Altered natural killer cell subset homeostasis and defective chemotactic responses in paroxysmal nocturnal hemoglobinuria

The glycosylphosphatidylinositol (GPI) anchor secures numerous proteins to the cell surface.¹  In paroxysmal nocturnal hemoglobinuria (PNH), there is expansion of hematopoietic stem cells (HSCs) harboring a somatic mutation in the X-linked PIGA gene, disrupting a key step in GPI anchor biosynthesis.^2,3  These mutant HSCs generate substantial populations of mature blood cells lacking GPI-linked proteins; red blood cells and platelets lacking the GPI-linked complement protectors CD55 and CD59 are highly susceptible to complement-mediated lysis, resulting in many of the clinical features of PNH.^2,3  Accordingly, inhibiting the formation of the complement membrane attack complex using a humanized anti-C5 antibody (eculizumab) is an effective therapy.^2,3 

Human natural killer (NK) cells are classified into 2 main subsets.⁴  The weakly cytotoxic CD56^bright cells constitute 10% of peripheral blood NK cells and are the likely precursors of the highly cytotoxic CD56^dim cells, which constitute the remaining 90% of peripheral NK cells in the blood. However, CD56^bright NK cells express lymph node homing molecules, and the ratio of CD56^dim:CD56^bright cells is reversed in secondary lymphoid tissue.^5,6  We analyzed NK cells from 47 PNH patients and identified a role for GPI-linked molecules in NK cell chemotactic responses and the homeostasis of NK cell subsets.

Peripheral blood samples were collected, after informed consent in accordance with the Declaration of Helsinki, from 47 PNH patients (32 receiving eculizumab; supplemental Figure 1, available on the Blood Web site) attending the PNH National Service clinic in Leeds, United Kingdom. The study was approved by the Leeds Teaching Hospitals National Health Service Trust ethical review board. Cells were stained with antibodies defining NK cells (CD56+CD3^neg) and GPI-linked molecules (CD48, CD59, CD160; see supplemental Methods). Discrimination of GPI+ and GPI^neg NK cells was also performed using fluorescent aerolysin, a modified bacterial toxin that binds to intact GPI anchors.⁷  For migration assays, NK cells (purified by indirect selection) were applied to 5-μM membrane transwells (Corning) and allowed to migrate in response to 100 nM stromal cell–derived factor 1 (SDF-1; R&D Systems). The GPI+:GPI^neg NK cell ratio was analyzed in the input and migrated populations by flow cytometry. Further details are provided in supplemental Methods.

The GPI^neg NK cells in PNH patients lacked cell surface expression of the GPI-linked molecules CD48, CD59, and CD160 (Figure 1A). GPI+ and GPI^neg populations of CD56^dim and CD56^bright NK cell subsets were detectable in the majority of patients (Figure 1B; supplemental Figure 1). However, in 10 PNH patients, the GPI+CD56^bright NK cells made up less than 0.02% of the total GPI+ cells (Figure 1B; supplemental Figure 1). This was not associated with eculizumab treatment, and 4 patients analyzed on multiple occasions revealed this to be a stable phenotype (Figure 1B; supplemental Figures 1-2). CD56^bright NK cells differentiate into CD56^dim NK cells,^9-11  and we reasoned that GPI+CD56^bright NK cells must be present in these individuals to generate the GPI+CD56^dim population. Furthermore, GPI+CD56^bright NK cells have the normal, not the GPI-deficient, phenotype. We speculated that the GPI+CD56^bright NK cells were present in all PNH patients but that GPI+CD56^bright and GPI^negCD56^bright NK cells were differentially distributed between the peripheral blood and other tissues. Indeed, CD56^bright NK cells have lymph node homing properties, and inflammation results in their loss from the periphery.^12,13  We suggest that the GPI^negCD56^bright NK cells are outcompeted by the normal GPI+CD56^bright NK cells for entrance to (or occupancy of) a niche such as the secondary lymphoid tissue. Such competition would reduce the occupancy of this niche by GPI^negCD56^bright NK cells, resulting in their enrichment in peripheral blood. Accordingly, analysis of the entire PNH cohort revealed that GPI^negCD56^bright NK cells were significantly more abundant in peripheral blood compared with their GPI+ counterparts and with the CD56^bright NK cells from healthy donors (Figure 1C).

