Neanderthals likely utilised behavioural adaptations to their new Arctic environment (Roebroeks et al. 2021; Zilio et al. 2021). Behavioural thermoregulatory strategies—namely fire, clothing, tool use, food storage, and dwelling construction—are widespread in humans (Ilardo and Nielsen 2018) and have even been observed in West African chimpanzees (Pan troglodytes verus) (Pruetz 2007, Pruetz 2018). In addition to behavioural cold buffers, Neanderthals likely possessed inherited biological adaptations to their cold environment. So far, morphological adaptations are those with the most robust support (Holliday 1997; Weaver 2003), but there is limited evidence for energetic thermoregulatory adaptations, such as high basal metabolic rates and diets designed to harness the thermic effects of digestion (Snodgrass and Leonard 2009). Most physiological adaptations remain chiefly theoretical, as it is challenging to obtain evidence for non-skeletal biology from the archaeological record (Ocobock et al. 2021).

In contemporary humans, biological thermoregulatory adaptations are not restricted to morphology and physiology. Inherited molecular-level adaptations, such as those that help Inuit cope with low-carbohydrate diets (Hale 2020), are also essential. I dissect pertinent examples in the section entitled Haemoglobin Adaptation in Order Primates. Molecular-level thermoregulatory adaptations in contemporary human populations imply that similar adaptations could have existed in the Neanderthals—our closest relatives.

Not all research aligns with the cold-adapted hypothesis (Rae et al. 2011). One recent study by Stewart et al. (2019) proposed that morphological differences between Neanderthals and modern humans may be explained as adaptations for an active lifestyle and that Neanderthals’ preferred environments were, in fact, more temperate than previously thought. However, climatic data strongly support glacial conditions for a majority of Neanderthal sites and palaeozoological evidence consistently associates them with glacial fauna (Bar-Yosef 2004). Additionally, no modern climate analogue exists for the mammoth steppe tundra that dominated the Eurasian landscape at this time (Zimov et al. 2012) and our ecological models, on which Stewart et al. (2019) rely, struggle when applied to such historical environments because they are parameterised using modern climate data.

Furthermore, the argument put forward by Stewart et al. (2019) that Neanderthals’ morphological differences are adaptations for an active lifestyle largely depends on their identification of genomic variants associated with power and endurance in modern humans. They believe these alleles indicate that Neanderthals were proficient at athletic performance and adapted to an active lifestyle. However, alleles that support athletic prowess and power in modern humans may not necessarily have had the same phenotypic effects in Neanderthals.

Stewart et al. (2019) additionally presuppose that a species will not be driven to evolve cold-specific adaptions unless its environment is consistently extremely cold. They argue that, due to the warmer interglacial periods experienced by Neanderthals, selection would not drive them to develop such adaptations. The underlying assumption here is flawed—thermoregulatory adaptations can benefit an organism even if harsh lows only occur periodically (Campbell et al. 2010). So long as the trait does not disadvantage the organism and lower its fitness during times of warmth, then it will provide an overall selective advantage (Schou et al. 2022). All evidence considered, cold adaptation in Neanderthals remains plausible.

The human globin genes are organised into two clusters located on separate chromosomes, the alpha-globin cluster and the beta-globin cluster (Bahr and Ohls 2022). Neanderthal globin genes have not yet been identified or studied; however, due to the globin genes of the extant non-human great apes maintaining strong similarities to those of Homo sapiens (Goodman et al. 1983), I suggest that those of H. neanderthalensis will also be exceptionally similar. Additionally, genomic reconstructions and sequencing indicate that Neanderthals possessed 23 chromosome pairs—the same number as modern humans (Petr et al. 2020). This suggests that, not only may Neanderthal globin genes look largely identical to our own, they may occupy analogous locations. This theory is supported by other research: the Neanderthal melanocortin type 1 receptor gene (MC1R) was identified at the same locus as that of modern humans and was nearly indistinguishable (Lalueza-Fox et al. 2007). The investigation of human globin genes as a direct comparison to those of Neanderthals is hence justified.

Haemoglobin’s ability to offload oxygen to respiring cells is inhibited at lower temperatures (Clementi et al. 1994; Weber and Campbell 2010). The deoxygenation of haemoglobin is an endothermic, or heat-dependent, process (Weber and Campbell 2010). Succinctly, haemoglobin requires sufficient heat to be present to successfully provide oxygen to the body’s tissues. This means that haemoglobin-oxygen affinity increases and decreases in negative correlation with temperature changes, with higher temperatures favouring oxygen unloading and lower temperatures hampering unloading. Extremely cold temperatures thus restrict the body’s ability to oxygenate tissue.

Haemoglobin is a highly adaptable molecule. As the protein responsible for oxygen transport in almost all vertebrates (di Prisco et al. 2002), its functionality is imperative for respiration and tissue oxygenation (Baldwin 1976). When a species habitually endures hypoxic conditions, whether due to altitude, temperature, or diving, haemoglobin efficiency can adapt to the environment (Storz and Moriyama 2008).

