Navigating the Limits: Unraveling Unidentified Fossil Bone and Tooth Fragments Through Histology, Chemistry, and Multivariate Statistics

Abstract

For paleoenvironmental reconstruction, paleontologists prefer large, well-preserved fossils. Yet, such specimens are rare, and countless small fragments, though abundant, often go unused. These fragments lack visible internal structure, thus requiring etching, a procedure not permitted on large, intact specimens. Our research introduces a three-step method to identify the nature of these small fragments. With their structures revealed, we can then analyze the chemical composition of identified tissues. The method was tested using samples of vertebrate fossils collected in Malawi. Even with a limited number of samples, multivariate analyses (Principal Component Analyses—PCA) of these chemical data effectively differentiate fossil and recent samples, as well as bone, dentin, and enamel. This approach successfully reveals the behavior of the mineralized tissues of fossil samples. Ultimately, by leveraging microstructural and chemical data, we can study previously unidentified fragments or rare fossils. This allows for the estimation of preservation state and helps to avoid biases in paleoenvironmental reconstructions.

1. Introduction

An organism becomes a fossil through structural and compositional changes. Three terms describe these processes: fossilization, diagenesis, and taphonomy, but their significances are not identical. The first use of “fossilization” probably occurred in the 1810s. “Diagenesis” refers to post-sedimentary, non-metamorphic transformations of sediments or sedimentary rocks [1]. Remains of organisms are now included in sediments. “Taphonomy” is “the laws of embedding [2]. The passage from the biosphere into the lithosphere is the result of numerous interwoven geological and biological phenomena.

The discovery of a new fossil site invariably raises questions regarding its formation: Is it attributable to diagenesis, taphonomy, or fossilization? Because the definitions of diagenesis and taphonomy are more precise than the definition of fossilization, the latter term is more appropriate for a new site whose origin is not yet known.

Focusing on the abundance and classification of preserved skeletal remains, research analyzes their external morphologies to determine if modifications—such as pits, striae, and fractures—originated from antemortem accidents or pathologies, embedding events, or postmortem processes ([3,4,5,6,7,8,9,10,11,12,13,14], among others). The accurate interpretation of these alterations is crucial because they directly impact our understanding of fossil preservation processes.

The original compositions of phosphatic bone and teeth are significantly modified during fossilization, offering insights into site formation [15,16,17,18,19,20,21,22,23,24,25,26]. While the original shapes and fracture patterns are also altered, studies often prioritize bone over teeth, primarily due to its abundance despite the higher taxonomic value of teeth.

To detect and identify alterations in small or large fossil mammals, most studies employ statistical analyses. Ideally, researchers would have access to a large number of samples with well-established taxonomic classifications. However, some fossil sites, despite being rich in samples, present a challenge because most of these samples are unidentifiable fragments. Identifiable samples, like skulls or mandibles with teeth, are often considered too precious for destructive analysis.

Lake-margin sites are among the sites that are very rich in fossil fragments and are concentrated along the Rift Valley in East Africa. The Chiwondo Beds (Pliocene–Pleistocene) in the Karonga District of Northern Malawi span approximately 70 km north–south and 10 km east–west (Figure 1) [27]. The Malema and Mwenirondo sites within these beds have yielded fossils of fishes, crocodiles, and large mammals [28,29,30,31,32,33,34,35]. Micro-mammals seem to be absent [30]. Previous studies have shown that fossil fish bones display distinct differences in surface modifications, crystallinity, and chemical composition [35].

A significant challenge in paleontology is the prevalence of unidentifiable bone and tooth fragments. This raises a key question: can statistical analyses effectively discern discriminant features among fossil and extant counterparts, or between tissue types, when applied to a small number of these fragments? To investigate this, we selected fragments and studied them to determine whether they were teeth or bones. We performed localized punctual analyses for chemical elements that are typically present in mammal skeletons. Furthermore, we included Fe, Mn, K, and Sr as indicators of modifications during the fossilization process, choosing these elements based on what was previously acquired [31,35].

2. Materials and Methods

2.1. Materials

Most fossil fragments collected in Malawi were unidentifiable under a stereo-microscope because they were often encrusted with sediments, obscuring their histology. Nevertheless, the discovery of skulls and mandibles from large mammals (Equidae, Suidae, Bovidae, Papio sp.) allowed for the selection of recent animals within the same taxa for comparative study. All recent animals were fresh and unburied.

