Revisiting KIR restriction and CD94 expression in diagnostic criteria for NK–large granular lymphocytic leukemia

TO THE EDITOR:

Recent diagnostic criteria for natural killer (NK)–large granular lymphocytic leukemia (NK-LGLL) proposed by Pastoret et al¹ incorporate immunophenotypic and molecular features, including killer cell immunoglobulin-like receptor (KIR) restriction and CD94/NKG2A overexpression by flow cytometry, as major diagnostic criteria. The authors describe a weighted scoring system combining 4 major criteria (NK-cell lymphocytosis ≥1.0 × 10⁹/L, KIR restriction, CD94 or NKG2A overexpression, and STAT3/5B, CCL22, or TET2 mutations without other myelodysplastic syndrome–defining alterations) and 4 minor clinicopathologic features, with diagnosis requiring either 3 major criteria or 2 major and 2 minor criteria.¹ Although this framework represents an important step toward standardizing diagnosis, the performance of these parameters depends critically on how they are defined and measured. Based on our experience evaluating NK-cell clonality by flow cytometry, we identified important limitations in the current definitions of KIR restriction and CD94 overexpression that may reduce the sensitivity and specificity of the proposed scoring system.

First, although KIR restriction is an important surrogate marker of NK-cell clonality, the operational definition proposed by Pastoret et al,^1,2 that is, reduced expression of at least 2 KIR markers (CD158a <9%, CD158b <12%, and CD158e/NKB1 <4% of NK cells), does not adequately account for the known heterogeneity and combinatorial expression of KIR receptors. KIR belongs to the type I transmembrane glycoprotein immunoglobulin family, with inhibitory or activating function.^3,4 To date, at least 15 KIR proteins have been identified, and these are heterogeneously expressed on each NK cell.⁵ Using a single-cell reverse transcription polymerase chain reaction method to evaluate 7 KIRs, it was demonstrated that each NK cell expressed 2 to 8 KIRs, resulting in the simultaneous expression of multiple KIRs on each NK cell.⁶ Therefore, clonal expansions may demonstrate restricted yet multiantigen expression patterns that exceed the thresholds proposed by Pastoret et al.¹ In fact, dominant clonal NK populations with coexpression of ≥2 CD158 antigens as well as clonal NK subsets with single CD158 antigen expression in a polyclonal background would be excluded from classification as KIR restricted.

Given the limited commercial availability of CD158 antibodies, many laboratories, including ours, assess CD158 (KIR) expression using 3 major markers: CD158a, CD158b, and CD158e (NKB1). To systematically characterize KIR expression, we proposed a combinatorial gating strategy after NK-cell identification (Figure 1A).⁷ Total CD158a, CD158b, and CD158e expression were first quantified, after which CD158b⁻ and CD158b⁺ populations were further subdivided based on CD158a and CD158e expression, yielding 8 discrete phenotypic combinations (A+B−E−, A−B+E−, A−B−E+, A+B+E−, A+B−E+, A−B+E+, A+B+E+, and A−B−E−). This framework captures the inherent heterogeneity of KIR expression, in which NK cells within a given sample may express 0, 1, 2, or all 3 measured CD158 antigens (Figure 1A), and enables identification of clonal populations exhibiting restricted yet combinatorial antigen expression. These combinatorial patterns are summarized across 145 healthy donors (Figure 1B), with each bar representing the 99% nonparametric reference interval and dashed lines indicating medians. Total CD158a, CD158b, and CD158e expression are shown on the right, with medians denoted by dashed lines and thresholds defined by Pastoret et al indicated by orange lines.

Figure 1.

Combinatorial analysis of KIR (CD158) expression and limitations of threshold-based KIR restriction criteria. (A) Combinatorial gating strategy for evaluation of KIR (CD158) expression in NK cells. After identification of NK cells, total CD158a, CD158b, and CD158e expression is quantified. CD158b⁻ and CD158b⁺ populations are further subdivided based on CD158a and CD158e expression, yielding 8 discrete phenotypic combinations (A+B−E−, A−B+E−, A−B−E+, A+B+E−, A+B−E+, A−B+E+, A+B+E+, and A−B−E−). This framework captures the combinatorial and heterogeneous nature of KIR expression at the single-cell level. (B) Distribution of CD158 expression patterns in 145 healthy donors. Bars represent 99% nonparametric reference intervals for each combinatorial phenotype, with dashed lines indicating medians. Total CD158a, CD158b, and CD158e expression is shown on the right, with dashed lines denoting medians and orange lines indicating thresholds proposed by Pastoret et al. (C) Representative patient demonstrating clonal NK-cell expansion with restricted coexpression of CD158a and CD158e in the absence of other CD158 antigens. Despite biologically restricted expression, both markers exceed proposed thresholds (>98%), and the case would not be classified as KIR restricted under current criteria. (D) Representative patient demonstrating an apparent CD158a-restricted NK-cell population (A+B−E−) within a subtle polytypic background. Minor NK-cell subsets expressing additional CD158 markers result in CD158b exceeding the proposed threshold, leading to exclusion from classification as KIR restricted. (E) Assessment of CD94 expression using CD3⁺CD8⁻ T cells as an internal negative control. In healthy donors, most NK cells demonstrate high CD94 expression, with most cases exceeding the proposed diagnostic threshold (>77%), highlighting the limited specificity of this criterion. SID, KIR: killer cell immunoglobulin-like receptor.

