Immature Neutrophils Released in Acute Inflammation Exhibit Efficient Migration despite Incomplete Segmentation of the Nucleus

This article is featured in In This Issue, p.1

Human neutrophils with a banded, horseshoe-shaped nucleus represent the last differentiation stage preceding mature neutrophils. At this stage, the segmentation of the characteristic polymorphonuclear nucleus of the neutrophils has not completed yet. In homeostasis, “band cells,” or banded neutrophils, reside in the bone marrow, but acute inflammation has long been known to increase the number of banded neutrophils in the circulation (1). This so-called “left shift” is one of the characteristics of the process termed emergency granulopoiesis (2). In this emergency response to severe systemic inflammation, more neutrophils are recruited to the circulation to respond to an increased burden of microbe-associated molecular patterns or damage-associated molecular patterns (2). Both immature and mature neutrophils are released from the bone marrow, and granulopoiesis is enhanced to produce de novo neutrophils (2). When the infection is cleared or the tissue damage has healed, neutrophil numbers will return to homeostatic numbers. The left shift has long been interpreted as a bystander effect of the call for more circulatory neutrophils, causing the bone marrow to release neutrophils prematurely.

However, circulatory banded neutrophils were recently shown to have superior antibacterial capacity in vitro, despite their immature morphology (3). This surprising result makes for an attractive concept in which banded neutrophils are not released as bystanders but are deliberately released as cells highly adept at pathogen killing.

In the study by Leliefeld et al. (3) and in the study here described, acute inflammation was experimentally induced in healthy subjects by i.v. administration of bacterial LPS, also known as endotoxin. Experimental endotoxemia is a well-characterized model for inducing a systemic innate immune response and, consequently, a neutrophil left shift (4). In homeostasis, the homogenous population of circulatory neutrophils displays nuclear segmentation, high CD16 (FcγRIII) expression, and high CD62L (l-selectin) expression (5). Upon LPS administration, three different neutrophil subsets can be distinguished in the circulation based on their CD16 and CD62L expression (5, 6). When sorting these subsets by flow cytometry, the subset of CD16^low neutrophils is enriched for banded neutrophils. The CD16^highCD62L^high neutrophils have segmented nuclei equal to homeostatic neutrophils. The third subset of CD16^highCD62L^low neutrophils is characterized by increased segmentation, or lobulation, of the nucleus (5, 6). The banded and hypersegmented subsets are not present in the circulation in healthy human controls (5).

Although the sorted subsets do not represent 100% pure populations, proteomics analysis of the sorted subsets clearly distinguished three separate populations (6). The differences in proteomic profile could not be explained by neutrophil activation, suggesting the CD62L^low neutrophils stem from a different lineage than the CD62L^high neutrophils (6). Furthermore, the sorted subsets have repeatedly been demonstrated to differ not only in phenotype (5–8) but also in function (3, 5, 7, 9). Notably, the hypersegmented CD62L^low neutrophil subset has a lower antibacterial capacity than normally segmented CD16^highCD62L^high neutrophils (3). The hypersegmented neutrophils phagocytose bacteria but fail to kill them intracellularly, even allowing bacterial proliferation inside the phagosome (3).

Because their antibacterial capacity differs considerably, the ratio between the different neutrophil subsets might dictate the outcome of the neutrophil response to infection. At a local site of infection in the tissue, the number of recruited neutrophils correlates to the rates of extravasation from the circulation to the tissue and subsequent chemotactic migration. The timing of the recruitment is important; if the banded subset only arrives at the infection site after the hypersegmented subset has phagocytosed the pathogens, the hypersegmented neutrophils may act as a physical barrier, shielding off the bacteria from the superior killers.

There is indeed reason to presume that banded neutrophils might migrate slower than their segmented counterparts because of their relatively bulky nucleus. The diameter of the nucleus has been described as a rate-limiting factor in cell migration (10). For T cells, the nucleus causes >99% of cells to be arrested at a pore cross section below 5–6 μm², whereas for mature neutrophils this degree of arrest was only reached at a pore cross section of 1–2 μm². Unfolding of the neutrophil’s segmented nucleus was thought to be responsible for this lowered limit (10). When migrating neutrophils were imaged in confined collagen matrices, the segments of the neutrophil nucleus were seen to rearrange into a “pearl-on-a-string” configuration, effectively decreasing the diameter of their nucleus (10).

