used in the present study as a dissolution index (CTX). Larger areas of low X-ray attenuation were found in the G. ruber albus specimens from the equatorial samples that are under the influence of the equatorial upwelling ( Figure 9C ). The increased dissolution of these samples may partly be related to the enhanced productivity of the region, which results in supralysoclinal dissolution (Archer et al., 1989) due to greater export and the subsequent degradation of organic matter at the sediment-water interface but also due to the thickness of the specimens, which gradually decreases towards the equatorial sites ( Figure 8A ). Thus, although the water temperature at the upwelling areas ( Figure 3 ) does not suggest great mineralogical differences to justify enhanced test dissolution, yet the thinning of the tests from the enhanced upwelling area along the western African margin make them most susceptible to dissolution and breakage (see section 3.5). In contrast to G. ruber, the XMCT analyses of G. truncatulinoides tests showed good preservation. The CTX estimates were stable for all the samples and showed that the low attenuation calcite areas (corroded areas) are less than 7% of the total voxels. This signifies that the preservation of the biogenic carbonate material of the studied samples was good and that changes in shell weights are a result of biotic forcing. Overall, although the shells were found to be best preserved below the subtropical gyres, the overall preservation of the scanned specimens was found to be good. This is also manifested by the mild bimodality of the CT number histograms (Iwasaki et al., 2015; Iwasaki et al., 2019) from samples where the highest CTX was recorded ( Figures 9C, D ).

Dissolution primarily affects foraminifera test integrity rather than its outer geometry. It initially takes place at the more soluble phases of CaCO₃ such as the higher magnesium calcite layers of the test wall or as it progresses, at the juvenile chambers that are secreted in the warmer surface waters. The pattern of shell material loss within the chamber and intact outer and inner walls has been previously described (Johnstone et al., 2010) and although with progressing dissolution the relative test density decreases, the thickness and size of the shell remains intact (Ofstad et al., 2021). Hence, the reported variations in species test thickness ( Figure 8A ) are not a signal of dissolution but mainly reflect some biotic architectural response. There is however some concern whether the XMCT derived thicknesses indicate true shell thinning or an increase in shell porosity. Foraminifera shell porosity is known to increase towards the equator (Frerichs et al., 1972) and the gradual G. ruber thinning inferred by the XMCT analysis can also be explained by increased shell porosities. On the other hand, Weinkauf et al. (2020) suggested that a change in porosity should only have a negligible effect on shell weight and thus the lighter G. ruber shells found at the equator may indeed be the result of shell thinning as the XMCT suggests.

The planktonic foraminifera shells along the studied transect were found to be lighter in the equatorial regions. When averaged together the shell weights of the 8 studied species decrease by ~32% from an average of 21.5 μg in the subtropical gyres to 16.3 μg in the southern equatorial site ( Table 1 ). The average foraminifera weights were found to be ~7.5% higher in the northern hemisphere samples, in line with previous observations for G. ruber (Hertzberg and Schmidt, 2013). The reason for this difference might be both the higher salinity (or density) of N. Atlantic subtropical waters ( Figure 3 ). G. truncatulinoides tests were found to be ~3% lighter in the region where the sample consisted of 100% dextral specimens (samples 6. RC13-189 to 12. VM22-175). This is in line with the observations of Feldmeijer et al. (2015) that dextral specimens were found to calcify shallower and would thus require lighter tests. On the other hand, G. truncatulinoides showed high test calcification and thickness variability especially in the extra-equatorial region ( Figure 8 ) perhaps due to a larger number of encrusted specimens, which is accordance with a more scattered, within the water column, growth/calcifying model that this species follows in the subtropics (Mulitza et al., 1997; LeGrande et al., 2004).

Since for all of the locations bottom water Δ CO 3 2 − did not vary considerably and the X-ray analyses did not suggest any severe shell dissolution, this collectively indicates that there is likely another factor controlling the weight differences between sites. When comparing with sea surface data, this shell weight variation in the study area was found also not to be related to surface water carbonate ion concentrations or to the degree of fragmentation but showed some moderate correlation to surface ocean density and salinity (section 3.2 and Supplementary Figure 1 ). The geochemistry of G. ruber shells from these samples was linked to surface ocean conditions (Arbuszewski et al., 2010). Since from the 8 weighed species only N. dutertrei and G. truncatulinoides occupy depths at or below the thermocline (Cléroux et al., 2007; Rebotim et al., 2019) averaged values are skewed towards the shallow-dwelling species. This might explain and validate associations with surface ocean properties.

