Towards a Biological Understanding of the Middle to Upper Palaeolithic Transition

Cover Page

The handle http://hdl.handle.net/1887/49751 holds various files of this Leiden University dissertation

Author: Welker, F.

Title: The palaeoproteomic identification of pleistocene hominin skeletal remains Issue Date: 2017-05-18

Frido Welker

The Palaeoproteomic Identification of Pleistocene Hominin Skeletal Remains:

Towards a Biological Understanding of the Middle to Upper Palaeolithic Transition

QLSYGYDEKSTGGISVPGPMGPSGPRGLPGPPGAPGPQGFQGPPGEPGEPGASGPMGPRGPPGPPGKNGD DGEAGKPGRPGERGPPGPQGARGLPGTAGLPGMKGHRGFSGLDGAKGDAGPAGPKGEPGSPGENGAPGQM GPRGLPGERGRPGAPGPAGARGNDGATGAAGPPGPTGPAGPPGFPGAVGAKGEAGPQGPRGSEGPQGVRG EPGPPGPAGAAGPAGNPGADGQPGAKGANGAPGIAGAPGFPGARGPSGPQGPGGPPGPKGNSGEPGAPGS KGDTGAKGEPGPVGVQGPPGPAGEEGKRGARGEPGPTGLPGPPGERGGPGSRGFPGADGVAGPKGPAGER GSPGPAGPKGSPGEAGRPGEAGLPGAKGLTGSPGSPGPDGKTGPPGPAGQDGRPGPPGPPGARGQAGVMG FPGPKGAAGEPGKAGERGVPGPPGAVGPAGKDGEAGAQGPPGPAGPAGERGEQGPAGSPGFQGLPGPAGP PGEAGKPGEQGVPGDLGAPGPSGARGERGFPGERGVQGPPGPAGPRGANGAPGNDGAKGDAGAPGAPGSQ GAPGLQGMPGERGAAGLPGPKGDRGDAGPKGADGSPGKDGVRGLTGPIGPPGPAGAPGDKGESGPSGPAG PTGARGAPGDRGEPGPPGPAGFAGPPGADGQPGAKGEPGDAGAKGDAGPPGPAGPAGPPGPIGNVGAPGA KGARGSAGPPGATGFPGAAGRVGPPGPSGNAGPPGPPGPAGKEGGKGPRGETGPAGRPGEVGPPGPPGPA GEKGSPGADGPAGAPGTPGPQGIAGQRGVVGLPGQRGERGFPGLPGPSGEPGKQGPSGASGERGPPGPMG PPGLAGPPGESGREGAPGAEGSPGRDGSPGAKGDRGETGPAGPPGAPGAPGAPGPVGPAGKSGDRGETGP AGPAGPVGPVGARGPAGPQGPRGDKGETGEQGDRGIKGHRGFSGLQGPPGPPGSPGEQGPSGASGPAGPR GPPGSAGAPGKDGLNGLPGPIGPPGPRGRTGDAGPVGPPGPPGPPGPPGPPSAGFDFSFLPQPPQEKAHD GGRYYRAKQYDGKGVGLGPGPMGLMGPRGPPGAAGAPGPQGFQGPAGEPGEPGQTGPAGARGPAGPPGKA GEDGHPGKPGRPGERGVVGPQGARGFPGTPGLPGFKGIRGHNGLDGLKGQPGAPGVKGEPGAPGENGTPG QTGARGLPGERGRVGAPGPAGARGSDGSVGPVGPAGPIGSAGPPGFPGAPGPKGEIGAVGNAGPAGPAGP RGEVGLPGLSGPVGPPGNPGANGLTGAKGAAGLPGVAGAPGLPGPRGIPGPVGAAGATGARGLVGEPGPA GSKGESGNKGEPGSAGPQGPPGPSGEEGKRGPNGEAGSAGPPGPPGLRGSPGSRGLPGADGRAGVMGPPG SRGASGPAGVRGPNGDAGRPGEPGLMGPRGLPGSPGNIGPAGKEGPVGLPGIDGRPGPIGPAGARGEPGN IGFPGPKGPTGDPGKNGDKGHAGLAGARGAPGPDGNNGAQGPPGPQGVQGGKGEQGPPGPPGFQGLPGPS GPAGEVGKPGERGLHGEFGLPGPAGPRGERGPPGESGAAGPTGPIGSRGPSGPPGPDGNKGEPGVVGAVG TAGPSGPSGLPGERGAAGIPGGKGEKGEPGLRGEIGNPGRDGARGAPGAVGAPGPAGATGDRGEAGAAGP AGPAGPRGSPGERGEVGPAGPNGFAGPAGAAGQPGAKGERGAKGPKGENGVVGPTGPVGAAGPAGPNGPP GPAGSRGDGGPPGMTGFPGAAGRTGPPGPSGISGPPGPPGPAGKEGLRGPRGDQGPVGRTGEVGAVGPPG FAGEKGPSGEAGTAGPPGTPGPQGLLGAPGILGLPGSRGERGLPGVAGAVGEPGPLGIAGPPGARGPPGA VGSPGVNGAPGEAGRDGNPGNDGPPGRDGQPGHKGERGYPGNIGPVGAAGAPGPHGPVGPAGKHGNRGET GPSGPVGPAGAVGPRGPSGPQGIRGDKGEPGEKGPRGLPGLKGHNGLQGLPGIAGHHGDQGAPGSVGPAG PRGPAGPSGPAGKDGRTGHPGTVGPAGIRGPQGHQGPAGPPGPPGPPGPPGVSGGGYDFGYDGDFYRA QLSYGYDEKSTGGISVPGPMGPSGPRGLPGPPGAPGPQGFQGPPGEPGEPGASGPMGPRGPPGPPGKNGD