Reference samples of recent mammals included Equus for long bone and incisor, Sus scrofa for enamel (molar, canine) and long bone, Bos sp. and Papio sp. for long bone, and an undetermined Cervidae from France for enamel and long bone. The Sus sample was about 5 years old and originated in the southwestern region of Paris.

Fragmented fossil teeth and bones were recovered by screening during the 1996–1997 excavations. These excavations targeted the Plio–Pleistocene Chiwondo and Chitimwe beds, which are known for their abundant vertebrate fauna (Suidae, Bovidae, Equidae). Thanks to their microstructures, the fragments were identified as bone (8), enamel (4), and dentin (6).

2.2. Sample Preparation

Fresh mammal samples were initially dried in open air. The bulk of the flesh was mechanically removed using a scalpel. To remove the remaining tissue, the samples were immersed in a commercial sodium-hypochlorite aqueous solution and subjected to ultrasonic cleaning. Subsequently, they were rinsed with Milli-Q water and dried at room temperature. Fossil teeth were extracted from water screening residuals, which were collected using a 0.5 mm mesh and then sorted manually. To preserve the teeth, only fragments were utilized. These fragments were identified based on the characteristic prismatic microstructure of the enamel, visible after cutting.

2.3. Microstructures

All samples were first cleaned using water and ultrasonic to remove the powdered sediment from their surfaces. Fractured, polished, and etched sections of both recent and fossilized teeth and bone were then examined using Philips 505 and XL 30 scanning electron microscopes (SEM) (Philips, Amsterdam, The Netherlands).

2.4. Chemical Analyses

Quantitative chemical analyses were performed using energy-dispersive spectroscopy (EDS), which utilizes a scanning electron microscope equipped with a solid-state X-Ray detector (Bruker, Berlin, Germany). Sample preparation involved embedding the teeth or bone in epoxy resin, followed by polishing the surface with progressively finer diamond paste. A brief 15-s etch with 5% formic acid was then applied. Samples were coated with a thin (approximately 400 A) carbon coat. The light etching process exposed structural details, including cracks containing sediment, enabling the precise placement of analysis points relative to observed structural features.

Quantitative chemical analysis was conducted on a Philips SEM 505 with the Link AN10000 ZAF/PB program, which is suitable for bulk specimens with rough surfaces (Oxford Instruments Microanalysis, High. Wycombe, UK). Measurements were taken with a live time of 100 or 200 s, an accelerating voltage of 15 kV., and and a spot size of 100 µm.

For individual characterization, up to 10 analyses were performed per tissue per sample and subsequently averaged.

3. Results

The fossil fragments exhibit broad taxonomic diversity, but precise identification (genus, species) is rarely possible. For this reason, recent specimens—such as Suidae, Bovidae, and Equidae—were selected based on their taxonomic proximities.

3.1. Microstructures

While the histology of recent bone, enamel, and dentin is extensively illustrated in the literature, a detailed discussion is beyond the scope of this work. Therefore, we have included only a short description and images for each tissue category. Given that only fragments were utilized for this study, neither their sizes nor shapes accurately reflect the main features required for a precise taxonomic identification. Similarly, the microstructural arrangements of the prisms of the enamel layer were not employed for taxonomic classification.

3.1.1. Enamel

The outer surfaces of teeth from a modern Sus sample exhibit brown deposits within their sharp furrows (Figure 2a). These deposits, initially sticky and colorless, are identified as dental plaque: a biofilm composed of bacteria and fungi that later turns brown.

The complex structure of the prismatic mammal enamel has become well known since the pioneer work of Korvenkontio [36]. Sections of modern Sus teeth reveal parallel, undulating bands composed of prisms or rods (Figure 2b). At higher magnifications, each prism appears to be made of elongated crystallites, which are more or less parallel to the long axis of the prism (Figure 2b,c). The shapes, orientations, and sizes of these crystallites are evident when an edge filter is used on SEM images (Figure 2g,h). The outer surface of the Cervidae tooth also displays a complex arrangement of prisms and interprisms, both formed from a single row of prisms (Figure 2d–f). Prism orientation differs in adjacent bands (Figure 2i), and each cylinder-like prism measures approximately 1–2 µm wide.