However, when averaging the oceanographic parameters for the top 100 m, mean population shell weights reveal a relatively strong correlation with seawater salinity, while some dependency with CO 3 2 − and alkalinity also became apparent ( Supplementary Figure 2 ). Salinity has previously been called upon to explain changes in planktonic foraminifera shell weight (Weinkauf et al., 2013), which might be linked to the concentration of ingredients required in shell formation. The seawater undergoes alkalization in internal foraminifera vacuoles by active ion pumping that elevates CO 3 2 − concentration and enhances calcification (Bentov et al., 2009; de Nooijer et al., 2009; Evans et al., 2018). Saltier, more alkaline ambient waters may have increased calcifying potential by reducing the energy requirements for ion transfer and promote foraminifera calcification or optimum shell growth. Notably any relation of shell weights to seawater density disappeared due to the complete elimination of the weight correlation with temperature ( Supplementary Figure 2 ). Since temperature in known to affect planktonic foraminifera calcification and thus their shell weights (Qin et al., 2020; Titelboim et al., 2021), the absence of any correlation with temperature might be an artifact of averaging. This might be because both horizontal and vertical temperature and salinity gradients do not change to the same extent and salinity dominates in setting seasonal stratification in the tropics (Johnson et al., 2012). Besides salinity, temperature, [ CO 3 2 − ], optimum growth conditions and nutrient availability have been reported to affect foraminifera shell weights (de Villiers, 2004). If this was the case, then shell weights would have been higher in the warm and nutrient rich equatorial region, but this is not supported by the observations. In fact, shells are found to grow heavier within the oligotrophic subtropical gyres.

Although it is shown that shell weights generally increase according to the position of the foraminifera in the water column ( Figure 4 ) there appears to be an inconsistency between the exact habitat depth known for each species and its sieve-based shell weights. That is because sieve-based shell weights are not normalized to their cell volumes and although size fraction is much constrained, it is still enough to allow for species-specific differences in size/volume. Within surface dwelling species this is evident for the different G. ruber chromotypes and the pink one although it is described to live shallower it is found heavier. This discrepancy is balanced out by the size of the specimens since the pink variety is larger than the white one (Hecht, 1976). The same is true for N. dutertrei and G. truncatulinoides (see section 3.5). However, G. aequilateralis appears to be an exception with anomalously light shells for its known habitat, which on the other hand shows how well protists can control their calcification.

Previous studies in the area confirm subsurface calcification depths for T. trilobus while O. universa is thought to have greater (Anand et al., 2003; Steph et al., 2009) or wider calcification depths (Farmer et al., 2007) in comparison to G. ruber. The apparently low shell weights of O. universa may stem from the fact that its final spherical chamber is not in geometric succession to the previous ones. Within the final calcification step O. universa changes its size abruptly and thus specimens that are found in a specific sieve fraction may be of a younger ontogenic stage (lower number of chambers) compared to the rest of the species that progress geometrically. A similar discrepancy exists for N. dutertrei that is found to have heavier shells than G. truncatulinoides, which is known to live deeper in the water column (Anand et al., 2003; Cléroux et al., 2013; Rebotim et al., 2016; Lessa et al., 2019). However, the XCMT analysis (see below) revealed that G. truncatulinoides is smaller in volume than N. dutertrei, which is found to be larger and thus its shell calcite is more sparsely distributed in space resulting in lower shell bulk densities ρ _shell ( Figure 5 ).

G. ruber albus weight values however may to some extent be compromised both by dissolution and detrital sediment contamination. On average G. ruber specimens’ detrital contaminants occupy 14% of the internal void space, while contamination is highly variable (43%). Since at the time the present weight analysis was performed the samples were only treated with water (Arbuszewski et al., 2010; Hertzberg and Schmidt, 2013), the degree of decontamination is similar to that described in previous studies (Zarkogiannis et al., 2020c) for treatment only with deionized water prior to wet sieving. The above suggests that for studies of sediment samples, unless an effective cleaning method is used, the reported results of foraminifera shell weights, especially of species with large or multiple apertural openings, may be compromised by contamination. G. truncatulinoides shells on the other hand were far less contaminated and their weights did not appear to be influenced by the degree of their contamination. Furthermore, the shell weight values were not related to their preservation state, while bulk shell densities (i.e., volume normalized shell weights) showed no correlation with dissolution as well. Consequently, since dissolution has not affected all specimens, the consistent decrease in planktonic foraminifera shell weights at lower latitudes may be interpreted to reflect a biotic response.