DGEAGKPGRPGERGPPGPQGARGLPGTAGLPGMKGHRGFSGLDGAKGDAGPAGPKGEPGSPGENGAPGQM GPRGLPGERGRPGAPGPAGARGNDGATGAAGPPGPTGPAGPPGFPGAVGAKGEAGPQGPRGSEGPQGVRG EPGPPGPAGAAGPAGNPGADGQPGAKGANGAPGIAGAPGFPGARGPSGPQGPGGPPGPKGNSGEPGAPGS KGDTGAKGEPGPVGVQGPPGPAGEEGKRGARGEPGPTGLPGPPGERGGPGSRGFPGADGVAGPKGPAGER GSPGPAGPKGSPGEAGRPGEAGLPGAKGLTGSPGSPGPDGKTGPPGPAGQDGRPGPPGPPGARGQAGVMG FPGPKGAAGEPGKAGERGVPGPPGAVGPAGKDGEAGAQGPPGPAGPAGERGEQGPAGSPGFQGLPGPAGP PGEAGKPGEQGVPGDLGAPGPSGARGERGFPGERGVQGPPGPAGPRGANGAPGNDGAKGDAGAPGAPGSQ GAPGLQGMPGERGAAGLPGPKGDRGDAGPKGADGSPGKDGVRGLTGPIGPPGPAGAPGDKGESGPSGPAG PTGARGAPGDRGEPGPPGPAGFAGPPGADGQPGAKGEPGDAGAKGDAGPPGPAGPAGPPGPIGNVGAPGA KGARGSAGPPGATGFPGAAGRVGPPGPSGNAGPPGPPGPAGKEGGKGPRGETGPAGRPGEVGPPGPPGPA GEKGSPGADGPAGAPGTPGPQGIAGQRGVVGLPGQRGERGFPGLPGPSGEPGKQGPSGASGERGPPGPMG PPGLAGPPGESGREGAPGAEGSPGRDGSPGAKGDRGETGPAGPPGAPGAPGAPGPVGPAGKSGDRGETGP AGPAGPVGPVGARGPAGPQGPRGDKGETGEQGDRGIKGHRGFSGLQGPPGPPGSPGEQGPSGASGPAGPR GPPGSAGAPGKDGLNGLPGPIGPPGPRGRTGDAGPVGPPGPPGPPGPPGPPSAGFDFSFLPQPPQEKAHD GGRYYRAKQYDGKGVGLGPGPMGLMGPRGPPGAAGAPGPQGFQGPAGEPGEPGQTGPAGARGPAGPPGKA GEDGHPGKPGRPGERGVVGPQGARGFPGTPGLPGFKGIRGHNGLDGLKGQPGAPGVKGEPGAPGENGTPG QTGARGLPGERGRVGAPGPAGARGSDGSVGPVGPAGPIGSAGPPGFPGAPGPKGEIGAVGNAGPAGPAGP RGEVGLPGLSGPVGPPGNPGANGLTGAKGAAGLPGVAGAPGLPGPRGIPGPVGAAGATGARGLVGEPGPA GSKGESGNKGEPGSAGPQGPPGPSGEEGKRGPNGEAGSAGPPGPPGLRGSPGSRGLPGADGRAGVMGPPG SRGASGPAGVRGPNGDAGRPGEPGLMGPRGLPGSPGNIGPAGKEGPVGLPGIDGRPGPIGPAGARGEPGN IGFPGPKGPTGDPGKNGDKGHAGLAGARGAPGPDGNNGAQGPPGPQGVQGGKGEQGPPGPPGFQGLPGPS GPAGEVGKPGERGLHGEFGLPGPAGPRGERGPPGESGAAGPTGPIGSRGPSGPPGPDGNKGEPGVVGAVG TAGPSGPSGLPGERGAAGIPGGKGEKGEPGLRGEIGNPGRDGARGAPGAVGAPGPAGATGDRGEAGAAGP AGPAGPRGSPGERGEVGPAGPNGFAGPAGAAGQPGAKGERGAKGPKGENGVVGPTGPVGAAGPAGPNGPP GPAGSRGDGGPPGMTGFPGAAGRTGPPGPSGISGPPGPPGPAGKEGLRGPRGDQGPVGRTGEVGAVGPPG FAGEKGPSGEAGTAGPPGTPGPQGLLGAPGILGLPGSRGERGLPGVAGAVGEPGPLGIAGPPGARGPPGA VGSPGVNGAPGEAGRDGNPGNDGPPGRDGQPGHKGERGYPGNIGPVGAAGAPGPHGPVGPAGKHGNRGET GPSGPVGPAGAVGPRGPSGPQGIRGDKGEPGEKGPRGLPGLKGHNGLQGLPGIAGHHGDQGAPGSVGPAG PRGPAGPSGPAGKDGRTGHPGTVGPAGIRGPQGHQGPAGPPGPPGPPGPPGVSGGGYDFGYDGDFYRA

The Palaeoproteomic Identification of Pleistocene Hominin Skeletal Remains:

Towards a Biological Understanding of the Middle to Upper Palaeolithic Transition

Proefschrift

ter verkrijging van

de graad van Doctor aan de Universiteit Leiden, op gezag van Rector Magnificus Prof. mr. C.J.J.M. Stolker,

volgens besluit van het College voor Promoties te verdedigen op 18 mei 2017

klokke 16:15 uur

door

Frido Welker

Geboren te Harderwijk, Nederland

in 1991

Promotores:

Prof. dr. J.-J. Hublin (Universiteit Leiden) Prof. dr. M.J. Collins (University of York)

Promotiecommissie:

Prof. dr. J.W.M. Roebroeks (Universiteit Leiden) Prof. dr. M. van Kolfschoten (Universiteit Leiden) Prof. dr. I. Barnes (Natural History Museum, London) Dr. M. Soressi (Universiteit Leiden)

Dr. S.P. McPherron (Max-Planck-Institute for Evolutionary Anthropology, Leipzig)

This research has been made possible through funding from the Max-Planck-Gesellschaft.

Cover design: Graphical representation of the COL1α1 and COL1α2 sequence from Homo sapiens (UniProt identifiers P02452 and P08123, respectively). Protein network derived from endogenous proteins identified in Chapter 5, analyzed through STRING (v.9) and modified in Cytoscape (v.3.2.1).

“If your fingers cannot manage this knot,

it has become so hard because no one has tried!”

Dante Alighieri - The Divine Comedy: Paradiso Canto XXVIII.

Contents

Chapter 1

Introduction. 1

Chapter 2

Using ZooMS to identify fragmentary bone from the 35 Late Middle/Early Upper Palaeolithic sequence of

Les Cottés, France.

Chapter 3

Ancient proteins resolve the evolutionary history of 45 Darwin’s South American ungulates.

Chapter 4

Variations in glutamine deamidation for a Châtelperronian 85 bone assemblage as measured by peptide mass

fingerprinting of collagen.

Chapter 5

Palaeoproteomic evidence identifies archaic hominins 101 associated with the Châtelperronian at the Grotte du

Renne.

Chapter 6

Discussion & Conclusion. 185

Summary

207

Samenvatting

209

Acknowledgements

211

Curriculum Vitae

213

1

Chapter 1:

Introduction

1. Perspectives on the Middle to Upper Palaeolithic Transition

The Middle to Upper Palaeolithic Transition (MUPT) in western Europe concerns a chronological period during which local Neanderthal populations are replaced by Anatomically Modern Human populations (AMHs). At the end of the transitional period AMHs are present in large portions of the territory formerly occupied by Neanderthal populations, while these themselves have gone extinct. The replacement is a permeable one from a genetic perspective with a modest proportion of Neanderthal ancestry present in non-African modern humans. This is commonly interpreted as the occurrence of gene flow between the two populations (Green et al., 2010). Because of its chronological position and association to the spread of AMHs out of southwestern Asia, the MUPT takes a central role in discussions regarding the development and spread of behavioural modernity, including aspects of cognition and its archaeological manifestations (McBrearty and Brooks, 2000), the shaping of genetic variation in present day modern human populations (Green et al., 2010; Vernot et al., 2016) and the timing and causes of Neanderthal extinction (Higham et al., 2014).

The exact processes facilitating this transition have sparked considerable controversy, especially in regards to several so-called “transitional” assemblages. These technocomplexes are placed chronologically between assemblages commonly attributed to Neanderthals and those attributed to AMHs and are present across Europe (Figure 1.1).

They are “transitional” in that they seem to combine various Middle Palaeolithic (MP) or Mousterian and Upper Palaeolithic (UP) or Aurignacian elements into a single behavioural system. Specifically, the Chȃtelperronian (extending from central France to northwestern Spain), the Uluzzian (in Italy and Greece), the Szeletian (in the Czech Republic and Hungary) and the Lincombian-Ranisian-Jerzmanowician (LRJ, present across northern Europe) are termed “transitional” based on their chronostratigraphic placement and cultural attributes. These transitional technocomplexes occur at various times in some areas of western Eurasia, generally exhibiting the expected chronological east-to-west pattern (Hublin, 2015). However, they are not present in some other regions of Europe such as the majority of the Iberian peninsula (Figure 1.1; but see comments the “transitional Aurignacian”

in Iberia (Bernaldo de Quirós and Maíllo-Fernández, 2009)). Furthermore, the presence of these transitional technocomplexes does not preclude the co-presence of late Middle Paleolithic assemblages in the same regions (Bobak et al., 2013; Higham et al., 2014;

Jaubert et al., 2011; Slimak, 2007). Most importantly, despite some insights into their behavioural characteristics (see §1.1) and chronological occurrence (see §1.2), little direct evidence is available for the hominin population(s) responsible for the transitional technocomplexes (see §1.3).

