Immunity and Inflammation in the Uterus


Author’s address (for correspondence): Prof Martin Sheldon, Institute of Life Science, College of Medicine, Swansea University, Swansea SA2 8PP, UK. E-mail: i.m.sheldon@


Microbes often infect the uterus and particularly the endometrium of animals. Infections are most commonly associated with natural service, pregnancy and the post-partum period, leading to inflammation with the elaboration of cytokines, chemokines and prostaglandins. Clinical diseases such as metritis, endometritis and abortion are important causes of infertility. The adaptive immune response to infection has been characterized previously, so the present review aims to highlight the emerging role for innate immunity in the endometrium. The detection of microbes and the innate immune response depends on the detection of pathogen-associated molecular patterns by pattern recognition receptors. The main families of pattern recognition receptors are Toll-like receptors (TLRs), nucleotide oligomerization domain-like receptors, retinoic acid-inducible gene I-like receptors and C-type lectin receptors. These receptors are most often expressed by hematopoietic cells, but the epithelial and stromal cells of the endometrium also possess functional receptors. For example, endometrial cells express TLR4 for recognition of the lipopolysaccharide endotoxin of Gram-negative bacteria, leading to secretion of IL-6, IL-8 and prostaglandin E2. It is likely that the epithelial and stromal cells provide a first line of defence in the endometrium to alert hematopoietic cells to the presence of microbes within the uterus.


The uterus and in particular the endometrium lining the uterus have important roles in normal reproductive cycles, implantation and placentation, and supporting a healthy foetus until parturition. However, microbial infections of the uterus are common in man and animals, and uterine infections are important because they cause infertility, abortion, pre-term labour and clinical disease (Wira et al. 2005; Jabbour et al. 2009; Sheldon et al. 2009; Mor and Cardenas 2010). Many infections ascend the genital tract via the cervix whilst some specifically adapted pathogens reach the uterus through the circulation. Regardless of the route of infection, the principal site of pathology is the endometrium or the endometrial-foetal interface during pregnancy. The importance of the uterus and especially the endometrium means that microbial invasion does not go unnoticed, with innate and adaptive immunity countering the infection (Wira et al. 2005; Sheldon et al. 2009).

Immunity and inflammation in the uterus has been well characterized and discussed in numerous reviews, which should be consulted for further details (Wira et al. 2005; Jabbour et al. 2009; Sheldon et al. 2009; Hansen 2010; Mor and Cardenas 2010; Sheldon and Bromfield 2011). Indeed, immunity and inflammation play roles in normal ovarian and uterine cycles, implantation, placentation and foetal development, as well as in the response to infection by microbes. The present review will not revisit these diverse matters but rather will focus on one particular area of new knowledge – the fundamental mechanisms of recognition and response to microbes in the uterus. Innate immunity comprises many aspects of non-specific defence against microbes, including antimicrobial peptides, the epithelial barrier and the complement cascade. However, impetus for the expansion of research in innate immunity came from seminal discoveries made in the laboratories of Jules Hoffmann and Bruce Beutler of specific cellular ‘pattern recognition receptors’ that detect molecules commonly associated with microbes (Lemaitre et al. 1996; Poltorak et al. 1998). The first discovery was that the Toll protein involved in dorsoventral polarity during embryonic development was also required for an effective immune response to fungal infection in Drosophila melanogaster (Lemaitre et al. 1996). The second discovery was that a Toll-like protein, Toll-like receptor 4 (TLR4), was found to be functionally important in mice and necessary for the inflammatory response to lipopolysaccharide (LPS), which is a component of the cell wall of Gram-negative bacteria but also acts as an endotoxin in animals (Poltorak et al. 1998). These discoveries in flies and mice led to the award of the 2011 Nobel Prize in Medicine or Physiology. Work using knockout mice or siRNA to target TLR4 has demonstrated that the pattern recognition receptors of the innate immune system are also important for detection of microbial infections in the uterus of animals (Sheldon and Roberts 2010; Cronin et al. 2012).

