A Late Pleistocene third molar of Hylochoerus (Suidae, Mammalia) from Rusinga Island, Kenya: paleoenvironmental implications and a note on the hypsodonty of African forest hogs

ABSTRACT

African forest hogs (genus Hylochoerus) are extant Afro-tropical suids that inhabit a variety of forest environments and thick bushlands and are predominantly herbivores. Hylochoerus likely evolved from a Pleistocene Kolpochoerus majus-like ancestor, but its recent evolutionary history is virtually unknown. Here, we describe a partial right lower third molar from the Late Pleistocene Wasiriya Beds of Rusinga Island (~50-36 ka). The crowns are mesiodistally compressed in a bunolophodont fashion and arranged in columnar pillars that resemble those of extant Hylochoerus. We provide accurate data derived from computed tomography on the hypsodonty index (HI) of extant Hylochoerus and show that the Rusinga third molar crown was as tall as those of its modern counterpart (HI = 1.8–2.0). Stable carbon isotope analyses suggest that the diet of the Rusinga specimen (δ¹³C = −17.0 ‰) was also like that of extant forest hogs (δ¹³C average = −17.6 ‰). This extremely negative value contrasts strikingly with those of other fossil large herbivores at Rusinga (δ¹³C average = −0.7 ‰.). Among the potential explanations for this anomaly, the most likely is that the Late Pleistocene paleoenvironments were more heterogeneous than previously considered and may have included closed-canopy woodland in the highlands of Rusinga.

KEYWORDS:

• Fossil
• stable isotopes
• ecology
• third molar
• computed tomography

Introduction

Despite the rich eastern African Plio-Pleistocene fossil record, the latest part of the Pleistocene is not well sampled, and the fossil record tends to be sparse. As a result, the recent evolutionary history of many African taxa remains obscured (Frantz et al. 2016). Today, sub-Saharan African Suidae form a clade with three genera, Potamochoerus, Phacochoerus, and Hylochoerus, all of which show distinct craniomandibular and dental adaptations and diverse ecological associations (Ewer 1958, 1970; Harris and White 1979; Harris and Cerling 2002; Souron 2012, 2017; Souron et al. 2015a; Lazagabaster 2019). The two living Potamochoerus species retain a relatively primitive suine dentition, characterised by low-crowned, bunodont teeth with thick enamel. Extant Potamochoerus species show a high degree of dietary flexibility, being capable of exploiting a wide array of dietary items throughout different habitats, from closed forest to open woodland/bushland (Leus and Macdonald 1997; Melletti et al. 2017; Seydack 2017). The craniomandibular and dental similarities of Potamochoerus with Pliocene Kolpochoerus and other archaic-looking specimens have brought confusion to the origin and evolution of these genera (White 1995; Bishop 2010; Souron 2012; Souron et al. J-R 2015b). Warthogs (genus Phacochoerus) have high-crowned third molars arranged in multiple rows of rounded columnar pillars. Their hypsodont dentition allows them to feed on grasses for most of the year in various types of open and occasionally dry savannah grassland/bushland habitats (Cumming 1975, 2013; Butynski and Jong 2017). Phacochoerus likely descends from a stock of Plio-Pleistocene Metridiochoerus via M. modestus but the early divergence of the two living warthog species inferred from molecular studies is strongly at odds with the fossil record because warthogs older than 1.5–1.0 Ma are unknown (Harris and White 1979; Gongora et al. 2011; Frantz et al. 2016).

African forest hogs are the largest on average of all the wild suids on Earth, though some wild boar populations and domestic pigs can be larger (D’Huart 1993; D’Huart and Kingdon 2013; Reyna-Hurtado et al. 2017). Only one species in the genus Hylochoerus has traditionally been recognised, H. meinertzhageni, with at least three subspecies accepted (Grubb 1993, 2005; D’Huart and Kingdon 2013), but it has been suggested that these forms could represent three different species (Groves and Grubb 2011; Gongora et al. 2017). The diet of forest hogs consists mainly of herbaceous plants and grasses (D’Huart and Kingdon 2013; Reyna-Hurtado et al. 2017), and they also feed occasionally on woody plants, ferns, seeds, larvae, worms, carrion, and eggs (D’Huart 1978; Kingdon 1979). Forest hogs mostly occupy Afrotropical montane forests and other closed habitats throughout western, central, and eastern Africa (Figure 1, D’Huart 1993, Cerling and Viehl 2004, D’Huart and Kingdon 2013, D’Huart and Yohannes 2014, Reyna-Hurtado et al. 2017). They are differentiated from other extant African suids, among various craniodental and mandibular traits, in having bunolophodont and relatively hypsodont molars (Harris and White 1979; Thenius 1981; Souron 2012). Hypsodonty is a relative measure of tooth crown height and while Hylochoerus molars are usually described as high-crowned, meticulous quantitative data on hypsodonty is lacking.