To complicate matters further, assemblages such as the Emirian (in the Levant), and the Bohunician and the Bachokirian (in central Europe) have been grouped together as Initial

2

Upper Palaeolithic (IUP) technocomplexes. Again, they follow a local MP or transitional technocomplex from a chronostratigraphic perspective but are suggested to be more fully rooted in UP technology compared to the transitional technocomplexes. The Szeletian in particular has been described as an MP acculturation in reaction to the arrival or development of the Bohunician (Nigst, 2012; Tostevin, 2007). The Szeletian and the Bohunician have significant geographic and chronological overlap (Škrdla, 2016), suggested to support such a scenario for the emergence of the Szeletian. This is in contrast to theoretical models advanced for the development of the Châtelperronian (Roussel et al., 2016). As with most transitional industries, these IUP industries are sparsely associated to hominin specimens. When they are present, such as at Üçağızlı Cave, they indicate an affiliation with AMHs (Kuhn et al., 2009).

The transitional technocomplexes are preceded by local MP industries such as the MTA, Keilmessergruppen and the Micoquian that are attributed to Neanderthals (Benazzi et al., 2014a; Gabori-Csank, 1983; Schmitz et al., 2002; Toussaint et al., 2010), including hominin specimens that are directly dated and genetically studied (Krings et al., 1997). The precise chronology of the transitional and IUP technocomplexes remains elusive (see §1.2), as does their biological attribution (see §1.3). Both the transitional assemblages and the IUP are replaced by Proto-Aurignacian or Early Aurignacian populations that are invariably linked to AMHs, supported by chronometric and genetic data (Benazzi et al., 2015; Nigst et al., 2014; Verna et al., 2012). These are sometimes described as Early Upper Palaeolithic assemblages (EUP) instead.

It must be noted that comparatively little is known about the “transitional” period in central and eastern Eurasia. Indeed, the existence of formal and well-defined analogous cultural technocomplexes has recently been re-addressed for these regions (Derevianko, 2011; Hoffecker, 2011; Zwyns et al., 2014). Examples include the identification of an IUP- related assemblage in central Mongolia (Zwyns et al., 2012). In part, this might be due to a low density of excavated sites and poor chronological control. As for western Europe, there are only a few hominin specimens associated with these assemblages, while the Ust’Ishim femur has been related to the IUP based on its direct AMS radiocarbon date (Fu et al., 2014). The recent report of AMHs in China by 80,000 BP (Liu et al., 2015) demonstrates that much remains to be learned about the cultural, biological and chronological arrival of AMHs in Eastern and Southern Eurasia and the local replacement (if any) of “archaic” hominins (Blinkhorn et al., 2013; Dennell, 2008; Groucutt et al., 2015; Timmermann and Friedrich, 2016).

For western Eurasia, the long history of research into the transitional period has resulted in a body of data large enough for some authors to propose explicit models on the behavioural origins of the transitional assemblages and their chronological relationship to Aurignacian assemblages. These models are dominated by inferences based on the Chȃtelperronian due to a prolonged research history into this technocomplex. Therefore, in the following a majority of archaeological, chronological and biological observations are based on the Châtelperronian. Generally, the proposed hypotheses can be divided into i) those that propose that Neanderthals are associated to a transitional industry and that they independently developed its behavioural characteristics, ii) those that propose that Neanderthals are associated to a transitional industry but acquired its behavioural characteristics directly or indirectly by contact with AMHs, or iii) the association of Neanderthals to a transitional industry and its behavioural characteristics is the result of taphonomic factors causing movement of artefacts or hominin fossils between Mousterian, transitional and/or Aurignacian layers. In the third (iii) case, the transitional industries would

3

be made by AMHs. The first two models state that the transitional technocomplexes were made by Neanderthals based on the presence of Neanderthal fossil specimens or indirectly through the interpretation of technological characteristics. If this would be disproven by direct biological evidence and supported by taphonomic control, then the transitional assemblages simply become a proxy for an earlier presence of AMHs. More complex scenarios such as dual authorship or “hybrid”-origin can be formulated based on genetic evidence for gene flow between Neanderthal and AMH populations (as tentatively hypothesized by Ruebens et al.

(2015) and d’Errico and Banks (2015)). They should not be excluded a priori, although they seem difficult to directly falsify based on the currently available data. In addition, the idea that all transitional technocomplexes would be related to either AMHs or Neanderthals does not need to be true (Hublin, 2015). The third hypothesis, however, proposes that taphonomic mixing of hominin specimens or cultural artifacts is the main cause of suggested attributions of Neanderthals to the transitional assemblages. This hypothesis needs to be refuted if one is to argue for either of the other two hypotheses.

These hypotheses were largely formulated after the biological assignment of a skeleton located in the Chȃtelperronian layers at Saint-Césaire to Neanderthals (Lévêque and Vandermeersch, 1980). This provides an interesting perspective on the importance of biological specimens in transitional contexts, as the Chȃtelperronian had previously been associated with AMHs through its technological and behavioural characteristics (Ashton, 1983; Bordes, 1972; Breuil, 1909; Peyrony, 1948). Following the publication of the Saint- Césaire hominin and the confirmation of a Neanderthal-Chȃtelperronian association by analysis of the inner-ear morphology of a Neanderthal temporal at the Grotte du Renne, Arcy-sur-Cure (Hublin et al., 1996), there was a need to explain this association in formal behavioural or cognitive models. The implications of these models regarding the Chȃtelperronian ranged beyond this particular technocomplex, as the Uluzzian in Italy was frequently interpreted as displaying a similar behavioural development based on the assignment of two Uluzzian molars excavated at Grotta del Cavallo as Neanderthals (but see comments below; Churchill and Smith, 2000).

After the establishment of the association between Neanderthals and some transitional assemblages, there was a general reinterpretation of these assemblages as more MP (e.g., Klein, 2009; Kozłowski, 1996; Otte, 1990; Valoch, 1990). After the publication of critiques of both proposed Neanderthal-Chȃtelperronian associations (Soressi 2011;

Higham et al. 2010) as well as the Neanderthal-Uluzzian association (Benazzi et al., 2014a), there has again be a theoretical shift that interprets these transitional lithic assemblage as more UP than MP (Bordes and Teyssandier, 2012; Roussel et al., 2016; Ruebens et al., 2015). Clearly, the presence and interpretation of hominin finds in transitional assemblages has a direct impact on the behavioural and chronological interpretation of those assemblages as well.

Because of the importance and research history associated with the Châtelperronian, three out of four chapters following will focus on this particular technocomplex (Chapters 2, 4 and 5). In addition, bone assemblages associated with the Châtelperronian are larger and more numerous compared to most bone assemblages associated with LRJ and Uluzzian assemblages, while few sites with bone preservation are known for the Szeletian. Below, several important behavioural and chronological aspects of the discussion regarding the makers of the transitional assemblages are presented. Following, an overview of the current biological evidence relating to this period and these European assemblages is presented.

4

Figure 1.1. Spatial distribution of the four main transitional technocomplexes and the approximate location of key sites in the MP-UP debate. Highlighted sites for the transitional assemblages include those playing a role in the biological attribution or those studied within this thesis (Les Cottés, Quinçay and the Grotte du Renne).

1.1 Behaviour

The discussion in relation to behavioural modernity present in transitional assemblages concerns cultural elements such as the presence of bone awls, pendants and minerals, the adoption of lithic systems aimed at blade and bladelet production, and those related to dietary adaptations.

