LAMP1/CD107a is required for efficient perforin delivery to lytic granules and NK-cell cytotoxicity

Natural killer (NK) cells comprise a subset of lymphocytes involved in protection against microbial pathogens and tumors.¹  Because of their cytotoxic capacity, NK cells are able to directly eliminate abnormal cells. The killing of these target cells is a multistage process that culminates with the localized secretion of lytic granules, containing perforin and granzymes, at the immunologic synapse.^2,3  Upon delivery to a target cell, perforin generates pores in the membranes,^3-5  allowing granzymes to access target-cell cytoplasm and induce apoptosis.^3,6,7  Defects in lytic granule secretion are associated with serious diseases, such as hemophagocytic lymphohistiocytosis and Chediak-Higashi, Hermansky-Pudlak, or Griscelli syndrome.^8-11  Knowledge about the proteins involved in regulation of lytic granule release is very limited. Soluble N-ethylmaleimide-sensitive factor attachment protein receptor proteins and small guanosine triphosphatases have been shown to be required for NK-cell degranulation and tumor killing,^12-16  but the granule-specific protein machinery involved in NK-cell exocytosis is poorly understood.

Lytic granules have the characteristics of lysosomes and are referred to as “secretory lysosomes.”^17,18  There are many proteins residing in lysosomal membranes that function in acidification of the lysosomal lumen, protein import, membrane fusion, and transport of degradation products to the cytoplasm.¹⁹  Surprisingly, the role of the most abundant proteins, lysosome-associated membrane protein (LAMP) 1 and LAMP2 (CD107a and CD107b, respectively), is largely unknown. In humans, mutations in LAMP2 cause Danon disease, believed to arise from aberrant autophagy.²⁰  LAMP2-deficient mice also accumulate autophagic vacuoles in several tissues.²¹  LAMP1-deficient mice have an almost normal phenotype but have elevated levels of LAMP2, suggesting that LAMP2 can compensate for the loss of LAMP1.²²  Recent studies using LAMP-depleted fibroblasts demonstrate that both proteins are required for proper lysosomal transport and efficient (auto)phagosome-lysosome fusion.^23,24 

Although LAMP1 is widely used as a marker for the identification of NK-cell degranulation,²⁵  as it appears on the cell surface following the fusion of lysosomes with the plasma membrane, the role of LAMP1 in NK-cell biology is unknown. Since LAMP1 is one of the most abundant proteins in the lysosomal membrane and NK cells use their lysosomes to induce the death of target cells, we sought to determine the importance of LAMP1 in NK-cell cytolytic activity.

Please refer to the supplemental Methods (see the Blood Web site) for the detailed description of methods and reagents used.

Antibodies (Ab’s) used included the following: anti-LAMP1, anti-LAMP2, and anti–granzyme B (Santa Cruz Biotechnology or eBiosciences); anti–Early Endosome Antigen 1 (EEA-1), anti-p150^glued, and anti–adaptin γ, α, and β (BD); anti–Ras-associated binding (Rab)7 and anti-Rab9 (Cell Signaling); anti-actin (Sigma); anti-pericentrin and anti–cation-independent mannose-6-phosphate receptor (CI-MPR) (Abcam); and anti-perforin (Mabtech, BioLegend, or Cell Sciences).

YTS, 721.221, and 293T cells were grown as described previously.¹⁶  YTS cells, transduced with short hairpin RNA (shRNA), were grown in complete RPMI 1640 medium with puromycin (2 μg/mL). NK92 cells were cultured in RPMI 1640 medium with interleukin 2 (IL-2) (100 U/mL). Blood samples from healthy volunteers were collected at the Department of Transfusion Medicine, National Institutes of Health (NIH), under protocol 99CC-0168, and used to isolate NK cells. NK cells were cultured in X-vivo medium (Invitrogen) supplemented with 500 U/mL of IL-2.

LAMP1 and adaptin γ short interfering RNA (siRNA) or vector-based shRNA was from Sigma. For YTS cells, nontargeting shRNA (Sigma) was used as a negative control, whereas for ex vivo NK cells, a scrambled siRNA was used. Both nontargeting shRNA and scrambled siRNA are collectively referred to as control (CTRL) RNA interference (RNAi). Generation of lentivirus particles and infection of YTS cells was done as described by Krzewski et al.¹⁶  siRNA was delivered to ex vivo isolated NK cells by nucleofection using Nucleofector II (Lonza), and the cells were analyzed 72 hours after the procedure.