Figure 1. Map of Africa showing the current geographical distribution of extant Hylochoerus according to the IUCN and the location of fossil sites discussed, including Kibish, in Ethiopia, and the Late Pleistocene Wasiriya Beds and the Nyamsingula locality of Rusinga Island, in Kenya

Hylochoerus likely descends from a lineage of bunolophodont suids formed by Kolpochoerus phillipi and its descendant K. majus that goes back at least 2.8 Ma. This lineage shows gradual evolution towards the mesiodistal compression of the molar main pillars, the elaboration of talon/id complexity, and the increase in coronal cementum cover thickness (Souron et al. 2015b; Lazagabaster et al. 2018). If the origin of Hylochoerus is traceable back to a K. majus-like ancestor, the characteristic features in the skull and the dentition of Hylochoerus must have evolved relatively rapidly during the Late Pleistocene. Unfortunately, the fossil record of Hylochoerus is very sparse. The only securely dated and well-identified fossils of Hylochoerus are those from Member III of the Kibish Formation ~0.1 Ma (Figure 1, Assefa et al. 2008). Therefore, the past diversity of Hylochoerus and the evolution of its derived craniomandibular and dental traits remain uncharted.

Here, we describe a distal fragment of a lower right third molar from Late Pleistocene (~50 to 36 ka) deposits at Rusinga Island, Kenya (Figure 1), that we attribute to an extinct form of Hylochoerus. We report for the first time rigorous Hylochoerus lower third molar hypsodonty values obtained from micro-CT scans and provide previously unpublished stable carbon isotopic data. These data allow us to discuss the importance of the fossil suid tooth in the context of the paleoenvironments at Rusinga and our understanding of the Late Pleistocene suid fossil record.

Depositional setting

The specimen was recovered in 2007 from the Late Pleistocene Wasiriya Beds at the Nyamsingula locality on Rusinga Island (Figure 1). Outcrops of the Wasiriya Beds are discontinuously exposed around the perimeter of Rusinga Island, typically ~15-36 m above the present-day level of Lake Victoria (Tryon et al. 2010, 2012). They document a succession of valley-fill sediments and tephra deposits, with the main lithologies including (1) poorly-sorted sandstones and conglomerates that represent phases of high-energy erosion and deposition in seasonal channels, (2) fine-grained sandstones and siltstones that have been pedogenically modified (i.e., poorly developed paleosols) and represent intervals of landscape stability, and (3) primary and variably reworked fine-grained tephra deposits that derive from sources in the Rift Valley (Tryon et al. 2010, 2012; Blegen et al. 2015). The maximum age of the Wasiriya Beds is constrained by U-series ages of 111–94 ka from a tufa deposit at the base of the sequence (Beverly et al. 2015), consistent with the composition of the basal Wakondo Tuff, which probably derives from eruptions of Mt. Suswa dated to 100 ± 10 ka (Blegen et al. 2015). The minimum age is provided by Menengai Tuff, which is at the top of the Wasiriya Beds sequence and has been ⁴⁰Ar/³⁹Ar dated to 35.62 ± 0.26 ka (Blegen et al. 2016). The sediments exposed at the Nyamsingula locality are stratigraphically above the ~50 ka Nyamita Tuff (Blegen et al. 2015, 2017) and suggest that the vertebrate fossils from this locality are between ~50 ka and ~36 ka in age.

Material and methods

Comparative specimens

We examined Hylochoerus mandibles from the mammal collections at the National Museum of Kenya, in Nairobi (n = 1), the Museum für Naturkunde in Berlin, Germany (n = 5), and the Chicago Field Museum, U.S.A. (n = 4). Additional fossil suid metrics were compiled from the literature (see references in Table 1). Third molar mesiodistal lengths were taken as maximal occlusal lengths, buccolingual widths were taken basally on the first and on the second pair of main lateral pillars, and heights were taken as maximal crown heights. All measurements were taken by one of us (I.A.L.) with digital calipers to ensure consistency. A summary of the metric data is given in Table 1.

Table 1. Lower third molar summary metrics of selected extant and fossil Suidae

	Length					Width					Height	Ref.
	n	min	max	mean	sd	n	min	max	mean	sd
Fossil Hylochoerus
Rusinga	1			>38					16.1		>29.2	a
Extant Hylochoerus (measured from micro-CT scans)
Royal Museum for Central Africa	10	35	48	40.8	3.6	10	12	17	14.7	1.3	30	b
Museum für Naturkunde	4	36.8	43.3	41.1	2.5	4	14.4	15.7	14.9	0.5	29.6	a
Total	14	35	48	40.9	3.0	14	12	17	14.8	0.9	30
Extant Hylochoerus (tooth inside the mandible, measured with caliper)
Museum für Naturkunde	4	37.1	46.5	41.6	3.6	4	16	18.5	16.7	1.0		a
Field Museum of Natural History	4	42.1	51.2	46.8	3.1	4	17.9	18.5	18.3	0.2		a
Total	8	37.1	51.2	44.2	3.3	8	16	18.5	17.5	0.6
Kolpochoerus majus
Olduvai Bed I upper	1			42.5		1			20.7			c
Olduvai Bed II lower	1			43.5		1			20.3			c
Olduvai Bed II upper	3	45	50	48.3	2.9	3	21.7	22	21.8	0.1		c
Olduvai Bed III/IV	1			43		1			19.7			c
Daka	9	37.5	55.3	46.5	6.2	16	20.6	28.1	23.2	2	32.6	d
Buia	3	38.7	42.7	40.5	1.7	3	20.9	21.7	21.2	0.4	18	e
Total	18	37.5	55.3	44	3.6	25	20.6	28.1	21.15	0.8	32.6
Kolpochoerus paiceae (=K. olduvaiensis)
Omo-Shungura H	1			70		1			24.7			c
Omo-Shungura J	2	62.9	70.3	66.6	3.7	1			21			c
Omo-Shungura K	1			74		1						c
Omo-Shungura L	3	72.5	83.7	77.1	4.69	3	20.5	27.5	23.8	2.9		c
Koobi Fora Okote	4	67	72.6	69.6	2.2	4	22.4	25.3	24.1	1.1	35	f
Daka	10	50.9	82.4	67.8	11.8	17	18.3	28.6	23.1	3.3	37.7	d
Konso	5	61	65	62.6	1.3	4	22.5	23.5	22.9	0.4	27	g
Buia	7	50.9	66.8	58.8	5.8	7	10	17.7	14.8	2.7	17.7	e
Elandsfontein	15	63	77	71.6	4.7	15	21	25	22.9	1.2	36	h
Vaal River Gravels	2	68	70	69	1	2	21	22	21.5	0.5	35	h
Skurwerug	1	60	60	60		2	19	20	19.5	0.5		h
Total	52	50.9	83.7	67.9	3.9	57	10	28.6	21.8	1.6	37.7