In vivo, neutrophils encounter narrow pores, for example, between or within endothelial cells, in the basement membrane, or in the interstitium (11–13). Over the years, many authors have proposed that the unique nuclear segmentation of the neutrophil has evolved to enable efficient migration through these narrow pores (14–16).

Based on the above information, we hypothesized that the nucleus of CD16^low neutrophils, characterized by incomplete segmentation, would have a larger diameter than the nucleus of mature CD16^highCD62L^high neutrophils, characterized by complete segmentation. Consequently, CD16^low neutrophils would migrate less efficiently than CD16^highCD62L^high neutrophils. In contrast, CD62L^low neutrophils, characterized by hypersegmentation of the nucleus, would migrate more efficiently than segmented CD16^highCD62L^high neutrophils. These hypotheses cannot be tested in vivo using animal models because the nuclear morphology of neutrophils differs per species (17, 18). Additionally, the increased antibacterial capacity of banded neutrophils has only been shown in humans (3). Therefore, we induced acute inflammation in the human endotoxemia model and isolated the human CD16/CD62L subsets for in vitro four-dimensional migration experiments.

To study migration patterns of neutrophils in vitro, we used a Transwell system with 3-μm-diameter pores covered by a monolayer of endothelium and dense collagen matrices similar to those used by Wolf et al. (10). In vivo, the space between collagen fibers is reported to range from 2 to 30 μm (13, 19, 20). In vitro, we adapted the collagen concentration of the matrices to obtain pore sizes on the lower end of the physiological range because rearrangement of the neutrophil nucleus would likely be necessary in such confined conditions.

Concurrently, these experiments allowed us to study the reverse hypothesis: can migration through narrow pores cause nuclear (hyper)segmentation? Because the hypersegmented neutrophils recruited upon LPS challenge are thought to originate from a different lineage, they could represent a pool of tissue neutrophils that moves to the circulation by reverse transmigration. In our experiments, we asked whether a history of migration can explain the difference in nuclear segmentation between the CD62L^high and CD62L^low subsets.

Blood samples were obtained from healthy male volunteers between the age of 18 and 35 y participating in human endotoxemia trials (NCT02629874, NCT02675868, NCT02922673, ABR NL61136.091.17). The studies were approved by the ethics review board of the Radboud University Medical Center in Nijmegen, the Netherlands, and written informed consent was obtained from all study participants. Subjects were healthy as determined by physical examination, electrocardiography, and hematological laboratory values. Subjects with febrile illness in the 2 wk before the study or taking prescription drugs were excluded from the study.

For the transmigration experiments, samples were obtained from the endotoxemia trial NCT02675868, “Effect of Vasopressors on Immune Response.” The morphology or marker expression of the neutrophil subsets as tested in microscopy and flow cytometry was not affected by the vasopressors tested compared with the placebo group. Nonlinear regression modeling indicated that the vasopressors tested did also not affect transmigration kinetics compared with placebo. Therefore, the results of all subjects were pooled.

The LPS challenge was performed as published previously (21, 22). In short, subjects were admitted to the research medium care unit of the Radboud University Medical Center and were infused with 1.5 l hydration fluid during 1 h (2.5% glucose/0.45% saline at a continuous rate). Subsequently, in the “bolus model” the subjects received at time t = 0 a single dose of 2 ng/kg bodyweight LPS (United States standard reference Escherichia coli O:113, National Institutes of Health Pharmaceutical Development Section, Bethesda, MD) and were then infused with hydration fluid at a constant rate of 150 ml/h. In the “continuous model,” the subjects received a single dose of 1 ng/kg at t = 0, followed by 3 h of continuous LPS infusion at 1 ng/kg per hour. During the challenge, heart rate, blood pressure, and the course of LPS-induced symptoms such as fever, muscle aches, and nausea were constantly monitored.

Blood samples were also obtained from anonymous, healthy volunteers between the age of 18 and 65 y, male and female, who gave informed consent under protocols approved by the Medical Ethical Committee of the University Medical Center Utrecht.

Human blood samples were collected 3 h after the start of LPS administration using sodium heparin as an anticoagulant. The erythrocytes were lysed using ice-cold lysis buffer with pH 7.4 (150 mM NH₄Cl, 10 mM KHCO₃, and 0.1 mM Na₂EDTA dissolved in double distilled H₂O), and the remaining leukocytes were washed twice and resuspended in FACS staining buffer (4 mg/ml human albumin and 0.32% [w/v] sodium citrate in PBS). Leukocytes were stained with Abs against CD14, CD16, and CD62L, and neutrophils (SSC^highCD14⁻) were sorted using a FACSAria (BD Biosciences, Franklin Lakes, NJ) into tubes containing FACS staining buffer supplemented with 20% FCS.