Total RNA was isolated with RNAqueous-4PCR kit (Ambion). Complementary DNA (cDNA) was generated with qScript cDNA Synthesis Kit (Quanta) and served as template for real-time PCR, using SYBR Green Master Mix and LightCycler 480 (Roche). Primers for real-time PCR were from Qiagen. The amount of the target gene messenger RNA (mRNA) was calculated from the standard curve and normalized to actin mRNA.

For immunoblotting, cell lysates or cell fractions were probed with the Ab’s indicated in the text. Immunoblots were developed using ChemiGlow West Substrate (Cell Biosciences). The images were acquired with FluorChem-Q imager (Protein Simple), using AlphaView (version 3.3) and automatic exposure.

NK-cell cytotoxicity was evaluated by Dissociation-Enhanced Lanthanide Fluorescent Immunoassay (Perkin-Elmer). Lysis percentage was calculated as described by Krzewski et al.¹⁶ 

YTS or NK cells were fixed, permeabilzed with Cytofix/Cytoperm buffer (BD), and stained with anti-LAMP1–fluorescein isothiocyanate, anti-LAMP2–AlexaFluor 647, and/or anti-perforin Ab, conjugated to fluorescein isothiocyanate or phycoerythrin.

Delivery of granzyme B to 721.221 target cells was assessed using GranToxiLux kit (OncoImmunin). In this assay, target cells are labeled with a cell-permeable fluorogenic granzyme B substrate; upon delivery of granzyme B to the target cell, the substrate is cleaved resulting in increased fluorescence in target cells.²⁶  Data acquisition and analysis were done using FACSort (BD) and FlowJo (version 7.6; Tree Star).

Activity of granzyme B in cell lysates was assessed according to Thiery et al.²⁷ 

The assay was performed as described in Krzewski et al.¹⁶ 

YTS cells were conjugated to 721.221 target cells at a 1:1 ratio at 37°C. Fixed and permeablized cells were stained with the Ab’s indicated in text. For the double staining of perforin, the cells were first stained with anti-perforin B-D48 Ab, followed by IgG1-specific DyLight 549–conjugated anti-mouse Ab, blocked with 5% mouse serum, and stained with directly conjugated anti-perforin δG9 Ab. Cells were visualized by a Zeiss LSM510 Axiovert-200M confocal microscope at room temperature. The images were obtained using 63× Plan-Apochromat objective and LSM510 (version 3.2).

The determination of characteristics of perforin clusters and colocalization analysis were done using ImageJ (version 1.45; NIH) and Imaris (version 7.3; Bitplane), respectively, as described in the supplemental Methods.

For live cell imaging, YTS cells were labeled for 30 minutes with 75 nM LysoTracker Red DND-99 (Invitrogen) and placed in Laboratory-Tek chambered coverglass (Nunc) in phenol-red free RPMI. Cells were imaged in all 3 planes for 3 minutes at 37°C using an Olympus IX81 microscope with 100× PlanApo objective. Image acquisition was performed with MetaMorph (version 7.7.3; Molecular Devices) using 0.5-μm z-axial dimension and 100-millisecond exposure/frame. Three-dimensional vesicle tracking was done using Imaris, using an estimated diameter of 0.3 μm, autoregressive motion tracking algorithm, 1.5 maximum distance, 3 maximum gap size, and track duration >2 seconds; track speed was calculated as the average speed of granule movement during the imaging period.

YTS cells were homogenized using Dounce homogenizer. Cell debris and nuclei were pelleted (1000g_av, 10 minutes), and the postnuclear supernatant was centrifuged (17 000g_av, 15 minutes) to pellet the crude lysosomal fraction (CLF). The CLF was adjusted to 25% iodoxanol. Ten percent to 40% discontinuous gradient of iodoxanol was centrifuged for 2 hours at 95 000g_av. In fractionation experiments, the postnuclear supernatants were overlaid on top of a 5% to 30% gradient of iodoxanol. The gradient was centrifuged at 52 000g_av for 18 hours.