Figure 2. Modern enamel structure: (a) Outer view of a Sus scrofa molar with dental plaque. (b) Parallel prisms composed of elongated crystallites in a molar. (c) Detailed view of the same tooth, showing elongated crystallites. (d) Wavy outer surface of an incisor from an undetermined Cervidae. (e) Detailed arrangement of prisms. (f) Further detail of the same tooth. (g,h) Edge filter image of Subfigure (c), illustrating the shapes, orientations, and sizes of elongated crystallites. (i) Edge filter image of Subfigure (f), demonstrating crystallite orientations in adjacent prisms. Subfigure (b): Reprinted/adapted with permission from Ref. [37]; Subfigure (d,e): Reprinted/adapted with permission from Ref. [38].

Figure 2. Modern enamel structure: (a) Outer view of a Sus scrofa molar with dental plaque. (b) Parallel prisms composed of elongated crystallites in a molar. (c) Detailed view of the same tooth, showing elongated crystallites. (d) Wavy outer surface of an incisor from an undetermined Cervidae. (e) Detailed arrangement of prisms. (f) Further detail of the same tooth. (g,h) Edge filter image of Subfigure (c), illustrating the shapes, orientations, and sizes of elongated crystallites. (i) Edge filter image of Subfigure (f), demonstrating crystallite orientations in adjacent prisms. Subfigure (b): Reprinted/adapted with permission from Ref. [37]; Subfigure (d,e): Reprinted/adapted with permission from Ref. [38].

The outer surfaces of fossil teeth exhibit diverse patterns (Figure 3). Beyond the tissue structure, some alterations are also visible (Figure 3a–c). These include numerous small holes, varying in diameter between 40 and 120 µm, which appear either grouped or scattered (Figure 3c). Despite these alterations, the enamel structure within the holes remains well-preserved. Locally, an alternating orientation of prisms is visible (Figure 3d). The inner structures of the prisms and the elongated crystallites are also preserved to varying degrees (Figure 3e,f).

3.1.2. Dentin

Fresh sections from a recent Sus scrofa tooth exhibit the typical pattern of mammalian orthodentin: round tubules measuring 1–2 µm in diameter, their apparent shapes varying with the obliquity of the section (Figure 4a). The regular parallel arrangement of the tubules seen in the longitudinal sections is absent in the enamel. The oblique sections clearly show the abundance of tubules and the presence of a peritubular zone (Figure 4b,c).

Parallel tubules are clearly visible in the fossil teeth (Figure 5a). Interestingly, not all of these tubules are empty; some show partial to significant infilling with secondary material (Figure 5b,c). While peritubular dentin is not displayed in either the transverse or longitudinal sections, it is important to note that it is not a permanent feature [40,41].

3.1.3. Bone

The intricate organization of the osteons, including Haversian canals, lacunae, canaliculi, and concentric lamellae, is clearly visible in a transverse section of a Bos long bone (Figure 6a). Edge filter image reveals the concentric structure of the Haversian system. The longitudinal sections further reveal a similar arrangement and superposition of the Haversian structures in both Bos and Equus long bones: the plywood structure (Figure 6b,c).

Fossil bone preservation is irregular. Despite this, the plywood structure remains preserved, as seen in Figure 7a. Thin organic filaments are visible, but the structure of the bone is not identifiable (Figure 7b). The inner structures of the broken zones are not well preserved.

SEM observations indicate inconsistent preservation across the fossil bones and teeth. While enamel is known for its high resistance to alteration, the fossil teeth are broken, and their enamel layers are eroded, appearing thinner than the layers of recent samples, and are riddled with visible holes. Surprisingly, dentin, often considered fragile due to its tubules, is still identifiable. Bone structures also show variable levels of identifiability. Distinguishing between tissue types, as well as between tissues and secondary deposits, is crucial for selecting the optimal analytical zone.

3.2. Composition

Our objective was to test the potential of multivariate statistics using a limited number of samples. To achieve this, we employed a microprobe on polished surfaces, enabling the identification of both tissue and sedimentary fillings within each sample (Figure 8). Our analysis focused exclusively on biological tissues, encompassing 18 fossil samples and 9 recent ones.