a, this publication; b, this publication. Data provided by A. Souron; c, Cooke (2007); d, Gilbert (2008); e, Medin et al. (2015); f, Harris (1983); g, Suwa et al. (2014); h, Hendey and Cooke (1985)

Micro-computed tomography

Four Hylochoerus skulls were scanned using the YXLON FF35 CT computed tomography system in the Micro-CT Laboratory of the Museum für Naturkunde in Berlin, Germany. All scans have a resolution between 106 μm and 113 μm. Three specimens are adults (two females and one male) and the other is a juvenile (with the lower third molar still unerupted). Lower third molar morphology was examined by segmenting the tooth enamel using the grey value variation tool in VGStudio Max 3.3. Third molar mesiodistal lengths and buccolingual widths were taken close to the cervix. The height was measured from the middle point of the cusp at the cervix level to the top (occlusal part) of the cusp on the first and on the second pair of main lateral pillars (Figure 2). Additional metrics obtained in a similar fashion on Hylochoerus CT scans from the Royal Museum of Central Africa (n = 10) were kindly provided by A. Souron. The results are provided in Table 2 and summarised in Table 1.

Table 2. Individual lower third molar measurements of extant Hylochoerus from CT-scans. Abbreviations: BL1 and BL2, buccolingual width of first and second lophid, respectively; H1 and H2, lingual height of first and second lophid, respectively; HI1 and HI2, hypsodonty index of first and second lophid, respectively. All measurements in mm

Specimen	MD	BL1	BL2	H1	H2	HI1	HI2	Sex	Tooth side
MRAC 603	36	14	14	14	13	1.0	0.9
MRAC 2226	43	16	16	24	23	1.5	1.4
MRAC 2228	43	15	14	9	9	0.6	0.6
MRAC 2490	40	15	15	14	14	0.9	0.9
MRAC 2908	42	15	15	30	30	2.0	2.0
MRAC 3418	38	14	14	13	13	0.9	0.9
MRAC 13842	40	12	13	19	19	1.6	1.5
MRAC 14002	48	17	16	30	30	1.8	1.9
MRAC 15292	35	14	14	14	14	1.0	1.0
MRAC 17783	43	15	16	21	20	1.4	1.3
ZMB 39651	42.6	15.7	15.8	23.5	22.5	1.5	1.4	Male	R
ZMB 39653	36.8	14.7	14.6	28.9	27.9	2.0	1.9	Female	R
ZMB 83342	43.3	14.4	14.3	26.4	23.7	1.8	1.7	Female	R
ZMB 83343	41.7	14.8	14.9	29.6	27.2	2.0	1.8	Juvenile	R

Figure 2. Lower third molar measurements taken on CT scans of extant Hylochoerus. The specimen pictured is the right lower third molar of ZMB 83342 in A) sagittal, B-C) coronal, and D) transverse planes

Carbon and oxygen stable isotope analyses

Enamel from modern Hylochoerus (n = 12) and from the fossil specimen were analysed for ¹³C/¹²C and ¹⁸O/¹⁶O ratios using the conventional δ¹³C and δ¹⁸O notation where δ¹³C = (R_sample/R_standard – 1) * 1000 and R_sample and R_standard are the ¹³C/¹²C ratios for the sample and the standard, respectively, with an analogous equation for δ¹⁸O, using the international reference material VPDB (Vienna Pee Dee Belemnite) as reference material. All samples were treated using conventional methods (Passey et al. 2002) in an area of about 5 mm x 1 mm x 0.5 mm (deep) mid-way from the cervix to the occlusal surface of the teeth. About 200 micrograms for each specimen were analysed using a Gas Bench at 50°C coupled to an Isotope Ratio Mass Spectrometer (IRMS). Temperature corrections for fossil and modern samples were made according to (Passey et al. 2007). The stable isotope results can be found in Appendix 1. In addition, a dataset of δ¹³C and δ¹⁸O values were compiled from the literature for extant and Pleistocene fossil African suid taxa, which includes the genera Hylochoerus, Kolpochoerus, Phacochoerus, Potamochoerus, and Metridiochoerus (Appendix 2). Note that taxonomic identifications have been updated by IAL and A. Souron and may differ from the original source. All δ¹³C values for modern taxa are corrected to the pre-industrial δ¹³C of atmospheric CO2 (δ¹³C_1,750).

