Selective concentration of iron, titanium, and zirconium substrate minerals within Gregory’s diverticulum, an organ unique to derived sand dollars (Echinoidea: Scutelliformes)

Specimens

Various sources supplied juvenile specimens of extant sand dollars pertaining to thirteen species from nine scutelliform genera known to possess Gregory’s diverticulum (Table 1). Field collection in Costa Rica was approved by the National System of Conservation Areas (SINAC-SE-CUSBSE-PI-R-131-2016) as well as in Mexico by the Comisión Nacional de Acuacultura y Pesca (PPF/DGOPA-291/17). Species identifications were made based on the most current taxonomic treatments (Harold & Telford, 1990; Mooi, 1997; Coppard, Zigler & Lessios, 2013; Coppard, 2016; Coppard & Lessios, 2017). In addition, twenty-four juvenile Echinarachnius parma specimens collected from Lubec, ME, USA were selected for detailed mineralogic analysis using petrographic, powder X-ray diffraction (XRD), scanning electron microscopy (SEM), and EDS methods.

Analysis using a strong light source

The above-mentioned twenty-four juvenile E. parma specimens were first analyzed by projecting an intense light through each individual to confirm the presence and extent of mineral grains inside the diverticulum (Fig. 1A). In addition, selected juvenile E. parma specimens were dissected under a stereomicroscope to expose the organ and its contents for subsequent photography (Fig. 1B).

Optical mineralogy

Selected specimens of E. parma were crushed, organic material was removed with sodium hypochlorite, calcite was dissolved with 10% hydrochloric acid, and the remains were washed in alternating baths of distilled water and ethanol. Diverticulum grains from five specimens were mounted on glass slides for optical identification of mineral species using a petrographic microscope.

Grain size measurements

The size of diverticulum grains was measured directly from SEM micrographs. Substrate samples from Lubec, ME, USA were sieved into size fractions using the sequence of US Sieves 5/10/18/35/60/120/230/pan, which retain grains in the size ranges of ≥4.0 mm/2.0 mm/1.0 mm/0.5 mm/0.25 mm/0.125 mm/0.062 mm/<0.062 mm, respectively.

X-ray diffraction analysis

Samples for XRD analysis were, depending on species, prepared by crushing 20 to 50 whole specimens after removing organic tissues with sodium hypochlorite. Carbonate material from spines and test was either washed from the samples (which left minor quantities of calcite and, in some cases, aragonite associated with microscopic amounts of sediment retained on the exterior of the specimens), or completely removed with 10% hydrochloric acid. The samples were powdered and mounted in holders for analysis following standard XRD procedures. Bulk samples of substrate sands were dried and powdered without washing or acid treatment. XRD data were acquired using a D5000 diffractometer (Siemens AG, München, Germany).

Scanning electron microscopy and energy dispersive X-ray spectroscopy

SEM imagery (Fig. 1D) and EDS data were acquired using a JSM-7200 FLV FE-SEM instrument (JEOL, Akishima, Japan). Individual mineral grains from one of the optically analyzed specimens as well as an additional similar specimen were analyzed with EDS for comparison between optical and XRD results. Samples of diverticulum grains were prepared from individual specimens and small splits were mounted on stubs with carbon tape. Some samples were coated with 4–8 nm of platinum, but charging effects were minimal with grains less than 100 μm in diameter and better results were thus obtained with uncoated grains.

X-ray imaging

Two-dimensional (2D) X-ray imagery was acquired using a Skyscan 1272 µCT scanning system (Bruker, Kontich, Belgium) in 2D mode.

Micro-computed tomography

Mineral grains were identified by their X-ray attenuation using μCT at source voltages of 50 to 80 KeV—Ziegler (2012) provides parameters for selected specimens. Attenuation of X-rays by minerals was previously described (Hanna & Ketcham, 2017), and minerals were identified using μCT based on the MuCalc software (https://www.ctlab.geo.utexas.edu/software/mucalctool/). The variety and general abundance of minerals was constrained by XRD data. MuCalc was used to calculate grain-by-grain mineralogy. Processing of μCT imagery was automated using MATLAB Version 2023a (The MathWorks, Natick, MA, USA). All µCT datasets used in the present study have been deposited in the MorphoBank repository and are available for public download (http://morphobank.org/permalink/?P4915).