To investigate the role of LAMP1 in NK-cell function, we evaluated the effect of LAMP1 silencing on NK-cell cytotoxicity. We generated YTS and NK92 NK cell lines with a stable LAMP1 knockdown (Figure 1A-B; supplemental Figure 1). Based on densitometric analysis of anti-LAMP1 immunoblotting, the estimated efficacy of LAMP1 knockdown was ∼75%. With siRNA-mediated silencing of LAMP1 in ex vivo isolated IL-2–cultured human NK cells, we achieved a 40% to 50% decrease of LAMP1 protein. The lower LAMP1 knockdown level achieved in the transiently siRNA-transduced ex vivo NK cells was at least partially due to the stability of LAMP1; LAMP1 has a half-life of at least 30 hours, and the time point at which NK cells were analyzed (72 hours) was likely too short for stronger or full silencing of LAMP1 (supplemental Figure 2). LAMP1 RNAi had no effect on LAMP2 mRNA levels (supplemental Figure 3), confirming the specificity of RNAi.

We used LAMP1 RNAi cells to assess their ability to kill a tumor cell line, 721.221 or K562 (Figure 1C; supplemental Figure 1). We found that compared with untransduced YTS cells, or cells transduced with nontargeting control RNAi, LAMP1 RNAi cells had a 70% to 76% decrease in cytotoxicity. Likewise, LAMP1 silencing decreased the cytotoxicity of NK92 cells by ∼65%. We obtained similar results using ex vivo isolated NK cells, where a transient knockdown of LAMP1 reduced cytotoxicity by 40% to 60%. We also used Ab-dependent cell-mediated cytotoxicity to evaluate the killing ability of LAMP1 RNAi NK cells (Figure 1D). Although CD16-mediated recognition of anti-Human Epidermal Growth Factor Receptor 2 (HER2) Ab, bound to the cell surface of SK-OV3 cells, resulted in pronounced lysis of target cells by control-transduced NK cells, cytotoxicity by LAMP1 RNAi NK cells was decreased by 35% to 40%. Thus, LAMP1 plays an important role in NK-cell cytolytic activity.

NK cells kill target cells by secreting granzyme-containing lytic granules at the immunologic synapse. We measured the ability of LAMP1 RNAi NK cells to deliver granzyme B to target cells. The delivery of granzyme B from LAMP1 RNAi YTS cells to target cells was decreased fivefold when compared with untransduced or control RNAi-transduced YTS cells (Figure 2A). We verified these results in NK cells, where LAMP1 silencing blocked the transfer of granzyme B from IL-2–stimulated NK to tumor cells almost twofold (Figure 2A). The smaller effect observed in the case of ex vivo NK cells was likely due to less efficient knockdown of LAMP1 in these cells and was consistent with the results of cytotoxicity assays. LAMP1 RNAi cells had normal levels of granzyme B (Figure 2B), and the activity of granzyme B from LAMP1 RNAi cells was undistinguishable from the activity observed in control cells (Figure 2C). LAMP1 RNAi cells formed conjugates with target cells at the same level as control cells (Figure 2D), excluding the possibility that blockage of granzyme B delivery and cytotoxicity resulted from inadequate cell-cell adherence. LAMP1 silencing did not affect the expression of NK-activating receptors on the cell surface; moreover, LAMP1-silenced cells accumulated Lymphocyte Function-associated Antigen 1 and filamentous actin, essential for NK-cell activation and cytotoxicity, at the immunologic synapse similar to nonsilenced cells, demonstrating that the synapse formation and NK-cell activation was likely not impaired in those cells (supplemental Figure 4). Following the stimulation with target cells, LAMP1 RNAi cells were able to upregulate LAMP2 and LAMP3 (CD63) on the cell surface (albeit at a slightly lower frequency than the controls), indicating that the lytic granules were able to fuse with the plasma membrane (supplemental Figure 5).

NK-cell cytotoxicity depends on the ability to transport their lytic granules to the immunologic synapse. Therefore, we next evaluated the granule transport to the cell-cell interface. As the results using YTS cells mimicked the data obtained with ex vivo NK cells, we used the YTS model to study the localization of perforin-containing lytic granules in response to activation with target cells.