Systematic palaeontology

Institutional Abbreviations – FMNH, Chicago Field Museum of Natural History, Chicago, USA. KNM, National Museums of Kenya, Nairobi, Kenya; MRCA, Royal Museum for Central Africa, Belgium; ZMB, Museum für Naturkunde, Berlin, Germany.

Dental terminology – Generally adapted from Harris and White (1979) and Boisserie et al. (2014).

Dental abbreviations – C, canines. I, incisors. P, premolars; M, molars; m3, lower third molar.

Class Mammalia Linnaeus 1758Order Artiodactyla Owen 1848Family Suidae Gray 1821Genus Hylochoerus Thomas 1904

Generic diagnosis

Adapted from Souron (2012). Relatively large suid, especially among the Suinae, with marked sexual dimorphism; males can weigh over 250 kg. Cranio-mandibular structure derived with respect to Potamochoerus and Sus. Dental formula is reduced: 1/3I 1/1 C 3–2/2P 3/3 M. Incisors and mesial premolars reduced relative to Potamochoerus. Incisors less hypsodont than in Potamochoerus or Sus, and with less lingual relief. Third molars elongated (longer than in Potamochoerus) and more hypsodont than Potamochoerus, but less than in Phacochoerus. Simple dental wear pattern, with mesiodistally compressed crown pillars and bunolophodont aspect. Developed troughs between main cusp rows, which are filled with coronal cement. Skull broad and short. Cross-section of the rostrum is rounded. Muscles of the rhinarium are reduced, leaving no marked insertions on the bones of the skull. Ridge of the supra-canine process of males is reduced with respect to Potamochoerus and restricted to a low crest. Upper canine oriented laterally and somewhat dorsally. Zygomatic arch oriented slightly more obliquely to the sagittal plane of the skull than in Potamochoerus. In males, the zygomatic arch has significant mid-lateral expansions covered by rugose bone. Parietal wide between the temporal lines, and usually concave. Occipital relatively low and wide. Mandible with a relatively wide mandibular symphysis.

Hylochoerus sp.

Referred material

KNM-RU 49738, partial right lower third molar with collection number RU 2007-449 from Rusinga Island, Kenya, and curated at the National Museums of Kenya in Nairobi (Figure 3).

Figure 3. The right lower third molar fragment KNM-RU 49738 from Rusinga Island (A) and comparative lower third molars of extant Hylochoerus (B and D-F) and Sus scrofa (C). The fossil tooth in A1) buccal, A2) lingual, A3) distal, and A6) apical views, and A4-5) reconstruction of the missing mesial portion in occlusal views. Occlusal views of B1) ZMB 39651, D) ZMB 83342, E) ZMB 39652, and F) ZMB 39653. Right hemimandible of ZMB 39651 in lateral view (B4), with the lower third molar enamel extracted from micro-CT scans in B2) buccal, B3) lingual, B5) occlusal, and B6) distal views. Coronal cross-section at the level of the cervix in B7) ZMB 39651 and C) Sus scrofa.

Locality and horizon

Late Pleistocene Wasiriya Beds at the Nyamsingula locality on Rusinga Island and dated to between ~50 ka and 36 ka (Figure 1).

Taxonomic remarks

The eastern African suid K. limnetes likely gave rise to two highly derived forms during the Pleistocene, the eastern African species K. olduvaiensis and its southern African counterpart K. paiceae. It has been suggested that there are no clear diagnostic characters separating these two species and that K. olduvaiensis is a junior synonym of K. paiceae (Souron 2012, 2017). We follow this recommendation here for simplicity, as it does not impact our anatomical comparative analyses.

Description

Specimen KNM-RU 49738 is a distal fragment of a right m3. There is marked space between pillar rows and the main pillars are mesiodistally compressed. The mesiodistal compression is stronger than the typical condition seen in the third molars of other derived kolpochoeres, including K. majus and K. paiceae, and resembles more the derived lophodonty of extant Hylochoerus. The occlusal morphology also matches that of extant Hylochoerus where the third molar main lateral pillars have mesial and distal invaginations of the enamel rim, though generally, suid tooth cusps (-ids) tend to have a simpler, rounded morphology as wear progresses. Nevertheless, in K. majus the third molar cusps (-ids) are more circular despite the mesiodistal compression. In K. paiceae, the cusp (-id) occlusal invaginations tend to get less marked towards the distal end of the third molar and the distal cusps (-ids) often show a certain degree of mesiodistal compression. In lateral view, the arrangement of pillars is somewhat reminiscent of other hypsodont suids, including K. majus, K. paiceae, Metridiochoerus spp., and Phacochoerus spp. The distal tapering of the tooth is like in Hylochoerus or Kolpochoerus, and unlike Phacochoerus. There is considerable coronal cementum cover occlusally, forming denser patches around and between cusps. Laterally, the coronal cementum cover is appreciably thinner and in some areas it is not preserved, revealing the corrugated enamel surface that is typical of other suids with cementum cover. In recent Hylochoerus there are usually cementum gaps between lateral cusps, though in some individuals, especially those with heavily worn teeth, the coronal cementum cover is extremely extended.