For healthy controls, neutrophils were isolated by density gradient centrifugation over a single layer of Ficoll-Paque Plus and the erythrocytes in the neutrophil layer were lysed using the lysis buffer described above.

Neutrophil nuclear morphology was determined by manual counting of May–Grünwald–Giemsa–stained cytospins in a blinded manner. An Axioskop 40 microscope (ZEISS, Jena, Germany) was used with a 100× oil immersion objective. A separation between nuclear lobes was defined as the connection between lobes being less than one third of the width of the adjacent lobes. Rare progenitors, recognized by nuclear morphology and blue cytoplasm, were counted as having one nuclear lobe.

HUVECs were a kind gift from Willem-Jan Pannekoek. FluoroBlok Transwell inserts for 24-well plates with 3-μm pores (Corning Life Sciences, Tewskbury, MA) and flat-bottom 96-well plates were coated with fibronectin and seeded with 35,000 HUVEC cells per insert or well. HUVEC cells were always cultured in endothelial growth medium (EGM-2; Lonza, Basel, Switzerland) supplemented with BulletKit (Lonza). After 24 h, formation of a confluent monolayer was confirmed by visual inspection under the microscope and by transendothelial electrical resistance testing with an Epithelial Volt/Ohm Meter (23). One electrode of the Epithelial Volt/Ohm Meter was placed in the bottom compartment and the other electrode in the top compartment of the Transwell system. HUVECs were activated with 100 U/ml TNF-α for 4 to 7 h prior to adding neutrophils, as described before (9).

After FACS sorting, neutrophils were stained with calcein AM (1 μM; Thermo Fisher Scientific, Waltham, MA), washed, resuspended in EGM-2 medium, and 2 × 10⁵ neutrophils were added per insert per well. To prevent differences in calcein staining biasing the comparison of subsets, the percentage of transmigration was always calculated based on a control well, for which the cells were placed directly into a bottom well of the same plate. To the lower compartments of the 24-well receiver plate, EGM-2 medium containing 1 × 10⁻⁷ M of fMLF was added, and the plate was immediately placed in the FLUOstar Omega plate reader (BMG LABTECH, Ortenberg, Germany). Fluorescence intensity in the lower compartment was measured every 5 min at eight positions per well and averaged. The excitation filter was 485/10 nm, the emission filter was 520/10 nm, and the incubation temperature was 37°C. Simultaneously, neutrophils in the 96-well control plate were incubated in EGM-2 medium containing 1 × 10⁻⁷ M of fMLF (37°C, 5% CO₂, humidified atmosphere). For selected experiments, after 1 h, neutrophils from the inserts, lower compartment, and control wells were washed and stained for flow cytometry analysis.

Fluorescence readings over time were analyzed by nonlinear regression modeling with GraphPad Prism 7 (GraphPad Software, San Diego, CA).

Abs used for cell staining were anti-CD16-Alexa Fluor 647 (3G8; BD Biosciences), anti-CD16 PE-Cy7 (3G8; BioLegend, San Diego, CA), anti-CD16 V500 (3G8; BD Biosciences), anti-CD62L-FITC (DREG-56; BD Biosciences or BioLegend), anti-CD62L-PE (SK11; BD Biosciences), anti-CD62L PE-Cy5 (DREG-56; BD Biosciences), anti-CD62L PE-Cy7 (DREG-56; BioLegend), anti-CD11b AF700 (ICRF44; BD Biosciences), anti-CD35 unlabeled (rabbit polyclonal; Santa Cruz Biotechnology, Dallas, TX) and/or anti-rabbit IgG AF568 (goat polyclonal; Thermo Fisher Scientific). Samples were analyzed on a LSRFortessa flow cytometer (BD Biosciences).

Abs used for cell sorting were anti-CD16-AF647 (3G8; BD Biosciences) or anti-CD16-BV785 (3G8; Sony Biotechnology, San Jose, CA), and anti-CD62L-PE (SK11; BD Biosciences) or anti-CD62L-FITC (DREG-56; BD Biosciences) or anti-CD62L-PE-Cy7 (DREG-56; BioLegend), and anti-CD14-allophycocyanin-H7 (MφP9; BD Biosciences).