Using this framework, we demonstrate that clonal expansion of NK cells coexpressing multiple CD158 markers can result in aggregate CD158 expression exceeding the thresholds proposed by Pastoret et al.^1,2 Consequently, such patients would not be classified as KIR restricted under the current definition. This is illustrated by a representative patient (Figure 1C), in which the clonal NK-cell population demonstrates restricted coexpression of CD158a and CD158e in the absence of CD158b; total expression of CD158a and CD158e both exceed 98% and thus does not satisfy the proposed definition. In our previously published cohort of 56 patients with NK-LGLL,⁷ 11 patients (19.6%) demonstrated coexpression of ≥2 CD158 antigens and would be excluded from classification as KIR restricted under the proposed criteria: 4 patients with A+B+E−, 4 with A−B+E+, 2 with A+B−E+, and 1 with A+B+E+ patterns (Table 1; cases 1-11). In addition, 3 patients (5.3%) exhibited apparent single-antigen restriction within a subtle polytypic background, resulting in 2 CD158 markers exceeding threshold values and similarly failing to meet criteria (Table 1; cases 12-14). A representative example (Figure 1D) demonstrates a dominant CD158a-restricted population (A+B−E−), in which minor NK-cell subsets expressing A−B+E− and A+B+E− lead to CD158b slightly exceeding the threshold, thereby excluding the patient from classification as KIR restricted. Overall, 14 of the 56 patients (25%) with biologically evident KIR restriction would not be classified as such using the criteria proposed by Pastoret et al.^1,2 This limitation may lead to underdiagnosis of NK-LGLL, particularly in settings in which molecular testing is unavailable, inconclusive, or does not definitively establish clonal origin, including patients in which no mutations are detected (approximately half of patients with NK-LGLL).^1,8, 9, 10

Table 1.

Clonal NK-cell populations from patients with NK-LGLL exhibiting restricted KIR expression were excluded from classification as KIR restriction according to the criteria by Pastoret et al

Case	KIR restriction pattern	CD158a, %	CD158b, %	CD158e, %
1	A+B+E−	96	95	0
2	A+B+E−	96	95	0
3	A+B+E−	85	100	0
4	A+B+E−	90	82	4
5	A−B+E+	4	79	87
6	A−B+E+	6	75	54
7	A−B+E+	0	75	64
8	A−B+E+	8	54	89
9	A+B−E+	99	0	99
10	A+B−E+	34	34	66
11	A+B+E+	35	94	44
12	A+B−E−	92	15	1
13	A+B−E−	74	5	9
14	A−B+E−	11	86	1

Percentages in red exceed the proposed cutoffs (CD158a <9%, CD158b <12%, and CD158e <4%).

Second, we question the use of CD94 overexpression (>77%) as a major diagnostic criterion because of limited specificity. The C-type lectin receptor CD94 is expressed on all human NK cells, and many normal individuals exhibit high levels of CD94 expression.¹¹ Given the continuum of CD94 expression observed by flow cytometry, an appropriate negative control is essential. In our laboratory, CD3⁺CD8⁻ T cells, which typically show limited CD94 expression, are used as a negative control to more accurately estimate the proportions of CD94⁺ and CD94⁻ NK-cell populations (Figure 1E). With this internal control, we have found that all 145 healthy donors had CD94 expression in ≥60% of the NK cells, and 125 (86%) had CD94 expression >77%, indicating a low specificity using CD94 >77% as a major criterion for the diagnosis of NK-LGLL.⁷ In addition, reactive NK cells in response to cytomegalovirus infection or reactivation often uniformly express CD94 and NKG2C (both >77%), further arguing against the use of CD94 >77% as a major criterion for the diagnosis of NK-LGLL.12, 13, 14

In summary, although the scoring system proposed by Pastoret et al represents an important step toward standardized diagnosis of NK-LGLL, its performance is highly dependent on the validity of its individual components. Our findings demonstrate that the current definition of KIR restriction has poor sensitivity because of failure to account for combinatorial CD158 receptor expression, whereas CD94 overexpression has insufficient specificity because of its high prevalence in normal and reactive NK-cell populations. These results highlight a fundamental limitation of 2 of the 4 major criteria for the diagnosis of NK-LGLL and underscore the need for more refined criteria in clinical practice.

Conflict-of-interest disclosure: E.H. has sponsored research from Eli LIlly and Equity - Abcon Therapeutics.