It is not possible to determine with certainty how much of the mesial part of the tooth is missing, whether it is missing one pair of main lateral pillars or more. The morphology of the apical portion of the tooth, however, provides a clue about this issue. In suines, the m3 cervix tends to be wider mesially, specifically close to the first and second pair of lateral pillars. Two examples showing extant Hylochoerus and Sus scrofa m3 axial sections at approximately the level of the cervix are pictured in Figure 3B-5 and 3C; the mesial widening is marked with a curved, dashed line. This widening reflects not only the larger size of the main four mesial cusps but also the larger size of the roots that support them. Distally to the first two pairs of laterals, the cervix outline continues more or less straight or narrows gradually towards the distal end of the tooth. This pattern mirrors what is seen in KNM-RU 49738 in apical view (Fig. 3A-6). Based on this comparison, we argue that the Rusinga m3 is likely missing just the first pair of lateral pillars (the metaconid and protoconid). We have restored an approximation of the occlusal outline of the mesial portion of the tooth using extant Hylochoerus m3s as template (Figure 3-A4 and 3-A5). The nomenclature used in the rest of the description below is followed based on this tentative reconstruction. We note that any conclusions derived from estimations of the length, the total number of main lateral pillars, or the morphology of the missing mesial portion of the tooth should be taken with caution.

Figure 4. Stable carbon isotope values of Late Pleistocene Rusinga Island suids and comparative modern and fossil data for other African suid species and genera. The values shown for Kolpochoerus spp. do not include those of K. majus.

Figure 5. Stable carbon isotope values of Late Pleistocene Rusinga Island large mammals, including Bovidae, Equidae, Hippopotamidae, and Suidae

Figure 6. Stable carbon and oxygen isotope values of Late Pleistocene Rusinga Island large mammals and extant African Suidae. Note that KNM-RU 49738 falls within the range of variation of extant Hylochoerus.

We infer that the Rusinga m3 was composed of a total of four rows of main lateral pillars, which is typical in Hylochoerus and K. majus, while in K. paiceae there is at least an additional pair and usually two more pairs distally. There is, however, variation in this character in all the species mentioned. Examples of extant Hylochoerus m3s with only three and four pairs of main lateral pillars are shown in Figure 3(d-e). Distal to the presumed hypoconid and entoconid (the second pair of main lateral pillars), there is another pair of well-developed lateral pillars, though these are smaller and have a simpler morphology than their mesial counterparts. The tooth is terminated by three additional pillars. The one situated on the lingual side is larger and taller than the other two and the tip is already in occlusion. The buccal pillar is unworn but through wear progression, the lingual and buccal pillars would have formed another pair of main laterals. The distalmost pillar is shorter but has a wide base. There would have been two rounded median pillars (only the distal one and the distal portion of the mesial one are preserved) between the presumed first and second pair of main lateral pillars. This feature is present in Hylochoerus but also in other kolpochoeres, including K. majus and K. paiceae. Between the second and third pair of main lateral pillars, there is another pair of oval-shaped median pillars. The presence of one or two median pillars between rows of lateral pillars in the talonid is variable but occurs often in K. majus, K. paiceae, and other Kolpochoerus. In extant Hylochoerus there is generally only one single median pillar separating the second and third pair of laterals, but sometimes there can be two. One example is the Hylochoerus m3 pictured in Souron et al. (2015a: Figure 1(c)).

The mesiodistal length of the preserved fragment is 38 mm. The total length cannot be accurately estimated but, in our reconstruction, the tooth would have been between 46–50 mm long. This is in the upper range of extant Hylochoerus and K. majus m3 metrics. The m3 mesiodistal length average of extant Hylochoerus in our dataset is 40.9 ± 3.0 mm (n = 14), with a specimen reaching 48 mm (Table 1). The m3 mesiodistal length average is 40.9 ± 3.0 mm (n = 14) in Hylochoerus, 44.0 ± 3.6 mm in K. majus (n = 18), and 67.9 ± 3.9 mm in K. paiceae (n = 52), with individual specimens reaching 48.0 mm, 55.3 mm, and 83.7 mm, respectively. It is noteworthy that our extant Hylochoerus m3 metrics derive from basal measurements taken on CT-scans, while most published measurements were taken with calipers and occasionally on specimens that had the base still inside the mandible and not visible. The different approaches can produce results that are not directly comparable. To demonstrate this point, we remeasured the m3s of the four scanned Hylochoerus individuals with a caliper on the original specimens. We show that there can be a significant overestimation of the m3 length (max m3 length: 43.3 mm from CT-scans, 46.5 mm with a caliper) and width (15.7 mm from CT-scans, 18.5 mm with a caliper).

The hypsodonty index (HI) is calculated through the division of the maximum unworn crown height by its buccolingual width at the base. The maximum crown height in KNM-RU 49738 is 26.3 mm labially and 29.2 mm lingually. However, the tooth is slightly to moderately worn and the estimated maximum height without wear would have been between 30 to 32 mm. The buccolingual width at the base is normally measured on the anterior portion of the tooth, which in the case of KNM-ER 49738 is missing. For this reason, we use the width at the base of the presumed second pair of main lateral pillars (in our reconstruction), which measures 16.1 mm. In a sample of H. meinertzhageni (n = 8), the difference between the width of the first and second lophid is negligible (Table 2; 17.5 mm and 17.9 mm respectively). The resultant HI of the fossil specimen is between 1.8 to 2.0. This HI estimation is considerably higher than the HI value of 1.36 for H. meinertzhageni reported in Janis (1988), with a maximum m3 height of 21 mm. The m3s examined in Janis (1988) must have had a certain amount of wear and/or were still inside the mandible, evidently producing underestimations of height and overestimations of width. The data derived from CT-scans presented here, however, indicate that extant Hylochoerus teeth are as high as 30 mm (Table 2), with a HI up to 2.0. The derived kolpochoeres K. majus and K. paiceae m3s can also reach this height, though in K. majus it is a rare condition while in K. paiceae m3 crowns are often taller. The hypsodonty of the Rusinga tooth is therefore compatible with extant Hylochoerus but also with other derived Pleistocene kolpochoeres.