Image processing

Basic image processing was performed using the open source software ImageJ Version 1.52a (https://imagej.net/ij/download.html). In addition, three open source three-dimensional (3D) imaging software packages were used for reslicing and analysis of the µCT scans, i.e. SPIERS Version 3.0.1 (https://spiers-software.org/downloads.html), MorphoGraphX Version 1.1.1280 (https://morphographx.org/software/), and SlicerMorph Version 1.4 (https://slicermorph.github.io/), all running on a 64-bit Windows operating system (Microsoft Corp., Redmond, WA, USA).

Data analysis

Data analysis, including statistical calculations and principal component (PC) analysis, were performed using MATLAB Version 2023a.

Biogeography

Biogeographic data for scutelliform sand dollar species were obtained from the Global Biodiversity Information Facility (GBIF) database (https://www.gbif.org/).

Optical mineralogy, magnetism, and grain size measurements

Results of the optical mineralogic analysis of five selected juvenile specimens of Echinarachnius parma are shown in Table 2. Opaque minerals comprise an average of 54.5% of the mineral grains found inside the diverticulum, but cannot be further identified optically. However, opaque minerals are often moderately to strongly magnetic and XRD shows that they comprise magnetite, hematite, and ilmenite. In fact, many diverticulum-bearing sand dollars respond to a strong magnet, a factor directly attributable to iron- and titanium-bearing mineral content and also a simple test for the presence of Gregory’s diverticulum. However, there is no evidence of magnetotaxis, as juveniles observed in captivity show no preferred orientation with regard to the geomagnetic field (L. Zachos, 2023, personal observation).

The remaining non-opaque minerals inside the above-mentioned specimens comprise primarily zircon, hornblende (an amphibole), and augite (a pyroxene), with minor amounts of other minerals. The light mineral quartz, lithic fragments (comprising in large part quartz and feldspar), and the light mineral clinochlore (chlorite) account for 0.8%, 3.0%, and 6.6%, respectively, of the total, resulting in an average of 89.6% heavy minerals as a remainder (Table 2).

Grain size in the diverticulum is generally in the range of 50 to 150 μm (Fig. 2). This size fraction is very fine to fine sand, which is finer than the fine to medium average grain size of the substrate that these animals inhabit, i.e., 85–90% from 180 to 350 μm. The major mineral composition of grain size fractions of sand substrate from Lubec, ME, USA does not differ substantially from the average composition of bulk samples (Table 3). There is a smaller proportion of free quartz in larger grain size fractions and a corresponding increase in proportion of polymineralic lithic fragments.

Powder X-ray diffraction analysis

Mineral identification and estimates of overall composition of grains in both specimen and substrate samples were determined using XRD. In every case examined in this study, heavy minerals comprise 70 to >90% of the grains inside the diverticulum, but only 1 to 20% of grains in the substrate (Table 3). Five minerals comprise the major proportion of the heavy mineral component found in Gregory’s diverticulum: the iron (Fe) minerals magnetite (Fe₃O₄) and hematite (Fe₂O₃), the titanium (Ti) minerals ilmenite (FeTiO₃) and rutile (TiO₂), and the zirconium (Zr) mineral zircon (ZrSiO₄). The remaining heavy fraction comprises primarily amphiboles (e.g., hornblende, Ca₂[(Mg,Fe)₄Al](Al,Si)₇O₂₂(OH)₂), pyroxenes (e.g., augite, (Ca,Na)(Mg,Fe,Al,Ti)(Si,Al)₂O₆), and garnets (e.g., andradite, Ca₃Fe₂Si₃O₁₂), with other minor components, and corresponds to the results from optical mineralogy (Table 2). However, the exact mix of minerals varies for the different species recovered from different substrates, and the corresponding XRD patterns can be complex.