Table 1. Average elemental compositions (in ppm) of recent and fossil tissues. M: average; Sd: standard deviation; B: Bone; D: Dentin; E: Enamel.

			Na	Mg	S	K	Mn	Fe	Sr	P	Ca	Ca/P
F	M	B	4260	891	1363	132	1214	6622	1395	153,784	349,965	2.27
O	Sd		479	133	343	94	1579	5580	238	2959	9165
S	M	D	5933	1701	1711	485	2018	8139	2101	164,818	358,038	2.17
S	Sd		2762	835	607	252	2195	3676	666	11,094	23,716
I	M	E	7665	2228	315	75	292	3777	1713	171,239	371,246	2.17
L	Sd		1111	365	245	57	200	489	321	4496	23,503
R	M	B	6783	3374	822	75	55	326	843	146,686	281,269	1.92
E	Sd		446	492	297	57	21	102	586	10,011	25,148
C	M	D	6788	10,621	2572	438	147	396	1319	127,343	222,162	1.74
E	Sd		178	2390	1691	204	69	68	87	16,653	25,972
N	M	E	7774	1676	796	309	0	300	1403	172,375	327,005	1.90
T	Sd		557	101	596	116	0	108	36	10,425	21,487

details the average elemental compositions of recent and fossil tissues. The taxa for the recent samples are known, but not for the fossils; therefore, the average of the 27 samples was calculated by tissue type. While certain elements (e.g., K, Mn) in the recent tissues fell below the detection limit of the microprobe, their inclusion remains valuable for understanding chemical alterations. Hypsodont teeth, such as Equus incisors and Sus canines, possess high Mg content within their dentin.

Table 1 and Figure 9 indicate similar values for Na, Sr, P, and Ca in both recent and fossil enamels. In contrast, fossil dentin is enriched in Mn, Fe, Sr, P, and Ca, but is depleted in Na, Mg, and S. Similarly, fossil bones show enrichment in S, Mn, Fe, Sr, Ca, and P, and depletion in Na and Mg. Across all fossil tissues, the Ca/P ratios are notably higher than those of the recent tissues. (Figure 10).

To better understand the modifications occurring in the fossilization process, we performed two principal component analyses (PCAs): one on recent and fossil samples combined, and another focusing solely on fossil samples.

The PCA that incorporated all samples reveals that the first principal axis accounts for 28.5% of the total variance, while axis 2 represents 27.3%, and axis 3 represents 16%. The first principal axis is characterized by higher Ca, S, and K eigenvalues. In contrast, the second principal axis sorts the samples based on higher Na, Mg, and Fe values. Despite the wide dispersion in the sample contents, the analysis successfully separated fossil and recent samples.

For the PCA focused on fossil samples, the first principal axis explained 47% of the total variance, followed by axis 2 with 20.6% and axis 3 with 11%. The first axis was primarily driven by higher Na, Mg, and P eigenvalues, while the second axis distinguished samples based on higher Mn, S, and Fe values. A significant overlap exists among bone, dentin, and enamel.

4. Discussion

4.1. Comparison Moder—Fossils from Malawi

To test the efficacy of multivariate analysis under conditions challenging the usual statistical requirements, samples were therefore selected from a fossil site with known significant modifications [31,35]. The fossils show greater mineralization than recent samples and are enriched in Ca and P. Both the fossil dentin and bone are additionally enriched in Fe, Mn, and Sr. Figure 11a reveals a clear classification trend: recent and fossil tissues are separated. Despite the “poor” quality of the samples—due to small size and imprecise identification—Figure 11b illustrates the distribution of the chemical elements within the fossil samples. Notably, the three fossil tissues do not exhibit clear separation. However, dentin is depleted in Mg, and a correlation is observed between Fe and S. A similar observation (loss of Mg) was noted in rodent incisors at various fossil sites.

One of the conclusions from the classical studies of the Malawi fossil sites is that there is a notable absence of small mammals, even despite the rigorous screening methods employed. Further examination of the cut fragments did not reveal any bones or teeth attributable to small mammals, suggesting that their absence in the sediment is not an artifact of preservation or recovery. This confirms the conclusion of the classical studies.

4.2. Why Has This Been “Successful”?

One problematic aspect of this study is the selection of modern comparative samples. The taxonomic proximity is often weak. Moreover, significant differences exist in the geographical distributions and diets of fossil and extant animals. As indicated in Figure 11b, the alterations to the chemistry are more pronounced than the original differences attributable to diet.