The importance of chemokines in leukocyte trafficking suggests that impaired chemotactic responses might underlie altered NK cell subset distribution in PNH. The expression of chemokine receptors by NK cells is complex,^14,15  and it is difficult to perform chemotaxis assays on the small numbers of CD56^bright cells available from these patients. We therefore analyzed the responses of GPI+ and GPI^neg NK cell populations to SDF-1/CXCL12, the CXCR4 ligand. CXCR4 is expressed by the majority of peripheral blood NK cells, aiding these studies.^14,15  Treatment of NK cells with phosphatidylinositol-specific phospholipase C (PI-PLC) removed GPI anchor–linked proteins but left CXCR4 expression intact. These PI-PLC treated NK cells exhibited significantly reduced migration in response to SDF-1, whereas sialidase treatment had no effect (Figure 2A-B). This suggested that GPI-deficient NK cells have impaired chemotactic responses. We repeated the SDF-1 migration assays using NK cells from 6 PNH patients in which all 4 GPI+/GPI^neg and CD56^bright/CD56^dim subsets were present in peripheral blood. We analyzed the ratio of GPI+:GPI^neg NK cells in the starting population and in the migrating cells and found that the SDF-1–induced migration significantly enriched the GPI+ NK cells (Figure 2C-D).

These results reveal that optimal chemotactic responses require GPI-linked molecules: GPI-linked proteins are abundant in lipid rafts, PIGA mutations disrupt raft formation, and CXCR4 requires incorporation into these structures for optimal activity.^16-18  In GPI^neg NK cells, SDF-1 responses may be impaired because of the redistribution of CXCR4 to membrane microdomains less able to support receptor signaling or into a local membrane environment that alters CXCR4 conformation.¹⁷  Impaired SDF-1 responses were not caused by defective CXCR4 expression, as GPI^neg NK cells expressed significantly higher levels of cell surface CXCR4 compared with their GPI+ counterparts (Figure 2E; supplemental Figure 3). Increased chemokine receptor expression has previously been reported in GPI-deficient cells,¹⁹  and we also found elevated expression of CXCR3 and CCR7 on GPI^neg NK cells (Figure 2E; supplemental Figure 3). Indeed, CCR7 is preferentially expressed by CD56^bright NK cells, and this receptor exhibited greater expression on the GPI^negCD56^bright NK cells (supplemental Figure 3). Elevated receptor expression may reflect in vivo selection, whereby GPI^neg NK cells respond to chemokines only if they express high levels of the receptor, favoring their survival. Alternatively, defective chemokine receptor activity may prevent activation-induced internalization in GPI^neg NK cells, increasing cell surface expression.²⁰ 

Why the GPI+CD56^bright NK cells were undetectable in the blood of just a fifth of PNH patients is unclear. However, NK subset homeostasis is regulated by both environmental (eg, infection/inflammation^9-13 ) and genetic factors,²¹  and separating their respective contributions within the confines of the rare PNH phenotype represents a considerable challenge. Defective SDF-1/CXCR4 activity cannot itself account for the altered subset homeostasis we observed in the patients, but it does highlight defective chemotaxis in PNH. Several models have been proposed to account for the abundance of GPI^neg cells in PNH, focusing on the maintenance and expansion of GPI^neg HSC.^3,22,23  Our results indicate that the balance of GPI+ and GPI^neg cells within a particular hematopoietic lineage is governed not only by the abundance of GPI^neg HSC but also by in vivo processes operating on mature cells. GPI-linked molecules are known to play a role in cellular trafficking and homeostasis of other hematopoietic cell types.^24,25  Furthermore, GPI^negCD34+ cells exhibit enhanced SDF-1 chemotaxis but reduced adhesion to SDF-1.²³  Our results appear contradictory to these findings, but such differences may be explained by the presence of cell-type–specific factors that regulate chemokine receptor activity.²⁰  Despite their differences, our data, and those of Ratajczak et al,²³  reveal that differential chemotactic responses of GPI+ and GPI^neg cells result in unequal competition for a particular niche, contributing to the balance and persistence of GPI+ and GPI^neg cells in PNH.

We are grateful to other members of our laboratory for technical advice and assistance and to Alan Melcher and Pam Jones for encouragement.

Contribution: R.J.K., A.H., and P.H. manage the PNH clinic and recruited the patient cohort; Y.M.E.-S. performed the experimental work; Y.M.E.-S., G.M.D., and G.P.C. analyzed the data; G.P.C. and Y.M.E.-S. designed the study; and G.P.C. wrote the article with input from all authors.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Graham Cook; Leeds Institute of Cancer and Pathology, University of Leeds, Wellcome Brenner Building, St James's University Hospital, Leeds LS9 7TF, UK; e-mail: g.p.cook@leeds.ac.uk.