Microbial Infection of the Uterus

Sexually transmitted infections are common in domestic animals and humans. Sexually transmitted infections in animals are important where natural mating is used. For example, bovine venereal campylobacteriosis is widespread in developing countries where infections cause infertility, embryo mortality and abortion (Mshelia et al. 2010). However, the introduction of artificial insemination in dairy cattle has almost eliminated the problem in much of the developed world. Microbial causes of abortion in food-producing animals remain a persistent problem, and as well as loss of the foetus they often lead to culling of the dam. Brucella abortus has a place in history not only as an infectious cause of abortion in animals but also as an important zoonosis causing chronic disease in man, until the spread of infection was controlled by pasteurization of milk for human consumption. In humans, sexually transmitted infections remain a major issue across the whole world and it is estimated that there are 340 million new infections each year (WHO 2001). Whilst many sexually transmitted infections are unapparent, they lead to substantive problems because they cause infertility and pelvic inflammatory disease, which costs $10 billion/annum for human health care in the USA (Washington and Katz 1991). Infections with bacteria also cause approximately a quarter of pre-term labour cases in women with devastating effects for parents and the child, although pre-term labour does not appear to be common in animals.

The remaining major disease of the uterus is important in both humans and animals and has had an influential role in shaping modern clinical practice. Bacterial infections around the time of parturition commonly cause disease of the uterus, called puerperal fever in women and metritis in animals. Puerperal fever killed many women up until the end of the 19th century when Semmelweis prevented disease by introducing hand antisepsis for the physicians attending women in labour. Peri-parturient and post-partum bacterial infections of the female genital tract are particularly common in dairy cattle (Sheldon et al. 2009). Parturition is a significant cause of trauma to the endometrium leading to the loss of surface epithelium, which exposes the underlying stromal cells to bacteria and additional trauma, such as difficult parturition or retained placenta, thereby increasing the risk of uterine disease (Potter et al. 2010). Between 20% and 40% of animals develop metritis within 10 days of parturition with inflammation of the endometrium characterized by an influx of neutrophils (Fig. 1a,b). Clinical disease persists beyond 3 weeks post-partum as clinical endometritis in ∼20% of cows and subclinical endometritis in ∼15% of animals. Endometritis is characterized by histological signs of chronic inflammation in the endometrium including the development of aggregates of lymphocytes (Fig. 1c,d) (Sheldon et al. 2006). Post-partum uterine diseases of animals are also of considerable economic importance. For example, metritis costs the European dairy industry €1.4 billion annually for treatments, reduced milk production and the cost of replacing infertile animals (Sheldon et al. 2009).

Figure 1.

 Inflammation of the endometrium. Photomicrographs of haematoxylin and eosin-stained sections of bovine endometrium collected during the acute phase of infection (a) with neutrophils visible in the highlighted area of the section inbetween the cells of the epithelium and stroma (b; arrow indicates neutrophil); and during the chronic phase of infection (c) with aggregates of immune cells including lymphocytes within the highlighted area of the section of endometrium (d)

Innate Immunity

Innate immunity is the non-specific defence against microorganisms including the physiological barriers of the skin and mucosa, antimicrobial peptides and the complement system, and cells that detect and respond to microbes. Hematopoietic cells involved in the innate immune responses include monocytes, macrophages, eosinophils, neutrophils and natural killer cells (Fig. 1a,b). In addition, the non-hematopoietic epithelial, stromal and endothelial cells of mucosa are also often involved in host defence against potentially harmful substances and microbial pathogens. Epithelial cells were once thought to simply form a passive barrier to infection but are now recognized to orchestrate defensive strategies by mediating innate immune responses and helping to direct the adaptive immune response through interactions with dendritic cells, B cells and T cells (Fig. 1c,d). The cells of the innate immune system use pattern recognition receptors to detect the presence of pathogens by sensing microbial molecules known as pathogen-associated molecular patterns.