Taxonomic discussion

According to Souron et al. (2015b), the origin of Hylochoerus dates to the Late Pliocene K. afarensis, which gave rise to a lineage of bunolophodont suids composed of two chronospecies, K. phillipi and K. majus. The species K. phillipi is known from the Ethiopian sites of Ledi Geraru ~2.8–2.6 Ma (Lazagabaster et al. 2018) and Matabaietu ~2.5 Ma (Souron et al. 2015b). The dental morphology of the Konso material ~1.9 Ma is morphologically intermediate between K. phillipi and K. majus (Suwa et al. 2014), while more typical K. majus is found throughout various eastern African sites dating from 1.8 Ma to ca. 0.5 Ma: Olduvai Beds I–IV, in Tanzania (Leakey 1942; Cooke 2007; Bibi et al. 2018); the Upper Lomekwi and Nariokotome Mbs. in West Turkana and Olorgesailie, in Kenya (Harris and White 1979; Harris et al. 1988; Souron 2012; Potts et al. 2018); Buia, in Eritrea (Medin et al. 2015); Mb. L of the Shungura Fm., the Busidima Fm. at Gona and Asbole, the Lower Herto and Daka Mbs. and the upper Bodo beds of the Bouri Fm., in the Middle Awash, and the Melka-Kunture Fm. of Garba IV, in Ethiopia (Harris and White 1979; Clark et al. 1994; White 1995; Geraads et al. 2004a, 2004b; Gilbert 2008; Everett 2010; Souron 2012). Some undescribed specimens from the Middle Awash ~0.1 Ma in Ethiopia may also be attributed to K. majus (Souron et al. 2015b). Sometime possibly during the Middle to Late Pleistocene, a population of K. majus likely gave rise to Hylochoerus but the only securely identified fossil remains of Hylochoerus are those from Kibish, in Ethiopia, also dated to ~0.1 Ma (Assefa et al. 2008).

The m3 from Rusinga described here was briefly noted in Faith’s (2014) overview of extinct Late Pleistocene African mammals. He reported it as Kolpochoerus, suggesting that the level of compression of the pillars and cementum most closely resembled K. majus. However, our reanalysis indicates that the mesiodistal compression of the main pillars and the distance between pillar rows are more accentuated than in K. majus third molars, and the occlusal morphology – with mesiodistal invaginations – aligns this tooth better with Hylochoerus. In K. paiceae, the m3 distal cusps can sometimes show mesiodistal compression but not to the degree of the Rusinga tooth. Furthermore, K. paiceae teeth tend to be longer and more hypsodont than our estimated metrics. Other Plio-Pleistocene suid genera, Metridiochoerus spp., and Phacochoerus spp., have taller crowns than KNM-RU 49738 and the main lateral pillars are not mesiodistally compressed. Unfortunately, the fragmentary nature of this fossil does not allow us to compare absolute numbers with confidence. For this reason, further specific taxonomic attribution of the Rusinga specimen is not feasible, especially considering that the taxonomic configuration of extant forest hogs has not been fully resolved (Reyna-Hurtado et al. 2017). Distal third molar fragments can sometimes share similar morphologies across different suid genera. Given the great intra-specific variability in extant and fossil suid third molars (Harris and White 1979), the possibility that this tooth belongs to another late-surviving Kolpochoerus lineage (e.g., K. paiceae) should be taken into consideration, though we find this hypothesis unlikely. Kolpochoerus paiceae evolved elongated third molars with tall-columnar pillars and cementum cover that resemble those of Metridiochoerus spp. (Souron 2012), which have resulted in several misidentifications (e.g., in Omo-Shungura suids, Cooke 2007). The latest known K. paiceae remains, however, are dated to 0.8 Ma (Souron 2012).

Palaeoecology and paleoenvironment: stable isotope analyses

Plants that use the C₃ photosynthetic pathway (tropical dicots: trees, bushes, herbs) differ in their discrimination against ¹³CO₂ from plants that use the C₄ photosynthetic pathway (tropical grasses and sedges). C₄ photosynthesis is typically prevalent in warm and seasonally dry, open habitats, whereas C3-plants dominate in conditions of lower water stress, cooler ground temperatures, and closed canopy (Pearcy and Ehleringer 1984; Cerling et al. 2011; Uno et al. 2011). In large mammals, higher enamel δ¹³C are associated with C₄-plant eaters (e.g., grazers; generally δ¹³C > −1 ‰) while more negative δ¹³C values are found in the enamel of C₃-plant consumers (e.g., browsers; generally δ¹³C < −8 ‰; Cerling et al. 2015).