Two sites containing Chȃtelperronian layers, Quinçay and the Grotte du Renne, have been proposed as transitional technocomplexes where ornaments such as perforated and grooved teeth are present (d’Errico et al., 2003; Granger and Lévêque, 1997). An additional lynx canine, possible pierced, is present in the Châtelperronian collections from Roc de Combe (de Sonneville-Bordes, 2002). In contrast, pierced shells as ornaments have only been reported from the Uluzzian site of Castelcivitta (Gambassini, 1997) while suggested personal ornaments from the LRJ at Ranis are now lost (Flas, 2011). Such objects are unreported from Szeletian contexts. Even for technocomplexes where such artefacts are present at some sites, they are not typical of the transitional assemblage as a whole.

Conversely, bone tools such as awls and bone points have been reported from numerous transitional assemblages (d’Errico et al., 2012, 2003; Peresani et al., 2016). Minerals, and the anthropogenic transport and modification of those minerals, are also widely distributed in transitional contexts (Dayet et al., 2014; Ruebens et al., 2015). They are not absent in European Middle Palaeolithic contexts, however, and so there role in debates specifically on the MUPT is limited (Heyes et al., 2016; Roebroeks et al., 2012).

The ecological context and faunal data on transitional assemblages remain poorly understood. In their assessment of a difference in hunting strategies from the Mousterian to

5

the Aurignacian in southwestern France, Grayson and Delpech (2002) compile a large overview of associated faunal assemblages for each technocomplex. They needed to exclude the Chȃtelperronian from their final analyses, because of a scarcity of Chȃtelperronian layers with representative amounts of identified bone specimens.

Nevertheless, Grayson and Delpech (2002) were able to provide significant statements on the absence of differences between the local Mousterian and the earliest regional Aurignacian (contra Mellars, 1973, 1996). The exclusion of the Chȃtelperronian highlights that an ecological perspective on this transitional assemblage itself was lacking, however.

This situation persists in the most recent studies (Discamps, 2014; Discamps et al., 2011).

The exception to these studies being those of Boyle (2007; 2010), where the author uses more bone assemblages but also includes those as small as three identified faunal specimens. If one only considers Châtelperronian sites with good contextual data (Ruebens et al., 2015), then 8 out of 15 such Châtelperronian sites contain faunal remains. All of these are cave sites (Cueva Morin, Grotte du Bison, Grotte du Renne, Quinçay, Les Cottés and Roc de Combe) or rockshelters (Saint-Césaire, Le Trou de la Chèvre). Some of these faunal collections have recently been (re-)analyzed, providing detailed information on Chȃtelperronian hunting strategies and seasonality (Morin, 2012; Soulier and Mallye, 2012).

Faunal analysis of bone specimens preserved at the open-air site of Les Bossats will provide a welcome open-air perspective on Chȃtelperronian subsistence strategies in the Paris Basin of northern France (Bodu et al., 2014). Such analysis is currently ongoing.

In the last decades there has been an increased understanding of regional Neanderthal MP technological variability (Faivre et al., n.d.; Ruebens, 2013), the development and production of standardized bone tools (Soressi et al., 2013), and their use of ochre, manganese dioxides and additional minerals (Heyes et al., 2016; Zilhão et al., 2010). Furthermore, Neanderthals were hunter-gatherers capable of acquiring both large and small mammals (Fa et al., 2013; Gaudzinski, 1995; Hardy and Moncel, 2011) that provided a large proportion of their dietary protein consumption (Naito et al., 2016; Richards and Trinkaus, 2009; Wißing et al., 2016), had diets that incorporated a substantial component of plant foods (Henry et al., 2014, 2011) and sometimes shellfish (Cortés- Sánchez et al., 2011), and perhaps obtained fauna for objectives other than direct consumption (Morin and Laroulandie, 2012; Peresani et al., 2011; Romandini et al., 2014).

Together, these cases have considerably narrowed the gap between a supposed cognitive difference between Neanderthal and AMHs, at least in material manifestations that can be observed in the archaeological record (Villa and Roebroeks, 2014). From this perspective, some, but not all, of the archaeological manifestations present in the transitional assemblages are within the Neanderthal behavioural repertoire observed in the preceding MP.

1.2 Chronology

Chronological arguments have played a prominent role in debates regarding the MUPT in the past two decades. The introduction of rigorous pretreatment chemistry for radiocarbon dating of bone (Higham et al., 2006) and charcoal (Bird et al., 2006), extended radiocarbon calibration curves (Reimer et al., 2013) and the use of Bayesian modeling (Bronk Ramsey, 2009a, 2009b), and improvements in tephrochronology (Lowe et al., 2012) have promised to provide increased chronological control. Combined with additional dating methods such as thermoluminescence, OSL and IRSL, these methods have been used to test the hypothesis

6

of regional contemporaneity of transitional and UP assemblages. In addition, chronological data has been used to estimate the presence of younger or older intrusive elements within transitional assemblages.

Nevertheless, considerable discussion has recently arisen concerning the extent to which archaeological models should inform radiocarbon sample selection and chronological modeling, or vice versa (Caron et al., 2011; Discamps et al., 2015; Gravina and Discamps, 2015; Pettitt and Zilhão, 2015). The debate ensuing from the dating of the Grotte du Renne stratigraphy by Higham et al. (2010) is a case in point. Proposing that the Neanderthal- Chȃtelperronian association at this site is the result of taphonomic mixing, this publication started a heated exchange between proponents of the independent development and acculturation hypotheses (Caron et al., 2011; Hublin et al., 2012; Zilhão et al., 2011) and those disagreeing (Higham et al., 2012, 2011a, 2011b). Similar exchanges concerning the chronology of the Proto and Early Aurignacian (Banks et al., 2013a, 2013b; Higham et al., 2013) demonstrate the diverse interpretations one can put forward based on chronological data.

Direct dating of hominin specimens associated with the transitional period would provide direct chronological control and an estimate on the stratigraphic position of the studied specimens through Bayesian modeling. Such direct radiocarbon dates do currently not exist because of the scarcity of hominin specimens attributed to transitional contexts.

The possible exceptions are the Saint-Césaire and Spy hominins (but see below). The strength of this approach has been demonstrated in other cases where supposed Palaeolithic hominin specimens were directly dated and found to be Holocene in age (Benazzi et al., 2014b; Conard et al., 2004; Hoffmann et al., 2011; Terberger et al., 2001).

These specimens were excavated before modern recording methodology was commonplace (except for Hahnöfersand, which was found out of context; Terberger et al., 2001), as is the case for the majority of transitional faunal assemblages present in museum collections.

These findings highlight the importance of providing direct chronological control for such assemblages and/or for hominin specimens contained within them.

For some authors, the hypothesis that Neanderthals made the transitional technocomplexes must be based on a prolonged coexistence of Neanderthals and AMHs in a restricted geographical area. Initially, claims in favour for this were based on interstratification of transitional and Aurignacian (Early or Proto) layers at several key sites (Bordes and Labrot, 1967; Champagne and Espitalie, 1967; Gravina et al., 2005). Genuine interstratifications have now been rejected for all sites involved in this debate (Bordes, 2002;

Riel-Salvatore et al., 2008; Zilhao et al., 2008; Zilhão et al., 2006), and proponents of the acculturation hypothesis currently favour the exchange of ideas over longer distances through concepts such as stimulus diffusion (Nigst, 2012; Roussel, 2013, Tostevin, 2007).

These are generally supported by chronological models on a continent-wide scale which demonstrate the presence of transitional assemblages and Proto/Early Aurignacian assemblages in the same chronological window (e.g., d’Errico and Banks, 2015; Higham et al., 2014). On a regional or local scale, however, there is often little high-quality data to confirm or refute such models based on chronological data alone.