Pathogen-Associated Molecular Patterns

Most pathogen-associated molecular patterns (PAMPs) are evolutionary conserved molecules like cell wall components and nucleic acids that are required for the function of microbes. Bacterial cell wall components are the most clearly characterized PAMPs and probably the most important in the endometrium. Bacteria are classified into two large distinct groups, Gram-positive or Gram-negative, based on the chemical and physical properties of their cells walls; both groups are important pathogens in the uterus (Sheldon et al. 2002; Sheldon et al. 2009). The most obvious distinguishing feature of Gram-negative bacteria is the presence of LPS in the outer membrane of the bacterial cell wall with important roles in the integrity and physiological function of the wall (Holst et al. 1996). The general structure of LPS is common across Gram-negative bacteria with three principal domains: a hydrophobic lipid A moiety, an oligosaccharide core and a distal polysaccharide known as the O antigen (Fig. 2). As well as the role of LPS in bacteria, LPS is an endotoxin causing disease in mammals and is the most widely studied PAMP. Another common PAMP is peptidoglycan, which is important for the rigidity of the cell wall of Gram-positive and Gram-negative bacteria (Fig. 2). Peptidoglycan is composed of repeated units of a disaccharide N-acetyl glucosamine (GlcNAc)-N-acetyl muramic acid (MurNAc) cross-linked by peptide side chains. Although peptidoglycan is common to all bacteria, variations exist either in the glycan strand, peptide stem or in the amino acids that form the interpeptide bridge. Another bacterial PAMP is the flagellin protein component of the flagella that bacteria use for motility, adhesion, invasion and the secretion of virulence factors. The nucleic acids of many microbes are also PAMPs, including the RNA and DNA of bacteria and viruses.

Figure 2.

 Structure of bacterial cell walls. The cell walls of Gram-positive and Gram-negative bacteria contain important pathogen-associated molecular patterns that are recognized by pattern recognition receptors on host mammalian cells. All bacteria have a phospholipid bilayer inner membrane surrounding the cytoplasm. Gram-positive bacteria have an external thick layer of peptidoglycan (PGN) embedded with teichoic acid and polysaccharides, which gives the bacterial wall strength. Gram-negative bacteria have a relatively thin layer of PGN but posses an outer membrane that contains lipopolysaccharides (LPS), phospholipids and lipoproteins

The PAMPs are recognized by pattern recognition receptors, which are found on the surface or within mammalian cells. The main families of pattern recognition receptors are Toll-like receptors (TLRs), nucleotide oligomerization domain (NOD)-like receptors (NLRs), retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs) and C-type lectin receptors (CLRs) (van de Veerdonk et al. 2008; Beutler 2009; Takeuchi and Akira 2010).

Toll-like Receptors

There are several TLRs in animals that are involved in the recognition of PAMPs (Takeuchi and Akira 2010). Table 1 outlines the major TLRs and examples of their principal ligands, and whether the receptors are localized to the plasma membrane or intracellular endosomes. The best characterized member of the TLR family is TLR4, which is widely expressed by hematopoietic and non-hematopoietic cells, and recognizes LPS in complex with two co-receptors, CD14 and MD-2 (Park et al. 2009; Bryant et al. 2010). Heterodimers of TLR2/TLR1 or TLR2/TLR6 recognize a variety of PAMPs from both Gram-positive and Gram-negative bacteria including lipopeptides and peptidoglycan, glycolipids and lipoteichoic acid (Takeuchi et al. 1999). The receptor dimerization is important, and tri-acetylated lipopeptides are usually bound by TLR2/TLR1 whereas di-acetylated lipopeptides are bound by TLR2/TLR6 (Takeuchi et al. 1999; Jin et al. 2007). Bacterial flagellin is recognized by TLR5 (Hayashi et al. 2001). Double-stranded RNA (dsRNA) from viruses is bound by TLR3, although a synthetic analogue (polyinosine-deoxycytidylic acid; poly I:C) is widely used in vitro to examine TLR3 activity. Uridine or guanosine-rich single-stranded RNAs from a variety of viruses and synthetic imidazoquinoline-like molecules such as resiquimod are recognized by TLR7 and TLR8. Finally, TLR9 recognizes the unmethylated CpG motifs of single-stranded DNA present in the genomes of many viruses and bacteria.

Table 1.   Toll-like receptors, their cellular location and principal ligands
TLRTLR locationTypical ligands
  1. TLR, Toll-like receptor.

TLR1Plasma membraneTriacylated lipopeptides
TLR2Plasma membranePeptidoglycan, lipoteichoic acid
TLR3EndosomeSingle-stranded RNA and double-stranded DNA
TLR4Plasma membraneLipopolysaccharide, LTA, fibronectin, mannan (Candida)
TLR5Plasma membraneFlagellin
TLR6Plasma membraneDiacylated lipopeptides, fungal zymosan
TLR7EndosomeSingle-stranded RNA, imidazoquinolines
TLR8EndosomeSingle-stranded RNA
TLR9EndosomeBacterial and viral unmethylated CpG motifs