Another key point is the management of the samples and observations. To minimize damage during the cleaning stage, an ultrasonic bath in water for 1–2 min was used. However, not all dust and secondary deposits were removed by this process. Therefore, in some cases, we applied a mild acidic decalcification. The duration and concentration of this treatment varied depending on the sample. This acidic bath was always applied to the polished surface prepared for microprobe analysis, ensuring that the sample structure remained clearly visible. Controlling for histological factors is crucial, as secondary deposits within dentin tubules and cavities—as well as encrusting sediments—are excluded from analysis. Once SEM observations were completed on the cleaned samples, in situ chemical analyses could be performed on the identified tissues. This ensures that only biological tissues were included in the electron microprobe investigations. The biases introduced by sediment incorporation in chemical analyses have been previously demonstrated in other types of porous structures, such as eggshells [43].

4.3. Is It Possible to Use These Data to Improve Interpretations of a Fossil Site?

To estimate the potential of the described method, two categories of remains found at a single site were compared: large identifiable remains and unidentifiable fragments. The fossil sites of Malawi were suitable for this purpose. Large remains, identified as Bovidae, Equidae, Suidae, and Hippopotamidae, were found. Micromammals and carnivores appeared to be absent [28,30], and postcranial elements were rare. The presence of fish and Hippopotamidae is a strong indicator of a freshwater lake.

In most cases, the structures of fragments from Malawi sites are unidentifiable. Consequently, they appear to be of limited utility and are typically disregarded in traditional fossil site analyses. Nevertheless, the abundance of fish remains enabled the identification of fish bones through SEM, EDS, and FTIR data [35]. Although the fragments from Malema and Mwenirondo exhibit similar appearances, their chemical compositions differ. Therefore, it can be proposed that, despite their geographical proximity, their geological histories are distinct.

The absence of small mammals is another pertinent point [30]. Examination of the fragment content confirmed that this is not a bias resulting from the collection techniques.

Fish and mammal bones are similar at this scale of observation, but teeth offer more complexity. For instance, dentin is often orthodentin, meaning that it does not help distinguish between fish and mammal samples, especially because orthodentin is also found in reptiles like crocodiles. However, while mammals and reptiles have enamel, fish have enameloid, which is useful for identifying the specific taxa. One crucial point to remember is that fish bones are less mineralized than reptile bones, making their preservation less reliable.

Enamel, being more compact and mineralized than dentin or bone, does not continuously remodel in living organisms. Its internal structure is complex and unique to each species [36,44], making it theoretically the best material for accurate species assignment. The significant challenge, however, is the lack of a complete catalog detailing the enamel structure for every mammal species.

Utilizing small fragments from what appeared to be a sterile site, the proposed method successfully revealed vertebrate skeletal remains in a previously unstudied location. While enamel was present, species-level taxonomic assignment was not achieved. Crucially, the presence of large mammal and fish teeth and bone fragments points to the existence of a freshwater lake in the past.

5. Conclusions

While paleontologists typically prioritize large, complete skeletons or shells, fragmented remains are often relegated to storage and overlooked. Yet, these seemingly “valueless” samples can be rich sources of structural and chemical data, offering crucial insights into structural and compositional alterations. Using large vertebrate fossils from Malawi, it was demonstrated that a systematic examination of these fragments enables low and high taxonomic (not done here) and tissue-level identification (bone, dentin, enamel). We also present a statistical procedure that allows for robust analysis, even with limited sample sizes and a poor correspondence between recent and fossil animals. The aim of this study is neither to classically describe the test samples nor to reconstruct an ancient environment. Rather, the samples demonstrate that the proposed method, which involves analyzing neglected fragments, can enhance our understanding of a fossil site.

First, this method is valuable in fossil sites where large, well-preserved fossils are present but fragments suitable for microstructural and compositional analysis are unobtainable. In these instances, the discovery of fragments within the sediment proves highly advantageous. Second, its utility extends to scenarios in which both large (though unidentifiable) and small samples are scarce at a given site. These analyses, therefore, pave the way for novel hypotheses regarding the origin of diagenesis—such as predation or postmortem marks—and will significantly enhance our reconstructions of ancient environments.