LAMP1 RNAi cells translocated the lytic granules toward the immunologic synapse (Figure 3A). The microtubule organizing center (MTOC), as detected by pericentrin staining, was positioned close to the immunologic synapse, just as in control cells, indicating that the migration of the MTOC was not affected by LAMP1 silencing. There was no significant difference in perforin polarization (Figure 3B); however, perforin appeared to be more dispersed around the MTOC in LAMP1 RNAi cells (Figure 3A). The analysis of shapes of perforin-positive vesicle clusters revealed significant changes in shape descriptors. The main area occupied by perforin in LAMP1 RNAi cells had smaller solidity and circularity, indicating that with reduced levels of LAMP1, lytic granules did not form tight clusters and were more spread out (supplemental Figure 6). We also observed a dispersed distribution of lytic granules when using LAMP2 as another marker for the granules (not shown). Thus, although the MTOC and perforin are able to move toward the immunologic synapse in response to cell activation, the lytic granules remain dispersed and do not cluster efficiently in LAMP1 RNAi cells.

The previous results suggested a possible defect in granule transport in LAMP1 RNAi cells. Hence, we performed three-dimensional time-lapse imaging of LysoTracker-labeled lytic granules in live cells (supplemental Videos 1 and 2) and plotted the trajectories formed by moving vesicles. We found that granules in LAMP1 RNAi cells formed shorter tracks, with a smaller displacement (ie, the straight-line distance between the beginning and end of the track) and decreased velocity (Figure 4A-B).

The activity of the motor complex dynein-dynactin is essential for the movement of lytic granules toward the MTOC.²⁸  We isolated lytic granules from control and LAMP1 RNAi cells and determined the levels of the dynein/dynactin complex regulatory subunit, p150^glued. Cells with LAMP1 knockdown had less p150^glued recruited to their lytic granules (Figure 4C). Rab7, directly involved in the recruitment of p150^glued, was present to the same extent on the granules from control and LAMP1 RNAi cells, suggesting that the decreased recruitment of dynactin is not related to changes in Rab7 levels. The lysosome-specific proteins, granzyme B and LAMP2, were absent in cytoplasmic fractions but present in the lysosomal fractions at similar levels, verifying the purity of the lysosomal fractions and indicating that the observed loss of p150^glued was specific. Thus, LAMP1 silencing leads to lytic granule movement impairment, likely due to the decreased recruitment of motor protein(s) to the granules.

As the altered rate of lytic granule transport in LAMP1 RNAi cells was likely not sufficient to fully block their cytolytic capacity, we explored if other factors could account for the reduced cytotoxicity. During imaging studies, we observed that LAMP1 RNAi cells had weaker anti-perforin staining (Figure 3A), suggesting that these cells could have less perforin. Reconstruction of multiple optical sections of YTS cells stained with anti-perforin Ab (Figure 5A) confirmed that, in addition to the greater dispersion of perforin-containing granules, the anti-perforin staining was much weaker in LAMP1 RNAi than in control cells. The fluorescence of perforin clusters was decreased by ∼35% when compared with untransduced or control RNAi cells (Figure 5B). Flow cytometry analysis revealed a 40% decrease in perforin level in LAMP1-silenced YTS cells when compared with control cells; we obtained similar results using ex vivo isolated IL-2–stimulated NK cells (Figure 5C). This difference was observed only at the protein level, as perforin mRNA was not affected by LAMP1 silencing (Figure 5D), and did not result from protein degradation, as the half-life of perforin was comparable between control and LAMP1 RNAi cells (supplemental Figure 2).

The Ab used for perforin detection (clone δG9) reportedly recognizes the mature, lysosomal form of perforin.^29,30  Another anti-perforin Ab, D48, has been described to detect multiple forms of perforin (ie, newly synthesized as well as lysosomal).^29-31  Therefore, we used these 2 Ab’s to determine if LAMP1 silencing affects the total amount or only the granule-associated perforin in the cell. Perforin staining with δG9 Ab showed that fluorescence of granule-contained perforin was decreased following LAMP1 knockdown, whereas the fluorescence profiles of anti-perforin staining using D48 Ab were virtually the same (Figure 5E; supplemental Figure 7). Additionally, granules isolated from LAMP1 RNAi cells had a similar level of granzyme B but less perforin than granules from control RNAi cells, whereas the amount of total perforin in cell lysates did not differ (Figure 5F).