The TLRs are type I transmembrane glycoproteins with the extracellular N-terminal end composed of leucine-rich repeats, which mediate binding to PAMPs and receptor dimerization (Akira and Takeda 2004; Takeuchi and Akira 2010). Binding of a PAMP activates the conserved cytoplasmic region of each TLR, which is often called the Toll/IL-1 receptor (TIR) domain because of similarity with the cytoplasmic domains of the interleukin-1 receptor family (Akira and Takeda 2004). Activation of the TIR domain initiates intracellular signalling cascades via adaptor proteins including MyD88, TRIF, TRAM and TIRAP (Watters et al. 2007). However, the MyD88-dependent pathway is predominant for most TLRs, and MyD88 recruits IL-1 receptor-associated kinase (IRAK)-4 to activate IRAK-1 and then TNF receptor-associated factor 6 (TRAF6). Then, TRAF6 activates the IκB kinase (IKK) complex resulting in translocation of the nuclear transcription factor (NFκB) from the cytoplasm to the nucleus. Alternatively, TRAF6 activates mitogen-activated protein kinases (MAPK) such as p38 and ERK1/2, which result in AP-1 transcription factor activation. These NFκB- and MAPK-dependent pathways lead to the production of cytokines, chemokines, prostaglandins and other inflammatory mediators (Fig. 3). In addition, epithelial cells produce a wide range of antimicrobial peptides in the female reproductive tract of animals and humans, constitutively and in response to PAMPs (Davies et al. 2008; Wira et al. 2011).

Figure 3.

 Toll-like receptor 4 pathway. The lipopolysaccharide (LPS) of Escherichia coli bind TLR4 to activate intracellular signalling via the mitogen-activated protein kinases (MAPK) (p38 and ERK1/2) and NFκB pathways. Activation of MAPK and NFκB pathways leads to gene transcription and production of inflammatory mediators including prostaglandins, cytokines and chemokines. These inflammatory mediators in turn attract and activate hematopoietic cells such as macrophages and neutrophils, which clear the microbial infection

Nod-like Receptors

The NLRs are a family of at least 23 cytosolic receptors that when activated by PAMPs or damage-associated molecules, lead to cytokine production (Davis et al. 2011). The NLRs have also been implicated in autophagy, a lysosomal degradation and cell death pathway that follows infection. The NLRs are characterized by the presence of a nucleotide binding domain flanked by leucine-rich repeats at the C-terminus, and a protein binding domain – caspase activation and recruitment domain (CARD), baculovirus inhibitor of apoptosis protein repeat (BIR), death effector domain (DED) or pyrin domain (PYD) – at the N-terminus. These N-terminal domains, also termed ‘the effector region’, are responsible for the protein–protein interactions needed to activate downstream signal transduction. The NLRs can be categorized into subfamilies based on this effector domain including: NLRCs containing a CARD domain, NLRPs containing a pyrin effector domain, and NAIPs or NLRBs containing BIR domains (Davis et al. 2011). The NLRs first identified were NOD1 and NOD2, which recognize components of peptidoglycan: NOD1 recognizes D-γ-glutamyl-meso-DAP dipeptide found in all Gram-negative but only some Gram-positive PGNs, whereas NOD2 recognizes the conserved muramyl dipeptide motif found in all PGNs (Strober et al. 2006; Shaw et al. 2008). NOD2 can also recognize viral ssRNA and mycobacterial N-glycolylmuramyl dipeptides. Activation of NOD1 and NOD2 by their ligands results in activation of NFκB, and the production of pro-inflammatory cytokines and antimicrobial peptides. Activation of NOD2 by MDP also can result in the activation of the MAPK pathways via the adapter CARD9. However, certain members of the NLR family that detect microbial components in the cytosol also trigger the assembly of a large caspase-1 activating complex termed the inflammasome. The inflammasome complex activates caspase-1 to process pro-forms of IL-1β, IL-18 and IL-33 to their mature biologically active forms ready for secretion from cells.

RIG-I-like Receptors

The cytoplasmic RNA helicases that comprise the family of RIG-I-like receptors (RLRs) play a role in anti-viral defence resulting in the production of type 1 interferons (IFN) (Loo and Gale 2011). The RLRs include RIG-I and melanoma differentiation-associated factor 5 (MDA5). The RIG-I receptor is involved in the recognition of Paramyxoviridae, Filoviridae and Rhabdovirdae among others, whereas MDA5 is important in the recognition of Picornaviridae (Loo and Gale 2011). The role of these RIG-I receptors in uterine disease of domestic animals is mostly unexplored but warrants attention because viruses such as bovine viral diarrhoea (BVD) virus and bovine herpesvirus 4 (BoHV-4) cause infertility and abortion. Indeed, bovine herpesvirus 4 is tropic for the bovine endometrial stromal cells where it drives the activation of the gene promoter for the chemokine IL-8 (Donofrio et al. 2010).