Haemoglobin Adaptation in Order Primates. Haemoglobin adaptation has been observed in both humans and non-human primates. Over 60 human haemoglobin variants have amino acid replacements that increase oxygen affinity or blood concentration (Jones and Shih 2009; Gassmann et al. 2019). Many of these alterations are for survival of hypoxic conditions of high-altitude living (Huerta-Sánchez et al. 2014; Zheng et al. 2023). Further human haemoglobin variants confer disease protection, for example, the sickle-cell anaemia and thalassaemia alleles for malaria resistance (Yuthavong and Wilairat 1993). Tibetan high-altitude adaptations were also found to have originated via admixture with Denisovans, proving the potential for adaptive genetic variation in extinct hominins (Huerta-Sánchez et al. 2014).

Most of these human haemoglobin adaptions are single-base point mutations, which can occur in short evolutionary timeframes (Clark and Thein 2004). Consequently, the known human haem-adaptations appear evolutionarily recent and have not been identified in ancient samples. Any ancient globin gene modifications in Neanderthals are thus likely to be analogous, but non-identical, to those in Homo sapiens. These types of mutation are hard to track due to their sizeable genotypic diversity despite phenotypic similarity, similar to the diverse thalassaemia haemoglobinopathies (Clark and Thein 2004).

Haemoglobin adaptation has also been described in a non-human primate. Researchers identified mutations in the haemoglobin sequences of Macaca fascicularis (Faust et al. 2020). These were associated with the same genes responsible for many human haemoglobin adaptations—HBA1 and HBA2. Researchers concluded that this variation likely evolved to confer protection against malaria via a similar process to that seen in humans (Faust et al. 2020). This study reinforces the likelihood of finding adaptive haemoglobin mutations in primates other than Homo sapiens—namely Neanderthals.

Haemoglobin Adaptation to Cold. Cold-adapted haemoglobin has been identified in various taxa, from fish to mammals. The haemoglobins of several Arctic mammals, including polar bears and reindeer, have been found to possess lowered enthalpy values, decreasing the effect of temperature change on haem-oxygen affinity (De Rosa et al. 2004). Notably, the average summer temperature in the reindeer’s habitat, circa 10°C (Grayson and Delpech 2005), is on par with the theorised average of 8–10°C in the Neanderthals’ range (Zimov et al. 2012). Reconstructions of the woolly mammoth’s (Mammuthus primigenius) globin genes also identified substitutions that minimised heat loss associated with haem-oxygenation (Campbell et al. 2010).

In an alternative approach to the low-temperature oxygenation problem, polar fishes developed haemoglobins with increased oxygen-binding properties (Anderson et al. 2009). The base sequences of these Arctic fish haemoglobins show closer similarity to Antarctic fish haemoglobins than those of temperate fish, implying analogue patterns and processes in cold adaptation (Verde et al. 2008). This recurrence emphasises the plausibility of finding thermoregulatory adaptations in Neanderthal haemoglobin. In another episode of convergent cold adaptation, the specific molecular mechanisms of haemoglobin-oxygen binding in Arctic cod (Arctogadus glacialis) are shared with high-altitude avians (Anderson et al. 2009).

Whilst cold-adapted haemoglobin has not yet been identified in humans, signals of strong natural selection for other cold-relevant alleles, specifically those that limit hepatic fatty acid oxidation, have been found in Inuit populations (Hale 2020). However, research has concentrated primarily on diet-driven vulnerabilities to iron-deficiency anaemia (see, for example, Appel et al. (2018)). There is, thus, a possibility of Arctic human populations also having undiscovered cold-adapted haemoglobin.

Ancient DNA (aDNA) refers to any DNA isolated from palaeo- or archaeological specimens (Orlando et al. 2021). aDNA is often more degraded than contemporary material, making it challenging to sequence and analyse (Huynen et al. 2012). Early aDNA work only recovered short sections of DNA from large amounts of well-preserved material. However, the development of next-generation sequencing and capture methods increased the information we can extract from fragments and enabled genomes to be sequenced at unprecedented speeds (Quail et al. 2012; Shapiro and Hofreiter 2014). Next-generation techniques are ideal for aDNA, as they can exploit tiny fragments. Full palaeogenomes have now been sequenced, including the mammoth to 0.1x coverage and Neanderthal to over 50x coverage (Campbell et al. 2010; Green et al. 2010). Next-generation techniques have also extended the upper temporal boundary for useable samples to ~1.65 million years (van der Valk et al. 2021).

The next step in palaeogenetics was to identify regions of potential evolutionary importance and analyse their functional consequences (Huynen et al. 2012). This molecular-level de-extinction enables us to travel back in time and observe characteristics of extinct organisms. If reconstructed genomes are compared with those of related extant species, it is possible to identify the differences responsible for target phenotypes (Shapiro 2017). Genome editing can then alter the living species’ genome and insert relevant portions from the extinct material. Once reconstructed, it is possible to create living cells containing extinct genes. If focusing on a single gene product, researchers can subclone the gene into an expression vector and express the long-extinct molecule in living cell culture (Thornton 2004).