Planktonic foraminifera shell weight variations are traditionally thought to reflect changes in the water column or sea floor [ CO 3 2 − ] (Lohmann, 1995; Barker and Elderfield, 2002; Osborne et al., 2016) but there is further evidence that this is not always the case. For example, at the same seawater [ CO 3 2 − ], foraminifera from different oceans have different shell weights, being heavier in the denser Atlantic (Broecker and Clark, 2001). Furthermore, both during the Pliocene (Davis et al., 2013) and the late Quaternary (Zarkogiannis et al., 2020b) the shell weights of planktonic foraminifera underneath the Intertropical Convergence Zone (ITCZ) were found to be invariant to atmospheric pCO₂ concentrations and thus changes in surface ocean [ CO 3 2 − ]. Zarkogiannis et al. (2019b) showed that glacial/interglacial shell weight changes vary with latitude and that the differences are greater at latitudes where the hydrology is sensitive to insolation gradients (Zarkogiannis et al., 2020b). Although moderate, the statistical correlation suggests that the observed foraminifera weight loss through shell expansion and thinning towards the equator may be the result of decreased water salinity and/or density. Tropical surface waters are relatively fresh due to excess rainfall associated with the convergent, ascending limbs of the Hadley circulation (ITCZ), whereas the subtropical oceans north and south of the equator are much saltier due to excess evaporation from the dry, descending limbs of the mean Hadley circulation. As a consequence, in these low latitude warmer and diluted waters, planktonic foraminifera would need to be lightly calcified to reach the ocean surface.

Changes in foraminifera calcite secretion according to ambient seawater densities has been previously proposed as a buoyancy regulation mechanism (Emiliani, 1954; Zarkogiannis et al., 2019a) so that foraminifera modify the mass of their shell accordingly to provide the “ballast” required for effective sinking at species specific depths (Marszalek, 1982). This is further demonstrated by the results of the present study, which show that shell weight is indeed a function of foraminifera position in the water column, and it increases with increasing depth habitat. Thus, despite the increase in acidity ( Figure 1E ) or the decrease in the degree of CaCO₃ saturation ( Δ CO 3 2 − ) with depth ( Figure 1D ), foraminifera calcite secretion increases, such that surface dwelling foraminifera shells are lighter than their deeper dwelling counterparts. The fact that foraminifera shell mass increases with depth is also manifested by the heavily calcified shells of the benthic forms (Davis et al., 2016) that calcify on the sea floor where waters can be undersaturated with respect to CaCO₃. The results suggest that the degree of calcification may provide foraminifera a mechanism to control vertical movement or positioning in the water column irrespective of the CaCO₃ saturation state of the ambient water by actively modifying it by alkalization in internal vacuoles through ion pumping (Bentov et al., 2009; de Nooijer et al., 2009; Evans et al., 2018; Gagnon et al., 2021). Optimum shell growth in dictated by habitat density requirements and it is aided by seawater salinity.

A control of ambient seawater densities on foraminifera tests and a mechanism of active calcite secretion for habitat acquisition and maintenance may perhaps explain some known foraminifera characteristics such as the gametogenic calcite, pustules at their wall texture, and the change in compositional banding within the calcite walls. Foraminifera tend to increase their lipid content prior to gametogenesis to furnish their offspring (Bé et al., 1983). An increase in the lipid content would give positive buoyancy to the cell and will displace the organism to shallower depths. Gametogenic calcite may function to offset the lipid deposition in the gametes. On the other hand, calcite pustules that are found on the shell surface of deep dwelling species (Hemleben, 1975) may serve for habitat fine tuning in depths close or below the pycnocline. At these depths small changes in seawater density render large changes in depth and small calcite depositions in the form of pustules may provide the necessary weight. Furthermore, some species undergo daily vertical migrations, rising to the surface at darkness and descending to greater depths during the day (Holmes, 1982). The temperature decrease during the periods of darkness may be enough to increase the medium’s density and force foraminifera towards the surface and thus to still warmer waters. If not biologically mediated, this may explain the higher magnesium calcite lamellae that are secreted during the night (Fehrenbacher et al., 2017) and cannot be justified by the drop in the diurnal temperatures that takes place at least up to the base of the mixed layer (Gille, 2012). Wall banding in foraminifera reduces at depths (Sadekov et al., 2005) where temperatures are more stable.