The loss of granule-associated perforin could result from its missorting, leading to increased secretion or trapping of perforin in nonlysosomal compartments. We were unable to detect any changes in the amount of perforin released to the cell culture supernatants (not shown), indicating that perforin is likely not prematurely secreted from cells after LAMP1 silencing. To address the second possibility, we analyzed cell fractionations for perforin content. In control and LAMP1 RNAi cells, the distribution of markers for several cellular organelles (EEA-1, Rab9, Rab7, and LAMP2) was the same (Figure 6A). Granzyme B was present in fractions positive for Rab7 and LAMP2, indicating that following LAMP1 knockdown granzyme B is properly sorted into lysosomes. In control RNAi cells, the majority of perforin was present in the fractions containing lysosomal markers. In LAMP1 RNAi cells, however, some perforin localized to lesser density fractions negative for LAMP2, granzyme B, or Rab7/9, demonstrating that in the absence of LAMP1, perforin does not traffic properly to lysosomes and is partially retained in nonlysosomal compartments. The nonlysosomal fractions positive for perforin in LAMP1 RNAi cells were also positive for 1 of the 2 MPR forms and overlapped with adaptin γ (Figure 6A), suggesting that in these cells perforin could be accruing in trans-Golgi network (TGN)–derived transport vesicles. To verify this hypothesis, we analyzed colocalization between perforin and MPR in YTS cells with normal or reduced LAMP1 expression.

Although the lysosomal perforin (δG9-reactive) minimally overlapped with MPRs (supplemental Figure 8A-B), the staining with D48 anti-perforin Ab showed evident colocalization with CI-MPR (Figure 6B). In control RNAi cells, 23% of perforin overlapped with CI-MPR, whereas 40% of perforin colocalized with CI-MPR in LAMP1 RNAi cells (Figure 6C). The overlap coefficient between perforin and CI-MPR was also significantly increased in cells with LAMP1 knockdown when compared with control cells (Manders coefficient M = 0.4 vs 0.21, respectively). LAMP1 was also found to partially colocalize with CI-MPR (supplemental Figure 8C). More perforin-positive vesicles were also positive for adaptin γ, and negative for LAMP2 in LAMP1 RNAi cells (supplemental Figure 9 and data not shown), further supporting the idea that without LAMP1, perforin is retained in transport vesicles and does not reach the lytic granules. The defect in perforin trafficking was not due to a general impairment of the protein transport, as the cell surface expression of NK-cell receptors and the ability of NK cells to secrete tumor necrosis factor α and interferon γ were not significantly affected by LAMP1 silencing (supplemental Figures 4A and 10). Thus, disruption of LAMP1 expression alters perforin trafficking and causes retention of perforin in transport vesicles, resulting in decreased perforin content in lytic granules.

The adaptor protein complex AP-1, important for sorting proteins leaving the TGN, is known to directly interact with LAMP1.³²  Therefore, we analyzed the effect of silencing the AP-1 complex subunit, adaptin γ, on perforin location in YTS cells. Adaptin γ silencing resulted in increased association of perforin with MPR-positive vesicles (Figure 7A). In adaptin γ RNAi cells, ∼51% of perforin overlapped with CI-MPR, whereas only 21% overlapped in control RNAi cells (Figure 7B). The overlap coefficient increased from 0.16 in control RNAi cells to 0.39 in adaptin γ RNAi cells. Adaptin γ knockdown had a similar effect on cytotoxicity as the silencing of LAMP1 (Figure 7C). Thus, disruption of AP-1 complex expression also causes retention of perforin in transport vesicles and produces effects similar to LAMP1 silencing, indicating that these 2 binding partners are important for proper trafficking of perforin to lytic granules and NK-cell cytotoxicity.