1.3 Biology

Obtaining a coherent biological understanding of the transitional period should be obtained independently of cultural inferences when the transfer of cultural behaviour between different

7

biological groups is probable or likely. There therefore is a need to identify hominin specimens in transitional archaeological contexts, and to obtain biomolecular data on the genetic ancestry of these specimens. Unfortunately, few remains of hominins have been directly associated with transitional assemblages (Churchill and Smith, 2000), and in most cases these associations have been disputed based on chronological or archaeological arguments.

Direct associations between hominin fossils and transitional technocomplexes have been proposed for each of the four “classical” technocomplexes introduced above (Figure 1.1). These include 1) the Spy Neanderthals, chronologically fitting with the LRJ but placed in an archaeological context including both LRJ and late Mousterian artefacts (Semal et al., 2009); 2) a maxilla potentially associated with the LRJ at Kent’s Cavern, morphologically undiagnostic for some authors but AMH for others (Higham et al., 2011c); 3) the Saint- Césaire Neanderthal, chronologically fitting with the Châtelperronian but placed in an archaeological context including Châtelperronian and late Mousterian artefacts (Hublin et al., 2012; Lévêque and Vandermeersch, 1980; Soressi, 2011); 4) the Grotte du Renne hominins, not directly dated and from an archaeological context that has been extensively debated (Bailey et al., 2009; Bailey and Hublin, 2006; Higham et al., 2010; Hublin et al., 2012, 1996);

5) two teeth from Cavallo Cave (Cavallo B and C), originally classified as Neanderthals but recently re-attributed to AMHs and potentially from a disturbed Uluzzian context (Benazzi et al., 2014a; Ronchitelli et al., 2014; Zilhão et al., 2015); 6) a tooth from the Grotta di Fumane (Fumane 6) with unclear taxonomic affinities, also from a disturbed Uluzzian context with Proto-Aurignacian intrusions (Benazzi et al., 2014a); and 7) a molar from Dzeravá Skala, from a cryoturbated Szeletian context and of unclear taxonomic affiliations (Hillebrand, 1914;

Kaminská et al., 2004).

Except for the Fumane 6 specimen, all these hominin specimens were excavated before modern excavation techniques were employed. The Spy and Saint-Césaire hominins have been dated directly, providing some chronological control (Hublin et al., 2012; Semal et al., 2009). The direct dates available for the Spy hominins in particular present a convincing case for an indirect association with the regional LRJ, as these dates fall outside those available for the local Mousterian, but within those available for LRJ assemblages dated elsewhere, and before the presence of the Aurignacian in the region (Flas, 2014, 2011). In contrast, a direct radiocarbon date for the Kent’s Cavern maxilla (Hedges et al., 1989) has been dismissed (Higham et al., 2011c). This recent attempt to provide chronological control through Bayesian modelling of bone specimens contextually related to the Kent’s Cavern maxilla (Higham et al., 2011c) has been criticized as well (White and Pettitt, 2012). The morphological affinity of the specimen is not entirely clear, while attempts to obtain ancient DNA were unsuccessful. The Spy Neanderthals might be seen as better biological candidates for the makers of the LRJ.

To this small body of specimens suggested to be associated with transitional archaeological contexts, one can add several specimens from Central Europe (Oase 1 and 2), and Central Siberia (Ust’Ishim) that are chronologically placed in the same period but that are not associated directly to an archaeological technocomplex (Fu et al., 2014; Trinkaus et al., 2003). Genomic analysis of the Oase 1 and Ust’Ishim individuals, as well as an additional genome from Kostenki 14 (Eastern Europe), has revealed that they represent AMHs that contributed variable proportions of their genetic heritage to present day modern humans, with the Oase 1 individual contributing very little or none (Fu et al., 2015, 2014;

Seguin-Orlando et al., 2014).

8

Recent genomic analysis of Pleistocene AMHs, Neanderthals and Denisovans has indicated that there has been an exchange of genetic information between all three clades (Green et al., 2010; Kuhlwilm et al., 2016; Meyer et al., 2012), in addition to gene flow from additional hominins currently unanalyzed genetically and potentially unknown morphologically (Meyer et al., 2012). The timing and extent of gene flow events within Eurasia varies between the specific exchange studied, with that from Neanderthal gene flow into AMHs occurring roughly between 60-40ka (Green et al., 2010). This estimate fits with most archaeological and chronological evidence regarding the dispersal of AMHs out of Africa or southwestern Asia. Evidence for the reverse, gene flow from AMHs into Neanderthals or Denisovans, is present in eastern Neanderthals (Kuhlwilm et al., 2016) but not in western Neanderthals. This is of importance when considering the geographic positioning and chronology of the transitional technocomplexes in western Europe, but could also be due to the absence of complete genomes of late Neanderthals from this region and time period. Future ancient DNA analysis of late Neanderthals associated with transitional technocomplexes might confirm this absence, in which case differences between late Mousterian and transitional assemblages need to be explained by (in)direct social interaction with AMHs or independent development but without requiring the transfer of genetic information. Unfortunately, because of a scarcity of hominin specimens in general and the obvious importance of the possible transitional hominin candidates mentioned above there is currently no genetic data available for any of these specimens.

From a biological perspective there is a clear lack of available hominin samples in transitional contexts that could be used for biological inferences (ancient DNA, ancient proteins) and chronological control. Obtaining additional hominin fossils is therefore of prime importance, and screening techniques that could cost-effectively provide taxonomic identifications for morphologically unidentifiable bone and tooth specimens are therefore of interest to archaeologists, archaeological scientists and geneticists. Both aDNA screening and ZooMS (Zooarchaeology by Mass Spectrometry) would be methodological candidates.

In fact, the draft Neanderthal genome was based on aDNA screening for Neanderthal bone specimens at Vindija (Green et al., 2010). Ancient DNA is relatively expensive, however, while ZooMS screening is known to be cheap, quick and less prone to biomolecular contamination (but see §2.3; Green et al., 2010). Furthermore, ancient proteins degrade at a slower rate compared to ancient DNA (Demarchi et al., 2016) and might therefore provide a suitable second biomolecular approach to ancient hominin taxonomy. Therefore, the work presented in this thesis focuses on i) developing a research framework in which ZooMS screening aims to identify additional transitional hominin specimens, and ii) exploring the potential of ancient proteins in relation to hominin taxonomy.

2. Proteins in archaeology, anthropology and palaeontology

The study of ancient proteins through mass spectrometry has seen increasing applications in recent years due to advances in mass spectrometry hardware and software (Cappellini et al., 2014). Mass spectrometry is now the method of choice when studying ancient proteins, compared to amino acid profiling or immunoassays, primarily because it allows the identification of amino acid sequences directly (Cappellini et al., 2014; Hendy et al., 2016).

Applied to a wide range of questions and tissue substrates, ancient protein analysis follows a workflow comparable between most studies where the primary aim is to obtain protein

9

identifications (Figure 1.2). These studies either utilize MALDI-TOF-MS or LC-MS/MS to obtain spectral data. Spectral data obtained through either mass spectrometry approach is then analyzed using a variety of bioinformatic tools to obtain peptide/protein (sequence) identifications (mMass, Byonic^TM, PEAKS, MASCOT, MaxQuant).

An extensive review of ancient protein analysis is outside the scope of this chapter, and some excellent reviews have recently been published (Cappellini et al., 2014;

Dallongeville et al., 2016; Vinciguerra et al., 2016; Warinner et al., 2015). Instead, the concepts of ZooMS (Zooarchaeology by Mass Spectrometry) and palaeoproteomic analysis will be introduced below as both methodological approaches provide distinct possibilities for the study of the MUPT. Next, an extensive overview of current approaches to prevent and identify protein contamination is given. The latter is an important point when studying ancient hominin or human specimens but rarely discussed extensively in the palaeoproteomic or clinical literature (Griss et al., 2016). In some ways, this is surprising as the field of ancient DNA struggled with contamination issues during the 1990’s and early 2000’s (Gilbert et al., 2005; Poinar and Cooper, 2000). Similar discussions have rarely taken place in the palaeoproteomic literature, those related to possible dinosaur proteins being an exception (Asara et al., 2008, 2007; Bern et al., 2009; Buckley et al., 2008; Pevzner et al., 2008;

Schweitzer et al., 2007). As there is a growing body of literature on ancient human and hominin tissues, it seems justified and timely to discuss current problematics regarding protein contamination in palaeoproteomic analysis.