Scanning electron microscopy and electron-dispersive spectroscopy analysis

While XRD analysis was here used on samples comprising thousands of mineral grains, SEM and EDS were used to focus on individual grains. The elemental composition of a given sand grain can be simple to complex, and direct correspondence between this grain and a particular mineral species is often impossible when crystallographic information is lacking. However, general patterns emerged from the data when studied using PC analysis to simplify the dataset (Table S1). The first three PCs accounted for 87% of the sample variance. The first (PC 1) was strongly correlated with silicon vs titanium and iron content, the second (PC 2) strongly correlated with iron vs titanium (Fig. 3A), and the third (PC 3) strongly correlated with aluminum vs zirconium (Fig. 3B). The elemental data plotted in the PC space define clusters assignable to broad mineral groups and are consistent with the results obtained from other analytical methods. Point distribution along the first principal axis discriminates the iron and titanium oxides from other oxide, carbonate, and silicate minerals, and distribution along the second principal axis discriminates among magnetite/hematite, ilmenite, and rutile. Point distribution along the third principal axis discriminates zircon and quartz from other silicate minerals. Only two rare earth elements (REE), i.e., Ytterbium (Yb) and Neodymium (Nd) were detected in the samples analyzed, but neither were found at greater than a small fraction of a percent.

Micro-computed tomography analysis

Use of μCT has the advantage of permitting identification of grain mineralogy and, by automated counting of the individual grains, estimation of the proportion of different minerals. The analyzed mineral composition from Gregory’s diverticulum found in twelve scutelliform species conforms, in general, with results derived from optical, XRD, and EDS analyses (Table 4). Because of similarity in X-ray attenuation, magnetite and hematite cannot be differentiated and are thus here combined. Regardless of the species or its origin, the proportion of heavy minerals exceeded 68%, with ilmenite, magnetite/hematite, rutile, and zircon making up the largest proportion of the heavy mineral fraction (Table 4). Analyses of individual specimens as well as multiple virtual μCT sections through the same specimen of E. parma (Table 4) demonstrate variation in the proportions of different minerals, but the mix of heavy minerals in all cases exceeded 90% of the total grains for this species.

Methodological approach

The identification and description of the mineral grains retained in Gregory’s diverticulum can be accomplished using a variety of methods and while each of these generates similar results, they are not equivalent in either application or interpretation. The most precise measurements involve analysis of individual grains, which themselves may be mono- or polymineralic. However, these methods (optical, XRD, SEM and EDS) require removal of grains from the specimen and are therefore inherently destructive. In contrast, non-destructive methods such as X-ray or μCT rely on an attenuation proxy that does not directly indicate the mineralogy of a mineral grain. However, given sufficiently large sample sizes, the results of the destructive methods can be combined with non-destructively obtained data to develop a general method for estimating the type, variation, and relative abundance of the various minerals across multiple species. Our research objective with regard to methodology was therefore two-fold: 1) choose species for which large numbers of specimens are available for destructive as well as non-destructive analysis, and then 2) extend these results to those species for which only a limited number of specimens is available or which could not be sacrificed using destructive methods.

The relatively common sand dollar species Echinarachnius parma was here selected to represent the model species. Hundreds of juvenile specimens with a test length <20 mm, collected from the Bay of Fundy at Lubec, ME, USA, were available for both destructive and non-destructive analyses. Two simple tests can be used to determine whether or not a specimen contains heavy minerals. 1) The specimens of E. parma from Lubec, ME, USA contain enough iron and titanium minerals to be susceptible to a magnet, and this was also the case with several other species tested. 2) Additionally, a strong light source can be focused through a specimen to reveal a silhouette of the mineral-filled diverticulum (Fig. 1A). The specimen can then be dissected (or crushed) and a quick estimate of the relative proportion of light to heavy minerals can be made by separating the opaque grains from the transparent grains (Figs. 1B, 1C and 1E). In case of E. parma, many of the remaining mineral grains are often zircon, which can be recognized by the distinctive prismatic euhedral shape of the grains and by the strong orange-yellow fluorescence when illuminated with short wavelength (254 nm) ultraviolet (UV) light. Identification of individual grains using a petrographic microscope is restricted to the non-opaque fraction of the total grain content and limited by the time-consuming nature of the process. Nonetheless, its primary utility is in identification of the relatively common amphiboles, pyroxenes, and garnets as well as some of the other minor fractions of mineral grains. In turn, XRD is useful for its definitive identification and quantitative measures of the major mineral fractions derived from the diverticulum grains. However, a major drawback is the amount of material needed for a valid analysis, usually requiring the extraction of grains from 20 to 50 specimens. The XRD patterns with a mix of mineralogy exhibiting overlapping diffraction peaks complicate strict quantitative measures of composition, but are invaluable for the qualitative interpretation of the overall mineralogy and relative abundance of the individual mineral species. Unfortunately, combined SEM/EDS analysis suffers from one of the same drawbacks as the petrographic method, i.e., the time-consuming focus on individual grains, although the process is more automated and is not restricted by grain opacity. The elemental data derived from the EDS are invaluable and can usually be used to identify either the mineralogy of the sampled grain or at least the major mineral group to which it belongs.