Although recent reports have highlighted the role of several proteins involved in lytic granule transport and exocytosis,^12-16,33  the function of granule-specific proteins in NK-cell cytotoxicity is poorly characterized. Here, we investigated the role of the lytic granule protein LAMP1 in the cytotoxicity of NK cells. We found that LAMP1 silencing resulted in inhibition of NK-cell cytotoxicity, demonstrating for the first time the importance of LAMP1 in this process.

In order to kill, NK cells must first translocate the MTOC and polarize their lytic granules toward the cell-cell contact site, a step that is required for the subsequent release of lytic granule content at the immunologic synapse.^2,3  YTS NK cells were able to form immunologic synapses, and move their MTOC and granules toward the immunologic synapse, regardless of the level of expression of LAMP1; contrary to cells with normal expression of LAMP1, in LAMP1 RNAi cells the granules were more dispersed and did not form thick clusters. This finding is consistent with a previous report showing that lysosomes in lamp1/lamp2-deficient mouse fibroblasts have severely reduced motility, supporting the hypothesis that LAMPs are involved in the transport of lysosomes.²⁴  Our analysis of the spatiotemporal movement of LysoTracker-labeled granules revealed that the reduction of LAMP1 expression is sufficient to disturb motility of lysosomes in NK cells and likely contributes to decreased cytotoxicity. The contribution of LAMP2 to lytic granule movement in NK cells is unknown, but the endogenous LAMP2 appears unable to fully compensate for defects in lysosomal movement caused by LAMP1 silencing.

Trafficking of lytic granules in NK cells depends on the activity of minus end–directed motor protein complex dynein-dynactin.²⁸  We found that lytic granules isolated from cells with LAMP1 knockdown had less p150^glued; the decreased recruitment of motor protein to lytic granules could explain the defect in granule movement caused by LAMP1 silencing. Interestingly, we found that the presence of Rab7, responsible for recruiting dynactin-dynein to lysosomes and endosomes,^34,35  was unaffected by LAMP1 knockdown, indicating that the LAMP1-mediated effect is independent of Rab7. Recruitment of dynein-dynactin to Rab7 is not sufficient to activate the motor complex, and dynein-dynactin has to be transferred from Rab7 to βIII spectrin to initiate transport of vesicles.³⁴  Whether LAMP1 could be involved in this process and/or directly bind dynein-dynactin or other motor proteins is unknown at present. The cytoplasmic tail of LAMP1 is very short and, with the exception of a lysosome-sorting sequence, lacks clearly defined signaling motifs. There is a paucity of information regarding binding partners of LAMP1, and we are analyzing the proteins that interact with LAMP1 to determine how LAMP1 could be involved in motor protein recruitment to lytic granules.

We found that disruption of LAMP1 expression not only slowed granule movement, but also interfered with perforin recruitment to lytic granules. Lysosomal membrane proteins are known to be important for sorting lysosome luminal proteins, and disruption of their function can result in secretion of lysosomal enzymes instead of sorting to lysosomes. For instance, deficiency of the lysosomal integral membrane protein 2 results in premature secretion of β-glucocerebrosidase.³⁶  We did not find any evidence of increased release of perforin from NK cells, even though LAMP1 silencing caused an alteration in perforin trafficking. The distribution of EEA-1, LAMP2, Rab7, and Rab9 was the same between cells with normal and reduced expression of LAMP1, indicating that the endolysosomal system was not disturbed following LAMP1 knockdown; this agrees with a previous report showing that LAMP1-deficient cells have normal lysosomal morphology.²²  We found, however, that whereas the distribution of granzyme B was unchanged in LAMP1 RNAi cells, a pool of perforin was present in compartments that were negative for protein markers of late endosomes and lysosomes, indicating that perforin was retained in nonlysosomal structures in these cells. The nonlysosomal fractions containing perforin were positive for markers associated with vesicle transport and sorting, and we found increased colocalization of perforin with CI-MPR in LAMP1 RNAi cells, suggesting that perforin is present in trans-Golgi–derived transport vesicles. The presence of perforin in MPR-positive vesicles is somewhat surprising, as perforin has been reported to lack mannose-6-phosphate modification.³⁷  On the other hand, lysosomal proteins are able to use MPR-positive carrier vesicles even though they do not have the modification; for example, neuraminidase uses the interaction with mannose-6-phosphate–modified cathepsin A to reach lysosomes.³⁸  Concurring with our findings, a recent report demonstrating the importance of perforin glycosylation in its traffic to lytic granules postulates that perforin primarily uses the MPR-dependent pathway to reach lytic granules.³⁹ 