Migration chambers and fibrillar collagen matrices were made as described before (10). In brief, the pH of rat tail collagen I, high concentration (Corning Life Sciences) was raised to pH 7.4 using NaOH and buffered by 25 μM of HEPES and 10× DMEM (Biochrom, Cambourne, U.K.). Neutrophils were suspended in the neutralized collagen solution, resulting in a final collagen concentration of 3.3 or 8.0 mg/ml, and the solution was transferred into a self-constructed chamber that was immediately placed at 37°C for collagen polymerization. For chemokinesis, fMLF was added to the collagen gel to reach a final concentration of 1 × 10⁻⁷ M. After 30 min, DMEM with 1 × 10⁻⁷ M fMLF was added on one side and the chamber was closed.

Neutrophil migration in matrices was recorded with a 5× air objective using a Leitz DM RXE microscope (Leica Microsystems, Wetzlar, Germany) and a custom-made macro in the QWin V3 software (Leica Microsystems). For each time lapse video, 100 images were taken at 30 s intervals. The jittering generated by the microscope during image acquisition was removed as described previously (24). Migration tracks were generated with custom-made scripts (https://github.com/jtextor/tracking) and were analyzed using MotilityLab (http://motilitylab.net) to quantify speed and mean autocorrelation. Mean autocorrelation, a persistence parameter not biased by migration speed, was calculated [as described in (25)]. In short, for each possible pair of displacement vectors in a track, the angles were compared over different time intervals (Δt) to obtain the angle difference θ. The autocorrelation coefficient is the cosine of the angle difference cos(θ). For all tracks in one matrix, the autocorrelation coefficients were averaged for all Δt to get a curve of the mean autocorrelation over time. To be able to compare donors, the persistence time was deduced with nonlinear regression modeling in GraphPad Prism 7 (i.e., the time after which the mean autocorrelation was at 0.5). Cells that did not move during the entire time lapse were excluded from analysis, although the percentage of nonmoving cells did not differ substantially between subsets (data not shown).

Collagen matrices were prepared as described above, but neutrophils were stained with Hoechst 33342 nuclear staining (20 μM; Life Technologies, Waltham, MA) and calcein AM (0.5 μM) before embedding in collagen matrices. Matrices were imaged with an AF7000 LX widefield fluorescence microscope (Leica Microsystems) using a 40× air objective at 37°C and 5% CO₂. For each time lapse video, 11 to 60 images were taken at 30 s intervals. For cells and nuclei, the circularity was determined with Fiji (26), and girth was determined with custom-made scripts (available upon request). A higher circularity value is closer to a perfect circle, which has a circularity value of one.

From the same collagen solutions, drops of ∼60 μl were pipetted on coverglasses and were allowed to polymerize at 37°C for 40 min. Next, the collagen matrices were submerged in DMEM with 1 × 10⁻⁷ M fMLF for 60 min, washed twice with PBS, and fixed in 4% paraformaldehyde in 1× PBS for 30 min, all at 37°C. Matrices were imaged with a LSM 710 confocal microscope (ZEISS), using a 20× air objective in z-stacks at a 1.5-μm slice interval and of 35–90 μm z-depth. The number of cells in the z-stack was determined as the number of Hoechst-stained nuclei, and the number of vesicles was determined as the number of calcein-stained particles of 0.2–10 μm².

Collagen matrices without cells or fixatives were imaged by confocal reflection microscopy as described previously (10). In short, z-stacks (1-μm slice interval, 30-μm total z-depth) starting at a minimal distance of 10 μm from the cover glass were generated using a 63× oil objective. After orthogonal reconstruction of xz images with a y-interval of 1 μm, pore cross sections were measured as areas free of reflection signal from collagen fibers. In Fiji, classic watershedding was applied to the binarized images before particle analysis with thresholds ranging 1 μm²–infinity was applied. Brightness and contrast were only adjusted for ease of viewing (Fig. 5), not for images to be analyzed.

Statistical tests, as indicated in the figure legends, were performed in GraphPad Prism 7. The Wilcoxon matched-pairs test was used to compare lobularity before and after transmigration. When comparing donor medians, parametric tests were used; repeated-measurements ANOVA for paired or ANOVA for unpaired data were used when applicable. When n was not high enough to test for nonnormality, nonparametric tests were used; Friedman test for paired or Kruskal–Wallis test for unpaired data were used when applicable. Because the number of groups was low (n = 3), the post hoc tests were not corrected for multiple comparisons. Results were regarded as significant when p < 0.05.

