Gut Immunology

NK cells were identified in rodent liver during the mid-1970s, at approximately the same time as their discovery in the spleen. A flurry of research on splenic NK cells followed immediately, but little work was done on liver NK cells until the early years of the 21st century. In a 2012 study that reignited interest in the topic, Gordon and colleagues observed a large population of ‘immature’ NK cells in mouse liver, in addition to ‘mature’ cNK [1]. On adoptive transfer, the ‘immature’ NK cells could differentiate to cells with a ‘mature’ phenotype. Therefore, the authors proposed that the two NK cell populations in the liver had a precursor–product relationship. Daussy and colleagues later repeated the adoptive transfers under more stringent conditions, providing evidence that the two populations in fact represented separate lineages [2]. Parabiosis experiments showed that one population (cNK) depends on the transcription factor Eomes and circulates freely, while the other is Tbet dependent, is found in liver sinusoids, and is unable to leave the liver (lrNK) [3, 4]. Further interest in lrNK has been sparked by evidence that they can mediate contact hypersensitivity and, therefore, could represent ‘memory’ NK cells [3].

In the light of these findings, several groups attempted to define lrNK in humans. CD49a is considered the definitive cell surface marker for lrNK in mice; thus, the first such attempt examined the expression of CD49a by human liver NK cells [5]. A small CD49a⁺ NK cell population was found in the parenchyma, but were absent in peripheral, portal venous or hepatic venous blood, suggesting they are liver resident.
NK cells in human blood are divided into CD56^bright and CD56^dim subsets, where most are CD56^dim. Human liver is enriched in CD56^bright NK cells and these express CD69, which is now recognised as a marker not only of activation, but also of tissue residence. Therefore, the next attempt to define lrNK in humans focussed on CD56^bright NK cells, finding that they express a distinctive panel of chemokine receptors, integrins, and L-selectin, which is likely to mediate their retention in the liver [6]. Similar to mouse lrNK, they are located primarily in the sinusoids and are likely to be retained there by the interaction of CCR5 and CXCR6, expressed by the NK cells, and their ligands on sinusoidal endothelial cells [6]. Indeed, another recent study suggested that human lrNK should be defined not as CD56^bright but as CXCR6⁺, similar to mouse lrNK [7].
Neither CD56 nor CXCR6 expression defines liver residence, since significant CD56^bright and CXCR6⁺ populations are present in blood [6, 7]. A consensus is now emerging that the best way to distinguish between cNK and lrNK in humans is by their expression of Tbet and Eomes [7, 8, 9]. Eomes^hiTbet^lo NK cells (henceforth ‘Eomes^hi’) account for some 50% of human liver NK cells, but are completely absent in blood, and largely, but not completely, overlap with the CD56^bright and CXCR6⁺ populations. However, the Eomes^hi population does not overlap with the CD49a⁺ population [5, 7]. It appears, then, that there are two nonoverlapping NK cell populations that are potentially liver resident in humans: CD49a⁺ NK cells and Eomes^hi (largely CD56^bright and CXCR6⁺) NK cells (Box 1).

The absence of Eomes^hi NK cells in blood, together with their expression of proteins associated with tissue retention, points to these cells representing lrNK, but is not definitive. Recently, however, it has been possible to carry out experiments in humans that were similar to the parabiosis experiments that defined lrNK in mice. In clinical liver transplantation, donors and recipients are not routinely HLA matched, allowing recipient-derived cells to be distinguished from those originating from the donor liver by their expression of HLA variants. Taking this approach, it was shown that Eomes^hi NK cells cannot exit the liver, where they can persist for up to 13 years, whereas Eomes^lo cNK cells recirculate freely [9]. This allows us to say with some certainty that Eomes^hi NK cells are indeed liver resident in humans.
While performing these experiments, the unexpected observation was made that recipient-derived Eomes^hi lrNK emerge within weeks of transplantation, indicating replenishment from the circulation. It was further found that, on culture with cytokines that are highly expressed in the liver, sorted Eomes^lo cNK upregulated Eomes together with cell surface markers associated with an lrNK phenotype [9]. This was surprising because, in mice, cNK and lrNK are thought to form separate lineages [2]. It is, of course, possible that human and mouse lrNK differ in this respect, but it is worth noting a small degree of flexibility between the two lineages in mice, even in the most stringent experiments [2, 3]. It is also possible that the immunosuppression to which transplant patients are subjected alters the ability of circulating cells to be recruited to the liver and fill a resident niche. There is some evidence that this is the case for Kupffer cells [10] and it has been suggested that the conflicting outcomes of adoptive transfers reported by Gordon and Daussy are a result of different conditioning regimes in the recipient mice.

In both mice and humans, lrNK cells are present and something of their origin and lineage is now understood. However, the function of these cells remains obscure. Since studies in humans have been informed by those in mice, experiments in mice are likely to be a necessary first step in defining the functions of lrNK in humans. Mouse lrNK may be ‘memory’ cells [3] and CD49a⁺ lrNK in humans are NKG2C⁺, reminiscent of NKG2C⁺ NK cells in the blood, which may have memory of human cytomegalovirus (HCMV) [5]. Meanwhile, the longevity of the Eomes^hi lrNK population could also be suggestive of memory [9]. Experiments to test the idea that either population represents memory cells will be challenging to perform in humans, although bulk hepatic NK cells in macaques display antigen-specific memory responses to vaccination. This supports the notion of memory NK cells in primate liver, although the subset responsible has not yet been defined [11].
Anatomical location and protein expression may also provide clues to function. CD49a⁺ lrNK are found in the parenchyma and express cytotoxic effector molecules and receptors for MHC class I; thus, it seems likely that they recognise and kill virally infected or cancerous hepatocytes. By contrast, Eomes^hi lrNK are found in the sinusoids and express fewer receptors for human targets, suggesting that they recognise nonhuman cells in the blood. The liver processes blood coming from the gut; thus, the possibility that they respond to bacteria or bacterial products is particularly attractive. In support of this hypothesis, mouse lrNK express high levels of AHR [3, 12], while Eomes^hi lrNK in humans express IL23R and RORgt [9] and these genes are required for the development of ILC3, which are involved in the recognition of gut bacteria. Whatever the function of these cells, their frequency, longevity, and conservation between mice and humans suggests that it is likely to be important and, therefore, defining it will represent a significant advance in our understanding of the hepatic immune system.

---

On-going Mechanical Damage from Mastication Drives Homeostatic Th17 Cell Responses at the Oral Barrier