The δ¹³C of KNM-RU 49738 is −17.0 ‰, which is well within the range of extant Hylochoerus, with an average of −16.7 ± 2.5 ‰ for 24 modern specimens corrected to the pre-industrial value and a range between −13.6 ‰ and −22.6 ‰ (Figure 4). The extremely negative δ¹³C value for KNM-RU 49738 contrasts with most of the other large herbivores at Rusinga Island, with an average δ¹³C value of −0.7 ± 2.6 ‰ (n = 58; Figure 5; data from Garrett et al. 2015). This is one of the most negative δ¹³C values for any fossil tooth enamel from Africa; other herbivores with such negative values are found in cave deposits in Southeastern Asia and are interpreted as being associated with closed canopy conditions (Bocherens et al. 2017; Ma et al. 2017, 2019; Bacon et al. 2018). Thus, this particular sample presents an interesting puzzle for interpretation since there is no other clear evidence for closed-canopy conditions in the associated fauna or isotopes, and most lines of evidence point to relatively dry conditions and an expansion of C₄ grasslands across the Lake Victoria basin during the Late Pleistocene (e.g., Tryon et al. 2010, 2012, 2016; Faith et al. 2015, 2020; Garrett et al. 2015; Beverly et al. 2017). The Rusinga Hylochoerus may derive from sediments reflecting a humid phase that is otherwise not represented by the faunas, but analysis of paleosols from Rusinga and nearby Karungu indicate that relatively dry conditions persisted from ~94-36 ka (Beverly et al. 2017). The complete absence of Later Stone Age artefacts (dating from < 36 ka in the region; Tryon et al. 2016; Blegen et al. 2017) at Nyamsingula also suggests that it is unlikely the specimen derives from undocumented or eroded Holocene deposits that sample a more humid climate with closed habitats.

Other alternative hypotheses that could result in comparable negative δ¹³C signals are selective diets of C₃ plants in arid open environments or C₃ graminoids or forbs in open but humid environments (e.g., close to water bodies that provide shade and humidity; Souron 2018). The consumption of aquatic plants can also result in negative δ¹³C values (Chappuis et al. 2017). In modern tropical lowland African ecosystems, C₃ herbaceous plants are generally scarce (abundance < 10%, Sage et al. 1999). However, C₃ graminoids, forbs, and ferns, can be relatively abundant in certain savannah ecosystems (e.g., Kruger National Park; Codron et al. 2005) and in some lowland African Plio-Pleistocene habitats (Bonnefille 2010; Feakins et al. 2013; Albert et al. 2015; Lüdecke et al. 2016; Magill et al. 2016). To our knowledge, however, there are no records of extant Hylochoerus feeding on aquatic plants or consistently venturing on open habitats to feed (D’Huart 1978; Kingdon 1979; D’Huart and Kingdon 2013; Reyna-Hurtado et al. 2017).

The Plio-Pleistocene suids with the most negative δ¹³C are those from the Karonga basin in the Malawi Rift (Lüdecke et al. 2016), Elandsfontein in the southwestern Cape of South Africa (Luyt et al. 2000), and Aramis in the Middle Awash, Ethiopia (White et al. 2009). The case of Elandsfontein is unique because the southwestern Cape has long had a winter rainfall regime that favours C₃ grasses and dicots (Luyt et al. 2000; Lehmann et al. 2016); today the region is dominated fynbos, renosterveld, and strandveld plant communities that unlike anything found in eastern Africa. In the case of Karonga and Aramis, these negative δ¹³C have been interpreted as reflecting closed-canopy habitats and gallery forests respectively (Luyt et al. 2000; White et al. 2009). Furthermore, forest/closed canopy species tend to have more negative δ¹³C than open-habitat species. For example, duikers (Cephalophus spp.), which inhabit rainforests and forests, display more negative δ¹³C than the eland (Taurotragus oryx), a woodland-savanna species (Cerling et al. 2003). Both bovids, however, are almost exclusive C₃ feeders. In Queen Elizabeth National Park in Uganda, which has both warthogs and forest hogs, where the forest hogs are living near the boundary of grassland/forest, the δ¹³C values are enriched (δ¹³C = −12 to −13 ‰; Cerling and Viehl 2004). These results suggest that the fossil forest hog would have consumed C₄ resources if it had inhabited open grasslands. Therefore, the extremely negative δ¹³C of KNM-RU 49738 is most likely an indication of the presence of closed canopy.

The δ¹⁸O of KNM-RU 49738 is −1.9 ‰, which is also comparable to the values of extant Hylochoerus, with an average of −2.6 ± 1.8 ‰ (n = 22; Figure 6). Oxygen isotopes in palaeoecological contexts are generally associated with external water dependency (Sponheimer and Lee-Thorp 1999, 2001; Harris and Cerling 2002). Variations in oxygen isotope composition of mammalian tooth enamel are mainly a function of the oxygen isotope composition of food and liquid water. Leaf δ¹⁸O water is enriched (typically 10–30% more) in comparison to the meteoric source (Epstein et al. 1977; Yakir and Others 1998). Thus, mammal species that obtain the majority of their water from the plants they eat should display more positive δ¹⁸O in comparison to obligate drinkers (Sponheimer and Lee-Thorp 1999, 2001; Harris and Cerling 2002). In African bovids, browsers tend to show more enriched δ¹⁸O than do grazers. Interestingly, in African suids the patterns seem to be reversed: taxa that inhabit forested habitats (Hylochoerus, Po. porcus, and some populations of Po. larvatus) have generally low values of both δ¹⁸O and δ¹³C (Figure 6), while both ratios get higher in grazer taxa (Ph. africanus and some populations of Po. larvatus). Nevertheless, KNM-RU 49738 fits clearly in the isotope ecospace of extant Hylochoerus (with negative δ¹⁸O and δ¹³C). It is worth mentioning that the only hippopotamid sampled in Rusinga also has negative values of δ¹⁸C and δ¹³O that are close to those of Hyochoerus (Figures 5–6; Garrett et al. 2015). The values for this tooth are also extremely negative points among Plio-Pleistocene African hippopotamids (e.g., Harris et al. 2008; Cerling et al. 2008, 2015); such negative values are more common in C3-grazers in swampy habitats (e.g., the pigmy hippopotamus Choeropsis liberiensis). The presence of grassy/sedgy wetlands, therefore, cannot be completely ruled out.