Research is now producing results that would have once been considered science fiction. Extinct gene products have been expressed from a range of species (see, for example, Huynen et al. (2012)). In the plant kingdom, molecular de-extinction synthesised scent molecules from extinct flowering plants (Dorado et al. 2019). In the animal kingdom, a transcriptional enhancer from the extinct thylacine was activated in vivo and found to affect chondrocyte expression (Pask et al. 2008). Two separate studies also sequenced the MC1R gene, an influencer of hair colour, in the woolly mammoth (Römpler et al. 2006) and Neanderthal (Lalueza-Fox et al. 2007).

A study by Campbell et al. (2010) on the woolly mammoth further expanded this approach to include structural functional analysis of the resurrected molecules, i.e., they investigated potential selective pressures that could have driven the evolution of differences. Campbell et al. used recovered DNA to resurrect the species’ globin genes. After expressing the genes in vitro, testing of the resurrected molecule demonstrated improved oxygen offloading when exposed to cold temperatures. The woolly mammoth was already assumed to be adapted to its high-latitude Ice Age environment, but the work of Campbell et al. put weight behind this theory.

Simply identifying differences between extinct genes and living analogues does not inform whether selection caused these differences. Discerning whether ancient differences were caused by positive directional selection is integral to linking potential adaptations to causative environmental changes. Neutral evolutionary theory must be excluded to claim that an observed mutation is adaptive (Kimura 1983): to demonstrate that a given sequence has been subjected to selective pressure, one must first reject the null that it evolved neutrally, i.e., was driven by genetic drift and stochastic processes alone (Koonin 2016).

Detecting positive selection can be achieved by searching for signatures left behind by selective sweeps (Koropoulis et al. 2020), where a beneficial mutation’s increase in frequency and subsequent fixation leads to the reduction or elimination of genetic variation in nearby sections. This condition is observed when strong selective pressures force allele frequencies to change rapidly; for example, in early Arctic populations, a selective sweep followed sudden extreme selection for alleles associated with metabolic adaptation to low carbohydrate diets (Hale 2020). During recombination, other genes can “hitch-hike” on to the focal allele and increase in frequency alongside it, reducing nearby variation. Such sweeps are primarily identified by measuring linkage disequilibrium (LD), where alleles in close proximity are non-randomly associated (Slatkin 2008). High LD indicates that an allele has recently increased in frequency under strong selection. Measurement of LD has already been used in studies on extant human populations (see, for example, Hale (2020)). Statistical methodologies used to ascertain the likelihood of given nucleotide changes occurring by chance range from binomial tests to Bayesian approaches (Koropoulis et al. 2020), with different methodologies befitting different circumstances (Slatkin 2008).

Due to the secondary nature of my data, I will not need to conduct extraction, amplification, or sequencing myself and, thus, will not need to apply for approval from my institution's relevant ethics committees. I will, however, need to verify that all human genetic information I use was previously obtained with appropriate ethical approval from equivalent institutions.

As I am working within an Aotearoa New Zealand context, my research must be performed in accordance with the 1940 Treaty of Waitangi (Boulton 2018), a formal agreement that Indigenous Māori interests would be recognised and protected in perpetuity (Hudson and Russell 2009). The Treaty remains an integral part of New Zealand’s legislative and organisational systems and honouring the Treaty in research is both a legal and ethical requirement (Health Research Council of New Zealand 2010). An important step will be ensuring that appropriate tikanga (Māori ethical protocol) for working with biological samples is consistently followed (Health Research Council of New Zealand 2010). As researchers often neglect cultural considerations (Boulton 2018), it may be impossible to obtain complete information on the nature of available data. At a minimum, I will ensure that no known breaches of ethics have been committed and that accepted international standards have been met.

I will utilise statistical tests to ascertain whether observed nucleotide differences were likely driven by positive directional selection. Neutral evolutionary theory will function as my null hypothesis. To search for the signatures left behind by selective sweeps, I will utilise a binomial test to compute summary statistics on the likelihood of my observations occurring in the absence of selection. A variety of other approaches to identifying LD and positive selection may also be beneficial, including a Bayesian framework (Koropoulis et al. 2020). If my functional difference analyses do not yield the hoped-for results, being able to report that Neanderthal haemoglobin shows signs of selection for an adaptive purpose is interesting in its own right—even if that purpose remains unclear. However, since the dataset available for Neanderthals is ancient and extremely small and most potential tests require samples of a minimum size to produce certainty, my tests are vulnerable to low statistical power. I thus plan to explore structure-function analysis as an alternate method of identifying selection.

I will conduct empirical structure-function analyses to investigate whether observed changes alter protein behaviour in a manner consistent with selection. I will obtain amino-acid sequences for the globin genes of other mammalian species via a cross-organism database, align the sequences, and note shared, deleted, or substituted amino acids. This will establish whether Neanderthal alterations are common across species and identify conserved regions. Nucleotide alterations at points that fundamentally alter a protein’s physiochemical properties rarely occur by chance and, if they do, are promptly eliminated from populations if they provide no fitness benefit (Perutz 1983). Thus, if I identify an alteration that affects the amino-acid sequence in a highly conserved region of structural importance, then the likelihood of that change being driven by positive selection is increased.