Figure 1.2. Generalized analytical scheme for the study of ancient proteins through proteomics.

2.1 ZooMS and COL1

ZooMS (Zooarchaeology by Mass Spectrometry) has been developed as a taxonomic screening method for mammalian tissues rich in collagen type I (Buckley et al., 2009).

Collagen type I is a triple-helical protein composed of two COL1α1 and one COL1α2 α- chains that wind around each other (Cowan et al., 1955; Rich and Crick, 1955). Hereafter, this triple helix will be named COL1. COL1 is the dominant bone protein, making up roughly 90% of the protein content of living bone and preferentially preserved over time compared to other proteins (Wadsworth and Buckley, 2014). The function of a protein is determined by its structure, which is ultimately reliant on the chemical properties of the amino acid side chains and the order of these side chains, and COL1 is no exception. As an extracellular molecule COL1 functions to provide structural support to connective tissues by agglomerating into larger COL1 fibrils and fibers, COL1-COL1 crosslinking and cross-linking to other proteins (Kalamajski and Oldberg, 2010; Orgel et al., 2006). The structural requirement for these various functions is that COL1 forms a triple-helical structure. The only way to achieve this from a protein point-of-view is by incorporating a glycine (G) at every third amino acid residue, as a glycine is the only amino acid without a side chain and therefore the only amino acid physically small enough to fit on the inside of the COL1 helix. The resulting motif (sometimes called a triplet), G-X-Y, is repeated ≈300 times along all three α-chains. In the G-

10

X-Y triplet, X and Y are most commonly proline (P) and hydroxyproline (Hyp), respectively, which further stabilize the triple helix (Shoulders and Raines, 2009; Vitagliano et al., 2001).

As there is seemingly little structural or functional requirement for the other X and Y positions, nonsynonymous substitutions at the DNA level are not selected against and are free to mutate over evolutionary time. In contrast, nonsynonymous substitutions at glycine- containing positions are related to connective tissue disorders such as osteogenesis imperfecta and Ehlers-Danlos syndrome and therefore undergo negative, purifying selection (Marini et al., 2007; Nuytinck et al., 2000). The nonsynonymous substitutions in X and Y coding positions at the DNA level are of interest to ZooMS and phylogenetic analysis of COL1 protein sequences as these cause different amino acids to be incorporated into the X and Y position of the G-X-Y triplets. Variation at such sites can become fixed during speciation events, and thereby have the potential to accurately depict the phylogenetic relationships between members of the Kingdom Animalia (formerly Metazoa).

The ZooMS method relies on trypsin digestion of COL1 extracts and proteomic analysis through the generation of peptide mass fingerprints (PMF’s) by MALDI-TOF-MS analysis. Next, the method analyzes the mass (m/z) of eight selected peptides in these PMF’s, termed peptide markers. The mass of individual peptide markers, or the combination of masses observed for multiple peptide markers, then results in a taxonomic identification through comparison with a peptide marker library composed of relevant species. Such peptide marker series are available for an increasing range of species (Buckley et al., 2014, 2009; Buckley and Kansa, 2011; Campana et al., 2013; Kirby et al., 2013), and generally indicate that the obtained taxonomic identifications are in the range of subfamily or genus level. Nevertheless, for the Late Pleistocene not all relevant species are present in the currently available databases. Furthermore, most peptide marker masses are unsupported by COL1 amino acid sequence data for the same species, meaning that identifications are solely based on the observation of a mass (rather than related to a known amino acid peptide sequence) and cannot be verified by subsequent LC-MS/MS analysis. Within the context of this research, a COL1 sequence database coupled with ZooMS PMFs for the Late Pleistocene of western Eurasia is built to amend these omissions, allowing confident ZooMS identifications and subsequent LC-MS/MS verification for the region and time period of interest.

One distinct advantage of ZooMS is the small sample size required, generally below 30 mg, which can be analyzed using different extraction techniques that are compatible with MALDI-TOF-MS analysis. Acid demineralisation (Buckley et al., 2009), semi-destructive ammonium-bicarbonate buffer extraction (AmBic; (van Doorn et al., 2011), and non- destructive eraser sampling (eZooMS; Fiddyment et al., 2015) have all been used as effective approaches to obtain suitable ZooMS PMF’s. Especially AmBic seems advantageous for large-scale applications as it is compatible with subsequent aDNA analysis of the same specimens (von Holstein et al., 2014), has a limited processing time, and provides reliable glutamine deamidation ratios (van Doorn et al., 2012; Wilson et al., 2012).

The latter would allow assessing the likelihood of a modern origin of the obtained ZooMS PMF (van Doorn et al., 2012). Although not employed yet, initial AmBic screening results could also be verified by subsequent acid demineralisation of the same specimens (see chapters 4 & 5).

ZooMS has been applied to worked artefacts (Ashby et al., 2015; Fiddyment et al., 2015; Ives et al., 2014; Kirby et al., 2013; von Holstein et al., 2014), to understand local extinction and recolonisation processes around the Last Glacial Maximum (Meiri et al., 2014), to differentiate between closely related species for which morphological identifications

11

are difficult (Buckley et al., 2011; Buckley and Kansa, 2011; Evans et al., 2016), and to determine the species composition of morphologically unidentifiable bone assemblages (Brown et al., 2016; Charlton et al., 2016). Furthermore, because of costs, sample requirements and simplicity, ZooMS has been applied as a cost-effective screening method before applying more expensive or invasive analysis (Meiri et al., 2014; von Holstein et al., 2014). Alternative approaches have been proposed for fish (Richter et al., 2011), tortoises (Van der Sluis et al., 2014), eggshell (Stewart et al., 2014, 2013), and hair and feathers (Hollemeyer et al., 2012). These differ either in the (combination of) target protein(s), in the computational approach to come to a taxonomic identifications, or in the taxonomic group considered.

2.2 Palaeoproteomics and the bone proteome

In contrast to MALDI-TOF-MS, tandem mass spectrometry (LC-MS/MS or MALDI-TOF/TOF- MS) has the ability to obtain protein amino acid sequence information for individual peptides.

Such a bottom-up approach requires more extensive investment in protein extraction protocols and bioinformatic analysis as individual MS/MS spectra need to be correctly identified when compared to a protein sequence database.

Palaeoproteomic analysis has several distinct advantages compared to MALDI-TOF- MS analysis, such as the identification of individual proteins in a protein mixture (Cappellini et al., 2012; Warinner et al., 2014b), the assignment of these proteins to specific taxonomic origins based on amino acid sequence information (Cappellini et al., 2012), the localization of amino acid substitutions compared to a reference proteome (Buckley et al., 2011), the analysis and localization of physiological and diagenetic protein post-translational modifications (PTMs; Cleland et al., 2015; Mikšík et al., 2016, 2014), and potentially to study quantitative differences in protein composition between samples (Cappellini et al., 2010;

Fiddyment et al., 2015).

Not surprisingly, with methodological improvements in tandem mass spectrometry hardware and software there has been a steady increase in the application of these methods to archaeological and palaeontological questions. For example, LC-MS/MS analysis has allowed ancient tissues representing protein mixtures to be explored (dental calculus, coprolites (Warinner et al., 2014a, 2014b)), and to investigate the ancient bone proteome beyond COL1 (Buckley and Wadsworth, 2014; Cappellini et al., 2012; Orlando et al., 2013;

Wadsworth and Buckley, 2014).