The most likely explanation for the presence of Hylochoerus on Rusinga, and its very negative δ¹³C value, is that the Late Pleistocene environment was more heterogeneous than previously considered. At the Nyamita locality (~3 km SW of Nyamsingula), spring-fed rivers and wetlands promoted locally dense vegetation cover of the sort that could favour Hylochoerus, including woody vegetation, as well as grasses and sedges (Beverly et al. 2015; Garrett et al. 2015). This alone may not explain the δ¹³C value, so it is also plausible that there were some patches of closed-canopy forest in the highlands on Rusinga where Pleistocene deposits are not found.

Conclusions

Despite being a single and fragmented specimen, the lower third molar crown morphology of the Rusinga fossil – namely, its bunolophodont aspect – matches that of extant Hylochoerus third molars. We show that its crown height and hypsodonty index were as high as those of extant forest hogs. The acquired hypsodonty in the K. phillipi-K.majus-Hylochoerus lineage mirrors what happened in multiple African suid lineages since the Miocene (Harris and White 1979; Souron et al. 2015b; Souron 2017; Lazagabaster et al. 2018), though not to the extreme of some derived taxa (e.g., Notochoerus scotti, Metridiochoerus compactus or Phacochoerus). What is unique about Hylochoerus among African suids is its derived lophodonty. Kolpochoerus is usually associated with open-grassland to mixed savannah habitats after 2.5 Ma, and stable carbon isotopes suggest its diet was mainly composed of C₄ resources (Harris and Cerling 2002; Bibi et al. 2013; Cerling et al. 2015; Lazagabaster 2019; Negash et al. 2020). While multiple suid taxa were evolving adaptations to eating more abrasive foods in the context of C₄-grasslands expansion in Africa, the available fossil evidence suggests that Hylochoerus turned away from the open habitats and into the forests. Comprehending exactly when and how did this happen is currently unattainable, as the third molar described here and the remains from Kibish are the only Hylochoerus fossil remains known to date (Assefa et al. 2008). The association of Hylochoerus with closed habitats (D’Huart 1993; Cerling and Viehl 2004; D’Huart and Kingdon 2013; D’Huart and Yohannes 2014; Frantz et al. 2016) may explain why its fossil record is so sparse and provide new insights into the paleoenvironmental at Rusinga, which appears to have been more heterogeneous than typically considered.

The Rusinga Hylochoerus adds to a growing body of evidence showing that geologically recent faunas are quite different from those of the present-day (Faith 2014; Faith et al. 2019, 2020). Extant species, including the Rusinga Hylochoerus, are often found well outside their historic ranges (e.g., Rowan et al. 2015; Lesur et al. 2016; Faith et al. 2020) – leading to the formation of non-analogue species combinations – and it is becoming increasingly clear that extinct species were numerous (Faith 2014). Considerable work remains ahead if we are to sort out the emergence of what we would recognise as ‘modern’ eastern African faunal communities and the impact of terminal Pleistocene climatic changes in the present configuration of African ecosystems.

Acknowledgments

IAL acknowledges a Humboldt Postdoctoral Fellowship. JTF acknowledges support from the Natural History Museum of Utah. The stable isotope work was supported by a National Science Foundation grant #1740383 and the research permit is NACOSTI//P18/95520/23873. We thank Kevin Uno and Julia Tejada for assistance in sampling. We thank the curatorial team a the National Museums of Kenya. We thank Kristin Mahlow from the Micro-CT Laboratory at the Museum für Naturkunde, and the mammal collections curatorial team, including Frieder Mayer, Doreen Breyer, and Detlef Willborn. We thank two anonymous reviewers and Antoine Souron for his critical comments and for sharing unpublished data obtained from CT-scans, which were obtained with the help of Emmanuel Gilissen and Wim Wendelen (MRAC, Tervuren), Franck Guy (PALEVOPRIM, Univ. Poitiers), and Walter Coudyzer (Universitair Ziekenhuis Leuven). We thank John Rowan and Valentina Rovelli for their suggestions and support. Lastly, we thank those who contributed to the collection of Pleistocene faunas from Rusinga Island, including Nick Blegen, Holly Dunsworth, Will Harcourt-Smith, Kirsten Jenkins, Tom Lehmann, Kieran McNulty, Dan Peppe, and Christian Tryon.

Data availability statement

All data used in this publication is reported in the text and in Appendices 1 and 2. Supplementary datasets (Appendix 1 and Appendix 2) are also available at a permanent repository: https://osf.io/rcjts/

Supplementary material

Supplemental data for this article can be accessed here.

Disclosure statement

The authors declare no competing interests.

Additional information

Funding

This work was supported by the National Science Foundation [1740383]; Alexander von Humboldt-Stiftung.