Minerals attenuate X-rays in accordance with their elemental composition and crystal structure, and therefore attenuation can be employed as a proxy for mineralogy. Simple 2D X-ray imaging is an efficient non-destructive method to display the grain-filled diverticulum in a specimen, but is of limited value in grain identification. However, μCT imaging is an effective method for non-destructively quantifying the mineral content of a specimen in 3D. In the present study, a single virtual horizontal section was selected from each μCT 16-bit image stack, for which a pixel value of 0 indicates no attenuation and a pixel value of 65,535 (i.e., 2¹⁶−1) indicates 100% attenuation. This virtual section was processed to calculate the local intensity maxima, which approximate the dense centers of the grains (Fig. 4). The resolution of the maxima is dependent on the pixel resolution of the image (i.e., on the isotropic voxel resolution of the µCT scan). Low-resolution images only resolve a generalized density map that approximates grain centers, each representing a single local intensity maximum. The relative attenuation, as measured by image intensity, was scaled between the known absolute attenuation value for calcite and zircon. If no zircon was present or if the image gain was too high and the zircon signal was saturated, then magnetite, the mineral with the next highest attenuation coefficient, was used as reference. This scaling resulted in the 16-bit image intensity values that could be expected for each of the analyzed minerals. Low-intensity peaks representative of the calcite skeleton were removed with a high-pass filter and the remaining mean intensities were compared with the expected values generated from the software MuCalc to estimate both the identity and relative volume percent of each mineral (see Figs. S1 and S2 for exemplary X-ray images and extracted intensity distributions). Workable solutions were also derived from 8-bit images upsampled to 16-bit, but here the significant loss of spectral resolution is undesirable, particularly in case of low pixel resolution.

However, one difficulty with mineral determination from μCT imagery is the thickness variation in free mineral grains. The relationship between total absorption and the coefficient of attenuation can be represented by the Lambert-Beer Law (Hanna & Ketcham, 2017):

$${\displaystyle \frac{I}{I_0}=e^{\left(-\mu x\right)}}$$where I is the final X-ray intensity, I₀ is the initial X-ray intensity, μ is the linear attenuation coefficient, and x is the path length through the grain. The maximum reduction in X-ray intensity is at the center of the grain, but a distribution of intensities across a grain will have a peak related to the average thickness. The virtual sections in a μCT image sequence have a constant thickness, but measured intensity values are dependent on X-ray transmission through a sample in 3D and are susceptible to beam-hardening effects. The probability density functions used to model the intensity peaks are broadened and skewed around the means by these non-linear effects, adding uncertainty to quantitative interpretation.

Nonetheless, the advantages of the approach developed for the present study include the non-destructive nature of the analysis, calculation of mineral distributions directly from grain populations of individual specimens, and the ability to automate the analysis computationally. One example for the usefulness of the approach introduced here is the case of the previously mentioned species Sinaechinocyamus mai—data on diverticulum grains obtained using optical mineralogy suggested approximately the same proportion of light to heavy minerals as found in the substrate (Chen & Chen, 1994). Instead, our calibrated and standardized µCT data strongly suggest the presence of ca. 25% magnetite and 67% zircon in the first and almost 100% zircon in the second specimen analyzed (Fig. S2F and Table 4). Unfortunately, while X-ray attenuation is directly related to the density of a mineral, it is often equivocal with regard to actual mineral identification. Where the X-ray attenuation of minerals is similar, such as for almandine and rutile or for magnetite and hematite, accurate mineral identification relies on knowledge of the expected mineral population of the substrate. In many cases, this information is available from local or regional mineralogical studies of nearshore sediments, or can be inferred from the geologic character of the source rocks.