The localization of amino acid substitutions compared to a reference proteome has been suggested previously as a useful tool in building protein phylogenies for extinct species (Buckley et al., 2011). The promise of a proteomic approach compared to ancient DNA being that proteins are more stable biomolecules and, therefore, available for analysis over longer time scales. Tissues only provide access to a part of the complete proteome from any given species, however, and these proteomes will be further restricted due to protein degradation over geological time (Wadsworth and Buckley, 2014). Especially in the light of prolonged protein preservation, some work has focused on COL1 sequencing (Buckley, 2013; Buckley et al., 2011) from ancient tissues to obtain phylogenetic information. Although a proof of concept, these papers do not provide evidence that the correct COL1 sequence is obtained - i.e. there is no demonstration that de novo/error-tolerant LC-MS/MS analyses result in a COL1 sequence already obtained through genetic analysis for the same species. Chapter 3 addresses this important caveat.

12

Palaeoproteomics has been applied to human remains from Holocene time periods to investigate oral health (Warinner et al., 2014a, 2014b) and infectious disease (Hendy et al., 2016; Kendall et al., 2016). Application to Pleistocene specimens has been limited to two MALDI-TOF/TOF-MS analyses of bone and dental proteins, respectively, from Neanderthals (Nielsen-Marsh et al., 2009, 2005). This is surprising, as protein immunology played an important role in establishing the place of our species among the great apes (Sarich and Wilson, 1967). Currently, there is a growing awareness of the limited availability of ancient DNA at environmentally unfavourable sites in the Late Pleistocene, or for the Middle Pleistocene in general (Meyer et al., 2016; Orlando et al., 2013). Available genomes and exome sequences from Neanderthals (Castellano et al., 2014; Prüfer et al., 2014), Denisovans (Meyer et al., 2012), and AMHs (1000 Genomes Project Consortium et al., 2015; Lek et al., 2016; Sudmant et al., 2015), however, suggest there are nonsynonymous substitutions present in several proteins commonly identified using LC-MS/MS. Among these are proteins on the archaic lineage that are significantly derived and involved in (bone) morphology and development which would be prime candidates for a protein-based approach to Late Pleistocene taxonomy of hominin specimens. Together with the above- mentioned capability of de novo/error-tolerant ancient protein sequencing, this holds enormous potential in contexts where ancient DNA might not be preserved. The final chapter of this thesis provides an initial exploration of this potential in a context where ancient hominin DNA has survived.

Figure 1.3. Schematic depiction of bone proteome compositional changes over time. Initially, the bone proteome is solely composed of endogenous proteins (green). After deposition the bone will be colonized (rapidly) by bacteria and fungi (orange). During and after excavation additional contamination can occur, primarily due to human handling and through protein-based consolidates (for example human keratins or animal-based glues; in red). A definitive source of contamination is introduced during protein extraction by proteins either present in purchased solutions (for example trace amounts of BSA), or deliberately added (Trypsin).

Throughout the scheme, endogenous bone proteome complexity and concentration decrease. Conversely, there is an increase in the proportion of contaminating proteins, both of vertebrate and non-vertebrate origin.

2.3 Protein contamination

The colonisation of bones and teeth by bacteria, fungi, etc. during deposition as well as the probable introduction of contamination due to human handling of archaeological and palaeontological specimens means that extracted bone proteomes are mixtures of diverse taxonomic origins (Figure 1.3). The addition of bacterial components to a bone proteome occurs at different stages, and some of these might be of interest for metaproteomic analysis as these could represent components of the burial environment in which the bones were deposited (Figure 1.3). It is likely that additional vertebrate proteins are added during the relatively recent history by handling or during protein extraction. It has recently been realized

13

that protein contamination is a source of spectral misidentification in clinical proteomic studies, further highlighting the attention protein contamination in palaeoproteomics should receive (Griss et al., 2016). Protein identification workflows should therefore allow the identification of such taxonomic mixtures, as is commonly the case when analysing dental calculus or coprolite samples (Warinner et al., 2014b).

Measures to prevent contamination are not sufficient to avoid these proteome mixtures, especially when both the tissue analyzed and the contaminating source are humans/hominins. For hominins this creates an analytical problem after determining the protein composition. Analytical tools should be developed that aim to resolve this problem, for example by studying diagenetic changes on a protein-by-protein basis or at the spectrum/peptide level. The complexity of the issue arises from a combination of i) a decrease in endogenous protein concentration due to protein degradation, ii) a decrease in endogenous protein composition complexity, iii) an increase in protein PTM complexity by the addition of diagenetically derived PTMs, and iv) an increase in exogenous protein contamination from vertebrate and invertebrate sources. The latter also causes an increase in taxonomic origins (Figure 2).

Palaeoproteomic workflows aiming to investigate ancient proteomes either take measures to prevent protein contamination, take measures to identify contaminating proteins or take measures to identify endogenous proteins. Preferable, all three possible avenues are used and discussed, taking into account that the null hypothesis should be that any identified protein represents a contaminant. In some published cases this is indeed true, while in others protein contamination is barely touched upon. This might be because of a lack in reporting, but could also represent a true absence in followed workflows.

Measures to prevent protein contamination during analytical steps are regularly employed and described in the palaeoproteomic literature. These include basic precautions such as wearing latex or nitrile gloves, usage of sterilized equipment and the usage of analytical-pure chemicals. In addition, several studies mention running “blanks” between LC- MS/MS analysis of samples to reduce the risk of protein carryover between samples (Table 1.1). Fortunately, there seems to be a general awareness among scholars, reviewers and editors that such measures should be employed and reported.

Next to measures to prevent protein contamination during analytical steps, several approaches have been proposed in the palaeoproteomic literature to demonstrate either the endogenous or the exogenous origin of an identified protein or peptide in a protein mixture (Table 1.1). The application of these approaches is varied and not widely shared between papers.

First, there are measures to monitor the addition of contaminating proteins during analyses, normally by employing an “extraction blank”. Proteins identified through such means are removed from further analyses when present in sample protein lists. Using an extraction blank is a powerful approach to identify contaminating proteins. A variant on this concerns the extraction of proteins from sediment samples as an external control to identify soil bacteria, fungi or arthropod proteins. Additional means to identify such proteins is by using an a priori list of proteins suspected to represent contaminating proteins. Some of these are known contaminants as they are added during the analytical process (Trypsin), are present as residual contaminants in purchased solutions (BSA, possibly bovine ALB), or are likely to derive from human handling (keratins). Employing such exclusion lists is useful in some cases, but could become problematic when analysing tissues that have a protein content known to be enriched in such proteins. For example, this would concern skin samples (containing keratins) or blood vessels (BSA, ALB, histones). Any of these

14

approaches require that bioinformatic workflows allow matches to non-endogenous proteins to be made confidently, either through adding a priori contaminant lists or by searching to large protein sequence databases such as UniProt. Literature that does not specify if such protein matches were facilitated, and that do not explain why not, is suspicious. Obviously, so are studies that do not employ or analyze an extraction blank.