To isolate different neutrophil populations, blood samples were obtained from the study subjects 3 h after the start of i.v. administration of LPS. After lysis of RBCs and neutrophil staining, flow cytometry distinguished three neutrophil subsets that could be sorted based on CD16 and CD62L staining (Fig. 1A). Subsequent counting of nuclear segments on cytospin slides showed that the sorted neutrophil subsets were enriched for banded, segmented, or hypersegmented neutrophils (Fig. 1B, 1C) as has been described before (5, 6). As expected, the CD16^highCD62L^high neutrophils isolated from experimental endotoxemia subjects showed similar nuclear segmentation as neutrophils from untreated healthy donors (Supplemental Fig. 1, Fig. 2F). Although the subsets showed some overlap, for ease of reading, sorted CD16^lowCD62L^high neutrophils will be referred to as banded neutrophils in the text of this paper, CD16^highCD62L^high will be referred to as segmented neutrophils, and CD16^highCD62L^low will be referred to as hypersegmented neutrophils.

To study how the nuclear segmentation of the neutrophil subsets affects extravasation, each subset was followed during in vitro transmigration through an endothelial cell layer and a membrane with 3-μm pores in a Transwell system. The neutrophils were fluorescently stained with calcein and placed in the top compartment of the Transwell system. As the used FluoroBlok Transwell inserts have a light-blocking membrane, only the neutrophils in the lower compartment were detected by a fluorescence plate reader. In this manner, the kinetics of the transmigrating neutrophils were recorded in real time.

All three subsets showed rapid migration toward the chemoattractant fMLF within the first 15–25 min (Fig. 2A). The rate of transmigration was represented by the slope of each curve, derived from nonlinear regression models (R² between 0.952 and 0.999) and did not differ between neutrophil subsets (Fig. 2B). After ∼50 min, almost all transmigration curves reached a plateau, represented by the top value of each curve in the regression models. The hypersegmented subset reached a higher plateau of transmigrated neutrophils than the other subsets (Fig. 2C). The percentage of transmigrated neutrophils at the start of the measurement (i.e., the start of the curve) was also higher for the hypersegmented subset (Fig. 2D). Together, although neutrophil subsets transmigrated with the same rate, the net transmigration efficiency was not the same.

To investigate the effect of transmigration on neutrophil phenotype, transmigrated neutrophils were retrieved from the lower compartments of the Transwell system after 1 h. Cytospin slides of the neutrophil subsets were generated to quantify nuclear segmentation under the microscope. For the segmented and hypersegmented subsets, transmigrated neutrophils showed similar nuclear segmentation as directly after sorting, indicating that transendothelial migration neither induced nor selected for neutrophils with a more segmented nucleus (Fig. 2E, 2F). Interestingly, for the banded subset, the nuclear segmentation increased in the neutrophils from the lower compartments compared with after sorting (Fig. 2E, 2F). However, this increase was also seen for neutrophils from control wells (Fig. 2G), indicating that incubation with fMLF alone was sufficient to increase the nuclear segmentation of the banded neutrophils. Possibly, the banded neutrophils were maturing in vitro. In conclusion, for banded neutrophils too, transendothelial migration neither induced nor selected for a more segmented nucleus.

The transmigrated neutrophils displayed increased CD11b and CD35 expression and decreased CD62L expression compared with the control neutrophils (Fig. 2H), suggesting that transendothelial migration activated neutrophils more than fMLF incubation alone. The three subsets showed similar changes in CD11b and CD35 expression, but only banded neutrophils showed significant transmigration-induced CD62L downregulation, probably corresponding to a higher CD62L expression before transmigration.

After extravasation, the next step in neutrophil recruitment is migration through the extracellular matrix, for which we used three-dimensional fibrillar collagen matrices as a model. Chemokinesis or chemotaxis was induced in the collagen matrices by fMLF, and migration speed was determined for a large number of neutrophils per matrix (Fig. 3A, 3B). Persistence of direction was evaluated using the autocorrelation within the direction of displacement (25), quantifying the straightness of the migratory track unbiased by migration speed. For each segment of a migration track, the autocorrelation coefficient is determined (Fig. 4A). If a cell takes random turns during migration, the autocorrelation coefficient will quickly decrease to zero. For a population of cells, the autocorrelation coefficients are pooled to obtain a mean autocorrelation curve (example in Fig. 4B). From such curves, the persistence time was deduced, the time after which the mean autocorrelation reaches 0.5.