Regardless of the trafficking path, LAMP1 appears to be important in routing perforin to lytic granules. The lack of complete blocking of perforin transport to lysosomes is not surprising. It is not uncommon for lysosomal proteins to use different pathways to reach lysosomes. For example, a portion of lysosomal enzymes that normally use the MPR pathway is still sorted to lysosomes in cells from I-cell disease patients (that do not have a functional MPR pathway).⁴⁰  Moreover, even though sorting of LAMP1 to lysosomes is mediated by adaptor complexes AP-1, AP-2, or AP-3, disrupting the function of any of these complexes does not completely block the delivery of LAMP1 to lysosomes.^32,41,42 

The interaction between LAMP1 and AP-1 complex is intriguing, as AP-1 is not required for the sorting of LAMP1.⁴³  AP-1 has been shown to interact with motor protein kinesin and function in delivery of cargo to lysosome-related melanosomes, and to facilitate movement of vesicles from the TGN.^44,45  That prompted us to ask whether LAMP1/AP-1 binding could play a role in the trafficking of perforin-containing transport vesicles to lytic granules. We found that disruption of expression of adaptin γ caused alteration of perforin trafficking and inhibited cytotoxicity, similar to effects observed in LAMP1 RNAi cells. Our data suggest that the direct interaction between LAMP1 and AP-1³²  could mediate transport of LAMP1-positive vesicles, containing perforin, to lytic granules. The lack of LAMP1 could prevent binding of AP-1 to perforin-positive vesicles and result in retention of perforin in the transport vesicles instead of its delivery to lytic granules. Similarly, the lack of AP-1 complex would prevent proper perforin transfer from the TGN to lytic granules because there would be no adaptor to interact with LAMP1 and enable the transport (Figure 7D).

Perforin is essential for the lytic activity of cytotoxic lymphocytes.^13,46  It is required for delivery and release of granzyme B to the cytoplasm of target cells.^47,48  Even a small decrease in perforin expression correlates with marked reduction in cytotoxicity.⁴⁹  NK cells do not release all of their lytic granules at once, only secreting a subset of the lytic granule pool following activation.⁵⁰  In addition, not all of the fusion events lead to full pore opening and release of granule content.⁵¹  Furthermore, perforin is processed in lysosomes to give a mature form of the protein; an interesting possibility is that the impaired delivery of perforin to lysosomes could also affect the rate of its maturation. Thus, a decreased amount of perforin in lytic granules, or its improper maturation, could have a detrimental impact on NK-cell cytotoxicity. In this light, our data show that disruption of LAMP1 expression has a pleiotropic effect on NK cells, causing impairment of trafficking of a critical lytic granule protein, perforin, and slower movement of lytic granules. This results in the inability of NK cells to deliver granzyme B to target cells and inhibition of NK-cell cytotoxicity. Thus, LAMP1 is not only a marker of NK-cell degranulation, but also a crucial component of NK-cell cytotoxicity.

The authors thank Laurette Femnou and Claudine Jones for their support and technical help, Eric Long and Linjie Tian for their critical comments, and the NIH Clinical Center Department of Transfusion Medicine for providing the blood samples.

This work was supported by the intramural programs of the National Institute of Allergy and Infectious Diseases.

National Institutes of Health

Contribution: K.K., A.G.-K., V.N., and G.P. performed the experiments; K.K., A.G.-K., and V.N. analyzed the results; K.K. and A.G.-K. designed experiments; and K.K. and J.E.C. wrote the manuscript.

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

Correspondence: Konrad Krzewski, National Institute of Allergy and Infectious Diseases, National Institutes of Health, 12441 Parklawn Dr, Twinbrook II, Room 205, Rockville, MD 20852; e-mail: krzewskikj@niaid.nih.gov; and John E. Coligan, National Institute of Allergy and Infectious Diseases, National Institutes of Health, 12441 Parklawn Dr, Twinbrook II, Room 205, Rockville, MD 20852; e-mail: jcoligan@niaid.nih.gov.