C-type Lectin Receptors

The CLRs recognize and bind carbohydrates via a carbohydrate recognition domain (CRD). The mannose binding lectin (MBL) in the peripheral circulation recognizes mannose and fucose, leading to activation of the lectin complement pathway and increased neutrophil chemotaxis. Cellular CLRs include Dectin-1 (also known as C-type lectin domain family 7 member A; CLEC7A), Dectin-2, macrophage-inducible C-type lectin (MINCLE), DC-SIGN and the mannose receptor (van de Veerdonk et al. 2008). Dectin-1 is the most widely studied CLR and is expressed on myeloid cells, including neutrophils, monocytes/macrophages and dendritic cells, with limited expression on other cell types. Dectin-1 recognizes specific glucose polymers found in the cell walls of fungi including Candida albicans and Saccharomyces cerevisiae.

Pattern Recognition Receptors in the Endometrium

Intact endometrium expresses mRNA for most pattern recognition receptors, at least in the species examined so far, which includes humans, mice, rats and cattle. Of course, this is not surprising as the endometrium comprises immune cells as well as epithelial, stromal and endothelial cells. However, what was less anticipated was that purified populations of epithelial or stromal cells also express most TLRs and functional responses are evident at least for cells treated with bacterial PAMPs in vitro.

Bovine and murine endometrial epithelial and stromal cells have immunoreactive TLR4, and LPS or Escherichia coli stimulate secretion of a range of cytokines and chemokines in vitro (Sheldon and Roberts 2010; Sheldon et al. 2010; Cronin et al. 2012). The responses to LPS measured in bovine cells included increased phosporylation of MAPK p38 and ERK1/2, as well as translocation of NFκB (nuclear factor kappa-light-chain-enhancer of activated B cells) into the nucleus (Cronin et al. 2012). In addition, there was increased expression of mRNA encoding IL-1β, IL-6 and IL-8 in stromal and epithelial cells from the bovine endometrium and increased expression of antimicrobial peptides such as LAP and TAP in epithelial cells (Davies et al. 2008). These findings are likely to represent biologically relevant defence mechanisms because endometrial biopsies or cytobrush samples from cows with uterine bacterial infection have a similar profile of genes with increased expression (Herath et al. 2009; Fischer et al. 2010). Furthermore, supernatants of cultured stromal and epithelial cells from the bovine endometrium also accumulate IL-6 and IL-8 when treated with LPS or Gram-negative bacteria. In the mouse, epithelial and stromal cells from wild type but not TLR4-null mice accumulate IL-6, CXCL1 and CCL20 when treated with LPS (Sheldon and Roberts 2010). One notable absence in endometrial responses to LPS is the secretion of TNFα, which does not appear to substantially accumulate when murine or bovine endometrial cells are treated with LPS in vitro (Sheldon and Roberts 2010). Again this observation somewhat reflects observations in cows with post-partum uterine infection where there is little evidence of a marked TNF response, at least in peripheral plasma (Williams et al. 2007). The specific role of TLR4 in the endometrium was confirmed using TLR4-null mice (Sheldon and Roberts 2010); siRNA targeting TLR4 or the adaptor MyD88 was also effective in reducing the bovine endometrial cell responses to LPS (Cronin et al. 2012). Whilst the interaction of LPS with TLR4 is the best studied paradigm for inflammation in the endometrium, cellular responses to PAMPs that bind TLR1, TLR2, TLR5 and TLR6 have also been reported in a range of species, including the intact bovine endometrium (Young et al. 2004; Borges et al. 2012).