Mineral selection

Our results demonstrate that five minerals account for most of the heavy mineral fraction of the scutelliform species analyzed, i.e., magnetite (Fe₃O₄), hematite (Fe₂O₃), ilmenite (FeTiO₃), rutile (TiO₂), and zircon (ZrSiO₄). Minor heavy minerals include various amphiboles, pyroxenes, and garnets generally at concentrations too small to identify reliably. Derived sand dollars with Gregory’s diverticulum significantly concentrate heavy minerals when compared to the composition of the substrate. A two-sample t-test using the data provided in Table 3 rejected the null hypothesis that the mean ratios of light to heavy minerals between the mineral grains of diverticulum and substrate are drawn from the same population, with a p-value < 0.05. However, the particular mix of mineralogy for any given species appears to be more a function of the heavy mineral composition of the substrate than any particular mineral selectivity on the part of the animal. For example, specimens of Mellita notabilis collected from Bahia de Banderas, Mexico contain more zircon than specimens of M. notabilis collected from Playa Buena Vista, Costa Rica (Table 3). A similar variability was documented for Scaphechinus mirabilis (Elkin et al., 2012; Begun et al., 2014; Elkin, 2019). In turn, different specimens of E. parma from Lubec, ME, USA contain the same minerals, but in varying percentages (Tables 2 and 4). Exactly how the animals differentiate heavy from light minerals that are both present in the substrate is not known, but there are two likely methods: (1) direct sensing of the specific gravity (or heft) of a mineral grain or (2) sensing of an electrochemical surface character of the mineral grain–so, essentially by touch or by taste. The mystery surrounding the mechanism whereby these animals select and retain heavy minerals is deepened by the consideration that some agglutinated foraminifera also select and retain heavy minerals in their cytoplasm or even test (Makled & Langer, 2010; Sabbatini et al., 2016; Garrison, 2019). These single-celled organisms select primarily the heavy minerals zircon, rutile, and ilmenite with a mean grain size of about 100 μm, although Waśkowska (2014) described a fossil species that preferentially selected tourmaline with a grain size of 39 to 69 μm (predominantly dravite, with a specific gravity of 3.0–3.2 g/cm³).

Minerals vary considerably in wettability (Ozcan, 1992), contact angle (Drzymała, 2007; Tang et al., 2018), magnetic susceptibility (Rosenblum & Brownfield, 2000), and zeta potential (Erdemoğlu & Sarikaya, 2006; Quast, 2006; Kursun, 2010; Nduwa-Mushidi, 2016; Nduwa-Mushidi & Anderson, 2017; Ruan et al., 2018). Taken together, these physical properties could indicate commonalities among the primary minerals found inside Gregory’s diverticulum (Table 5), which in turn could play a role in selectivity. However, while it has previously been hypothesized that sand dollars select minerals based on REE content, particularly in zircon mineral grains (Panichev, 2015), REE content of the minerals analyzed here was found to be insignificant, although our EDS analyses are incomplete with regard to the full range of species retaining heavy minerals.

Some size selectivity for grains obtained from the substrate is observed, with a clear preference for grains in the 50 to 150 μm range (Fig. 2). Observations on living E. parma demonstrate that, for the smallest juveniles (1 to 5 mm test length), the diameter of the periproct (i.e., the anal opening through the test) limits the ability to pass grains larger than 150 μm out of the rectum. Interestingly, the size of diverticulum grains does not change as the animal grows, whereas the diameter of the periproct increases with age and size. Thus, while the upper limit of grain size appears to be constrained by the diameter of the periproct and the maximum opening of the teeth in the smallest juveniles, it may also be constrained by the diameter of the duct connector throughout ontogeny (Fig. 1C). If the limiting factor is indeed the duct connector then this would imply that grain selection occurs internally and is likely a passive process, whereas grain selection before ingestion implies active selection by the tube feet and/or spines. For example, stereomicroscopic studies on living E. parma demonstrate that heavy mineral grains can be selected by tube feet on the oral surface of the animal, then transported to the ambulacral grooves and finally—with the aid of the spines—into the peristome (i.e. the oral opening) as shown in Video S1. The buccal tube feet, in conjunction with the oral spines, actively select and incorporate (Video S2) or reject and discard (Video S3) mineral grains, supporting the hypothesis that selection of heavy minerals occurs before the grains are ingested (Chia, 1985). However, the precise means by which the tube feet and spines distinguish between individual mineral grains remains unknown at this point.