Foraminifera shell weight normalization to shell size is achieved with different methods that use either a linear dimension for normalization like measurement-based weights (Barker and Elderfield, 2002) and size-normalized weight (Beer et al., 2010) or methods that use two-dimensional approaches like area normalized shell weight (Osborne et al., 2016) and shell area density (Marshall et al., 2013) as approximations to a normalization against cell volume ( Figure 2B ). From the above methods the ones that use a liner measure although less accurate produce results that are still expressed in micrograms and are thus easily comparable between studies. On the other hand, the result of two-dimensional normalization is in μg/cm², which is thus less intuitive and not comparable to most studies. Ideally, weight would be normalized to the mean foraminifera cell volume, but cell volume has been more difficult and time consuming to establish. In the present study we used high resolution XMCT to accurately measure total foraminifera volumes and attempted a three-dimensional normalization of shell weights that led to the determination of foraminifera shell bulk densities.

Determination of non-motile plankton bulk densities is particularly important in the marine realm since they are driving the organismal positioning in the water column. It was only after the normalization over cell volume that foraminifera shell weights agreed with the preferable depth habitat for each species ( Figure 5 ). This transformation revealed some further physiological characteristics of the different foraminifera species. From the scanned species of the 300-355 μm sieve fraction, surface dwelling G. ruber albus had the smallest and thinnest shells. T. trilobus that had (+10%) thicker test and was considerably heavier than G. ruber (~40%), manages to buoy at almost the same depths (Farmer et al., 2007) or usually only slightly deeper (Steph et al., 2009; Venancio et al., 2017) because its test is bigger and thus (+23%) more voluminous. Another such example is N. dutertrei, the test of which was found to be the heaviest of the analyzed species, yet it manages to live shallower in the water column than G. truncatulinoides (Cléroux et al., 2013), which produces the second heaviest tests ( Figure 4 ). This is because although heavier, N. dutertrei is 13% bigger and so its shell bulk density is 13% lower than that of G. truncatulinoides.

The spatial information that resulted from the XMCT analyses allowed the determination of the size and percentage of the test within the cell of the scanned species. G. ruber albus has the smallest test but its average size increases towards the equator ( Table 2 ), while its mean thickness decreases ( Figure 8 ). Since its test does not simply become thinner by retaining the same size but also grows larger in the diluted equatorial waters, this suggests that as G. ruber approaches a thickness limit it may adjust its buoyancy by scaling its volume to match ambient seawater densities and meet certain depths. Accordingly, changes in test size may to some extent be physiological adaptations for buoyancy regulation. However, this is not entirely clear for G. truncatulinoides, which although it reaches its larger volumes in the tropics, in two of the equatorial samples, sizes decrease. Thus, size changes as a buoyancy regulation mechanism may apply mostly to surface dwelling species that encounter the lowest and most variable seawater densities. We interpret variations in test size and thickness to reflect a physiological response, rather than dissolution states, since dissolution lowers the test relative density ( Figure 2D ) but it leaves the thickness and the size of the test intact (Johnstone et al., 2010; Ofstad et al., 2021).

The XMCT analyses of the specimens showed that the lowest concentrations of solid material within the foraminifera cell is 30% and it is found in G. ruber albus primarily at the equatorial sites. On average the test of the G. ruber occupies 34% of its total (cell) volume and these values characterize tropical waters. These percentages are lower than those reported for similar sized specimens by Todd et al. (2020) from the Caribbean during the late Pliocene. Since late Pliocene atmospheric pCO₂ concentrations are comparable to today (Seki et al., 2010; Martínez-Botí et al., 2015) and global temperatures 2 to 3°C warmer (Haywood et al., 2016), higher pelagic calcification may be the result of the influence of the intense, warm and saltier (~1 kg/m³ denser) Mediterranean outflow to the Atlantic during the late Pliocene that altered the N. Atlantic oceanography (Khélifi et al., 2009). Along the transect, the solid test fraction of G. ruber albus varies by 55%, while for other globigerinids in the Atlantic the test fraction was found to vary up to 110% across the last glacial maximum (Zarkogiannis, 2021). This enhanced plasticity of the extent of sea surface plankton shell precipitation may be an important factor in regulating atmospheric CO₂ over the geologic time, since the production of biogenic CaCO₃ affects the alkalinity of seawater and thus the capacity of the ocean to absorb CO₂ from the atmosphere (Zondervan et al., 2001).