Second, after exclusion of proteins present in various controls and contaminant protein lists, the composition of the remaining proteome is often used as demonstrating the endogenous origin of the identified proteome. The extent to which this is explicitly tested varies between sources (Table 1.1) and requires the availability of reference proteomes from modern sources for the same tissues. These are available for important mineralized tissues such as bone, cementum, dentine and enamel (Alves et al., 2011; Jágr et al., 2012; Jiang et al., 2007; Salmon et al., 2013). It should be realized, however, that in some of these studies very different extraction procedures are used and that these studies cannot always exclude the possibility that significant amounts of bone marrow, blood vessels, periosteum and endosteum were included. The presence of proteins unique to specific tissues (Hendy et al., 2016; Maixner et al., 2013) or cellular types/stages (Cappellini et al., 2014) is another criterion to argue for the endogenous origin of a given protein. These would represent the strongest possible case that could be presented in a palaeoproteomic paper, especially when supported by the presence of diagenetic alterations (see below). Such argumentation will only be available for a subset of the proteins identified within a proteome, as most proteins are present in multiple cell or tissue types. The presence of proteins unique to a single tissue does not directly support the endogenous origin of other proteins identified in the same proteome, but does provide a strong indirect argument.

The presence of non-human amino acid sequences has also been mentioned in this context as the presence of such sequences suggests these do not derive from human contamination. Although powerful for vertebrate palaeoproteomes especially, this approach is likely not applicable when studying ancient human/hominin proteomes. For such studies, alternative arguments must be provided.

COL1 studies, such as those employing ZooMS through MALDI or those extracting bone proteomes with a focus on COL1 for phylogenetic purposes through LC-MS/MS analysis, often defend the origin of their detected collagen peptides by stating that COL1 is the dominant bone protein and therefore unlikely to derive from exogenous contamination.

This is a legitimate argument, but it should not be forgotten that COL1 is also present in other tissues, including the dermis (Lovell et al., 1987; Smith, 1994). The latter could be considered to be a possible source of contamination, although COL1 concentration is low in the outer layer of the skin (the epidermis). Just as important, studies employing large-scale ZooMS screening would be advised to include extraction blanks to demonstrate the absence of COL1 contamination through intensive use of lab space or (shared) chemicals, especially when large amounts of bone extracts are processed simultaneously. Currently, this is rarely reported in the ZooMS literature.

Third, there are various studies that identify and report diagenetic PTMs in ancient proteomes (Table 1.1). The assumption made is that diagenetic alterations should increase during deposition or burial and should be distinctly different (either in nature or extent) from those encountered in modern samples from the same tissues. Ideally, such comparisons are made between comparable extraction protocols as it is known that these can induce diagenetic changes themselves (Hao et al., 2011; Ren et al., 2009; Simpson et al., 2016).

Among others, those discussed in the contemporary literature are deamidation of glutamine and asparagine, oxidation of methionine, and the presence of non-enzymatically cleaved

15

peptide sequences (interpreted as representing peptide backbone cleavage prior to protein extraction). In addition to these, there is a growing awareness that protein diagenesis is varied and can be studied confidently in ancient samples (Cleland et al., 2015; Hill et al., 2015). Currently, the analysis of diagenetic PTMs often focuses on a single protein within a proteome (for example IgG in Kendall et al. 2016 or COL1 in Orlando et al. 2013) or on proteome-wide levels (Warinner et al., 2014b). Although very useful in their own right, such approaches do not allow classification of all individual proteins as exogenous or endogenous. Instead, the palaeoproteomic community might benefit from studying these types of modifications on a protein-by-protein basis for all the proteins present in a palaeoproteome. This would allow assessment of the extent and type of modifications caused by the extraction method employed and provide a (semi) quantitative basis of classifying proteins as exogenous or endogenous (including those regarded a priori as representing contaminants).

A combination of measures to prevent and detect protein contamination in palaeoproteomics is advisable. Unfortunately, there are palaeoproteomic papers that do not mention “contamination” or “controls” when describing their analytical workflow or palaeoproteomic results, or only describe some measures taken to avoid contamination. The claims made in such papers are difficult to evaluate, and replication studies might not be able to support previous studies (for an example, see Kendall et al., 2016).

16

Table 1.1. Measures to prevent and/or identify protein contamination in palaeoproteomic studies on human/hominin tissues. Various approaches have been reported to prevent or to identify protein contamination from non-endogenous sources.

They can be separated in measures to 1) prevent protein contamination, 2) identify exogenous proteins, and 3) identify endogenous proteins. Only studies employing LC-MS/MS or MALDI-TOF/TOF-MS are included in the table. Studies focusing on non-human/hominin tissues are not included in the table, but display a similar diversity in the approach taken concerning protein contamination. The use of blanks between analytical runs is not applicable (NA) to MALDI-TOF/TOF-MS studies as samples are located on separate MALDI plate spots.

Contamination

prevention Contamination detection

Study Tissue type Species

Extraction blanks

Blank runs

Contaminating protein list

Proteome composition

Deamidation (N/Q)

Semi- tryptic peptides Nielsen-Marsh

et al. 2005 Bone Neanderthal NA X

Nielsen-Marsh

et al. 2009 Enamel Neanderthal NA X

X (one individual

protein)

X (one individual

protein) Boros-Major et

al. 2011 Bone

Human / Mycobacterium

tuberculosis

NA X

Corthals et al.

2012

Blood/saliva

deposits Human X X

Hajdu et al.

2012 Bone

Human / Mycobacterium

tuberculosis

NA

Maixner et al.

2013 Brain tissue Human X X

X (global proteome levels) Kendall et al.

2016 Bone Human X X

X (one individual

protein)

X (one individual

protein) Mikšík et al.

2014 Bone Human X X

X (one individual

protein) Bona et al.

2014

Osteogenic

sarcoma Human X X

Warinner et al.

2014

Dental

calculus Human X X X X

X (total proteome

levels)

X (total proteome

levels) Warinner et al.

2015

Dental

calculus Human X X X

X (one individual

protein) Hendy et al.

2016 Lung tissue

Human / Mycobacterium

tuberculosis

X X X X (textual

reference) Mikšík et al.

2016 Bone+Muscle Human X X (textual

reference) Brown et al.

2016 Bone Neanderthal X X

17

3. Aims and objectives

The research presented in this thesis aims to provide a proteomic workflow that allows the screening of large numbers of morphologically unidentified bone and dental specimens by ZooMS at sites containing transitional assemblages in western Europe, followed by subsequent palaeoproteomic analysis of the same or additional protein extracts (Figure 1.4).

Archaeologically, the work focuses on the Châtelperronian as this technocomplex is central to existing theoretical replacement scenarios. Throughout these analyses, an adequate understanding of protein contamination, and where possible contamination identification and exclusion, must be presented. Specific objectives are:

● Demonstration of the feasibility of performing ZooMS screening on medium-sized transitional faunal assemblages with a positive success rate of identifications (Chapters 2 and 4).

● Development of a research methodology by which correct COL1 sequences can be obtained through de novo/error-tolerant LC-MS/MS analysis from modern and Pleistocene bone specimens (Chapter 3).

● Development of adequate measures to identify contaminating proteins (bacterial, vertebrate, human) in hominin palaeoproteomes (Chapters 4 and 5).

● Exploration of the use of amino acid sequence data to distinguish between Late Pleistocene clades of the genus Homo (AMH, Neanderthals and Denisovans) without the need to resort to ancient DNA (Chapter 5).

These objectives will be explored in the following four projects.

Figure 1.4. Proteomic workflow presented in this research for the identification and utilization of hominin bone specimens in transitional contexts. ZooMS and MALDI-TOF-MS analysis provide initial taxonomic identifications, allowing exploration of faunal assemblage species composition and the identification of hominin specimens. Next, these hominin specimens become available for in-depth multidisciplinary analysis, including LC-MS/MS analysis.

3.1 Project 1: ZooMS screening at Les Cottés, France

The first project aims to demonstrate the feasibility of performing ZooMS screening on a medium-sized transitional faunal assemblage for which no morphological identifications were available. In order to achieve this goal, the project aims to:

● Provide a blind test on bones with known morphological identifications from the same site (Les Cottés) to demonstrate that ZooMS and morphological identifications are in agreement with each other;