The mean grain size of the substrate is larger than that of the diverticulum grains and is 85 to 90% in the range of 180 to 350 μm, or fine to medium sand. This is in agreement with previous studies on substrate preference of E. parma (Stanley & James, 1971; Serafy, 1978; Harold & Telford, 1982; Serafy & Fell, 1985), although Brown (1983) reported a preference for coarse sand. Mellita tenuis shows a similar preference for fine to medium substrate grain size (Pomory, Robbins & Lares, 1995), as do Leodia sexiesperforata, Mellita quinquiesperforata, and Encope michelini (Weihe & Gray, 1968; Telford & Mooi, 1986; Hilber & Lawrence, 2009). While it has previously been claimed that L. sexiesperforata inhabits only biogenic carbonate substrates (Telford & Mooi, 1986), a strict preference for biogenic sands has not been verified (Mooi & Peterson, 2000). While we did not have access to substrate samples for the L. sexiesperforata specimens analyzed in this study, the bioclastic beach sands on the western side of Barbados, where one specimen was collected (Table 1), can contain up to 10% combined quartz and heavy mineral components (Limonta et al., 2015).

The observation that the concentration of light minerals (primarily quartz and feldspar) can range from 0% to 25% of diverticulum grains and that also minor amounts of other minerals may occur provides evidence that sometimes the animals simply pick up what is readily available in the substrate. Sedimentologic studies distinguish shallow-water coastal provinces characterized by specific heavy mineral suites. In the western hemisphere, these studies are limited in scope to the Atlantic coast of North America (Van Gosen & Ellefsen, 2018), dominated by ilmenite and zircon; the Gulf of Mexico (Davis, 2017), dominated by rutile and zircon; and the Pacific coast of Mexico (Carranza-Edwards et al., 2009, 2019), dominated by magnetite and zircon. Heavy minerals in the Okhotsk Sea are predominantly hornblende, hypersthene, and epidote, with only minor amounts of ilmenite, magnetite, and zircon (Wang et al., 2021). In the southern Sea of Japan, shore sediments are predominantly hypersthene and hornblende, with only minor amounts of zircon (Yokota et al., 1990). Heavy minerals in the lower reaches of rivers in western Taiwan are predominantly zircon and garnet (Deng et al., 2016; Garzanti et al., 2023). In addition, while the overall heavy mineral concentration of the substrate can be determined and compared with the concentration inside the diverticulum, it is also possible that the animals selected mineral grains from a naturally concentrated source. For example, heavy minerals in littoral sands are usually deposited in thin layers or “stringers” with much higher concentration than the overall deposit (Van Gosen & Ellefsen, 2018).

Phylogeography

The diverticulum-bearing scutelliforms are restricted to the eastern and western coasts of North and South America as well as the northeastern coast of Asia (Fig. 5). The Astriclypeidae, a scutelliform taxon hypothesized to be sister to all other extant scutelliforms (Lee et al., 2023) is not found in these regions (Ghiold & Hoffman, 1986) and lacks Gregory’s diverticulum (Ziegler et al., 2016). The oldest fossil records of scutelliform echinoids (Eoscutellidae and Protoscutellidae) occur in North America, and it has long been proposed that this clade originated there (Stefanini, 1924). Strictly European fossil scutelliform genera (Scutella, Parascutella, Remondella, Scutulum, Samlandaster) were distinguished by Durham (1955), and no evidence of Gregory’s diverticulum has been reported for any of these. Such lack of evidence is supported by a recent study of juvenile specimens from several protoscutellid species from Eocene deposits of eastern North America that has failed to produce definitive evidence for the presence of a mineral-filled diverticulum (Zachos & Ziegler, 2021). To date, the oldest proven occurrence of a mineral-filled diverticulum is therefore that of Kewia marquamensis (Echinarachniidae) from the late Oligocene of Oregon (Linder, 1986; Linder, Durham & Orr, 1988), suggesting that the lineage of scutelliforms with the organ evolved in western North America in the middle to late Paleogene.