The deep dwelling G. truncatulinoides is more heavily calcified and test percentage varies only 27% along the transect. On average its test accounts for 47% of the total cell volume, while the maximum sample-averaged test percentage was recorded to be 52%. This small calcification variability implies that G. truncatulinoides has a better control on its calcification. If this narrow calcification window was the result of environmental and habitat stability, then calcification depth reconstructions for this species would have been steady. However, adult shell geochemistry suggests large regional variation of its mean calcification depths along the transect (Cléroux et al., 2013). We interpret this consistency in the amount of calcification as reflecting precision in buoyancy control, since adult G. truncatulinoides along the transect were found to calcify at or below the permanent (Feucher et al., 2016) pycnocline (Cléroux et al., 2007; Cléroux et al., 2013). Within this part of the water column (compared to the surface, mixed layer) densities vary less with depth and over wide depth ranges seawater density differences (Δρ) are small. Hence, small changes in the calcification of species that inhabit these depths result in large shifts of neutral buoyancy depths (Eq. 5). Therefore, deep-dwelling species may afford only small changes in their shell bulk densities, if they are to maintain certain depth ranges.

The organism (protoplasm and skeleton) will be neutrally buoyant only if the protoplasm compensates for the weight of the skeleton and the protoplast can indeed achieve a specific gravity significantly less than that of the ambient sea water (see below). The above biomass and calcite volume estimates confirm the hypothesis of Emiliani (1954) that the most important factor for foraminifera to live at certain depths is the ratio of the mass of the protoplasm and inclusions to the mass of the test and that when species have the largest ratio (G. ruber albus) they will prefer shallow habitats, while species with a smaller ratio will occupy deeper habitats (such as G. truncatulinoides).

The subtraction of the seawater density at neutral buoyancy depth from the foraminifera shell bulk density allowed the determination of foraminifera intracellular density and are in agreement with previous theoretical calculations (Marszalek, 1982). Intracellular density, a cumulative measure of the concentrations of all cellular components, impacts the physical nature of the cytoplasm and can globally affect cellular processes (Odermatt et al., 2021), yet cytoplasmic density regulation of foraminifera remains poorly understood. Although intracellular densities were found to be less than water or surface seawater density for both the shallow and deep dwelling groups, they varied considerably both between and within species. The present approach showed that on average the cell of G. ruber albus may have intracellular densities similar to that of mammalian (Kim and Guck, 2020) or yeast (Odermatt et al., 2021) cells. However, reconstructed values vary greatly with the cytoplasm appearing very diluted in some equatorial samples. This uncertainty may be the result of some apparent positive buoyancy that drive foraminifera to calcify at depths shallower than that dictated by organismal density. For example, within the mixed layer, foraminiferal spines may help heavier organisms to harvest turbulence and upwelling flow velocities to ascend to shallower depths or decrease their settling velocity (Lipps, 1973; Takahashi and Be, 1984; Furbish and Arnold, 1997). In such environments ρ _cyt from Equation 4 may be underestimated. If we hypothesize that ρ _cyt does not vary much between the studied surface and deep dwelling species and we subtract ρ _cyt of G. truncatulinoides from ρ _shell then we come up with organismal densities lower than that of water, which cannot be true. Thus, we conclude that the cytoplasm of spinose foraminifera is lighter than that of the non-spinose species but cannot be constrained solely from Equation 4.

The intracellular density differences both within a group and between groups may partly be due to the high variability in the cellular carbon content (Michaels et al., 1995), osmolytes (Boyd and Gradmann, 2002) and the variability of the water content and the density (i.e. salinity) of the intracellular water. Although foraminifera may to some degree osmoregulate, through a narrow tubular channel system, which resembles the contractile vacuole that is documented to be involved in osmoregulation in other protists as well (Gerisch et al., 2002), in terms of energy efficiency, it is logical to assume that they are largely isotonic with ambient seawater. Isotonic conditions would require higher density cellular fluids in deep dwelling-species, and this may explain the increased intracellular density reconstructions for G. truncatulinoides. Furthermore, the intracellular density may also be a function of the state of the lipidome. The effect of increasing pressure on the structure of lipids and lipid assemblies is generally qualitatively similar to the effect of decreasing temperature (Brooks, 2014). Hence, at thermocline temperatures the lipids of these ectotherms rather than being fluid must be ordered in a densely compacted gel phase (Schneider et al., 1999), such that the aggregate density of the animal will become greater as well. However, the greater compressibility of lipids compared to seawater makes any depth of neutral buoyancy unstable (Campbell and Dower, 2003).