The abundant accumulation of IL-6 when bovine intact endometrium explants or endometrial cells are treated with PAMPs is likely important not only locally but also systemically (Borges et al. 2012; Cronin et al. 2012). Increased concentrations of IL-6 in peripheral plasma stimulate the production of acute phase proteins such as haptoglobin, αl-acid glycoprotein, ceruloplasmin and serum amyloid A from the liver, which aid tissue repair after damage or infection (Baumann and Gauldie 1994). Dairy cattle have increased plasma concentration of αl-acid glycoprotein, haptoglobin and ceruloplasmin after parturition, and bacterial infection stimulates further increases (Sheldon et al. 2001). There is some evidence of local cellular expression of genes for acute phase proteins in peripheral tissues but the liver is the principal site of their generation.

An aspect of innate immunity that may have particular importance in cattle is the role of prostaglandins. Prostaglandin E2 is an important inflammatory mediator in many mammals and endometrial stromal cells produce abundant prostaglandin E2 in response to LPS or Gram-negative bacteria, except when TLR4 expression is disrupted (Herath et al. 2009; Sheldon and Roberts 2010). In cattle the epithelial cells of the endometrium also produce prostaglandin E2 when treated with LPS or E. coli in vitro, and the same inflammatory mediator is evident in the blood of cows with post-partum uterine infections in vivo (Herath et al. 2009). The increased prostaglandin E2 associated with LPS was specifically because of increased expression of phospholipase A2 group VI, which generates the arachidonic acid for conversion to prostaglandin E2. One of the consequences of the increased prostaglandin E2 concentrations may be to delay luteolysis, which depends on prostaglandin F produced by the endometrial epithelial cells. Indeed, this is one of several mechanisms by which infection and inflammation can impact ovarian function [reviewed in (Sheldon et al. 2009; Richards et al. 2008)].

Our overall concept is that the epithelial and stromal cells provide an opportunity for the initial detection of PAMPs and defence against microbes, attraction of specialized immune cells such as monocytes and neutrophils to the site of infection, and also assists with the development of the adaptive immune responses. An important question is how the cellular and inflammatory responses are regulated.

Regulation of Innate Immunity and Inflammation in the Uterus

Steroids are potent regulators of innate immunity and particularly the TLR pathways (Ogawa et al. 2005). So, the most obvious regulators to consider for innate immunity and inflammation in the endometrium are the ovarian steroids, which exquisitely control the physiology of the endometrium, antimicrobial peptides and a multitude of regulatory immune cells that can be found in the endometrium (Wira et al. 2011). Progesterone and estradiol certainly seem to have an impact on the expression and function of the TLRs in the endometrium in vivo and in vitro (Herath et al. 2006; Aflatoonian et al. 2007). However, the mechanisms are not clearly established and ovarian steroids regulate numerous components of the inflammatory pathways including the roles of NFκB and MAPK (King et al. 2010).

Diet in dairy cattle or at least the transition to lactation and often a negative energy balance can influence post-partum immunity. Cows in negative energy balance have perturbations of gene expression in the endometrium that may prevent cows from mounting an effective immune response to bacteria (Wathes et al. 2009). Some of these effects are likely mediated by effects of negative energy balance on the function of neutrophils, which are important cells of the innate immune system and defence of the endometrium. Negative energy balance can impact multiple parameters that represent neutrophil function, including glycogen content, cytochrome c reduction and myeloperoxidase activity (Hammon et al. 2006; Galvao et al. 2010).


There can be no doubt that specifically adapted microbes and non-specific infections of the uterus cause infertility, foetal mortality, economic losses and unnecessary suffering. This article focused on the emerging role of innate immunity in the uterus. Pattern recognition receptors are expressed by endometrial as well as immune cells. Binding of these receptors to PAMPs of microbes that infect the uterus leads to inflammation mediated by MAPK and NFκB intracellular signalling pathways in particular. The host cell responses include the release of cytokines, chemokines and prostaglandins. Some of these effects are regulated by the ovarian steroids and probably by other factors found in the pregnant animal. The remaining challenges relate to how innate immunity drives adaptive immunity in the female genital tract, developing effective treatments that target the uterus, and understanding how inflammation in the uterus is regulated. Addressing these questions will likely depend on understanding the underlying fundamental biology of host–pathogen interactions in the female genital tract.


This work was supported by the Biotechnology and Biological Sciences Research Council ( Matthew Turner holds a BBSRC Doctoral Training Grant (Grant number: BB/F017596/1), and Gareth Healey and Martin Sheldon are supported by a BBSRC project grant (BB/I017240/1). The funding body had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Conflicts of interest

None of the authors have any conflicts of interest to declare.

Author contributions

All the authors contributed to the writing of the paper.