The three neutrophil subsets were each embedded in separate collagen matrices with a concentration of 3.3 mg/ml. The directed movement of chemotaxis displayed higher speed (Fig. 3D) and persistence (Fig. 4E) than the random movement of chemokinesis (Figs. 3C, 4D). To our surprise, no significant speed or persistence differences were observed between the three neutrophil subsets in either chemokinetic or chemotactic migration in 3.3 mg/ml matrices (Table I). Neither were differences in speed and persistence observed between the circulatory neutrophils from experimental endotoxemia subjects and circulatory neutrophils from healthy, untreated donors (Table I).

However, the calculated chemotaxis speeds close to 10 μm/min (Fig. 3D) were far higher than Wolf et al. (10) had reported, indicating that we did not yet reach the physical limits of migration. Additional imaging revealed the pores in our 3.3 mg/ml matrices had a cross section around 8 μm² (Supplemental Fig. 2), in contrast to the expected 2–3 μm² based on the work by Wolf et al. (10). This was a result of using rat tail collagen from different manufacturers (K. Wolf, unpublished observations).

Possibly, more stringent conditions could uncover an effect of nuclear segmentation on migration speed. Hence, we repeated the migration experiments in matrices of 8.0 mg/ml collagen, resulting in a pore size around 5 μm² (Supplemental Fig. 2). During fMLF-induced chemotaxis in the 8.0 mg/ml matrices, migration tracks were shorter (Fig. 3A), and migrational speed (Fig. 3E) and persistence (Fig. 4F) were decreased compared with the 3.3 mg/ml matrices. Interestingly, in the 8.0 mg/ml matrices, the migration speed significantly differed between the banded and hypersegmented subsets (Fig. 3E, Table I), although there was no difference in persistence (Fig. 4F). In contrast to our hypothesis, hypersegmented neutrophils migrated significantly slower than banded neutrophils (Fig. 3E).

To find an explanation for the speed difference observed during chemotaxis in 8.0 mg/ml matrices, for some donors, the migration of the subsets was also imaged with fluorescence microscopy at a higher magnification (Fig. 5A, Supplemental Video 1–3). Notably, the hypersegmented neutrophils more often assumed an elongated shape when trying to squeeze through narrow holes between the collagen fibers (Fig. 5A, Supplemental Video 3). In parallel, the nuclei of those hypersegmented neutrophils assumed an elongated, pearls-on-a-string configuration (Fig. 5A, Supplemental Video 3). In contrast, the nuclei observed in the banded and segmented subsets unfolded less, retaining a more compact shape (Supplemental Video 1, 2).

This behavior was quantified by analyzing the circularity and girth of the nucleus, based on the Hoechst signal. The girth is defined by us as the diameter of the largest circle that fits inside the nucleus (Fig. 5A) and, therefore, corresponds to the minimal pore diameter this nucleus would be able to traverse at that time. In accordance with our visual observations, the circularity and girth of the hypersegmented nuclei were decreased compared with the banded nuclei (Fig. 5B, 5C). The circularity and girth of the cell body were determined based on the calcein signal and followed the same trend as the nuclei. Although the difference in cellular circularity between banded and hypersegmented neutrophils was not statistically significant (Fig. 5D, 5E), the difference between banded and segmented neutrophils was statistically significant, confirming that banded neutrophils elongated less than (hyper)segmented neutrophils.

So why did the reduced girth of the hypersegmented neutrophils not correlate to higher migration speeds as expected beforehand? The migration videos suggested that the hypersegmented subset suffered from a reduced rear release. In some instances, the front of the cell kept moving forward over multiple frames, while the uropod was not yet released, until both parts were only connected by a very thin stretch of membrane (Supplemental Video 4). Sometimes, one nuclear segment was present in the uropod, suggesting the nuclear membrane was also very stretched out (Supplemental Video 4). Occasionally, we observed how the hampered rear release even led to pieces of membrane breaking off from the migrating cell body (27, 28). To quantify this phenomenon, collagen matrices in which calcein-stained neutrophils had been migrating for 1 h were fixed and analyzed for the presence of calcein-positive vesicles. Interestingly, the matrices with hypersegmented neutrophils contained more vesicles than the other matrices (Fig. 5F). Although this approach did not distinguish between active and passive release of vesicles, the link between reduced rear release and increased vesicle release would be interesting to investigate in follow-up studies.

The human experimental endotoxemia model provided us with a unique opportunity to isolate human neutrophils with banded, segmented, as well as hypersegmented nuclei from a single blood sample. The transendothelial migration assay and the three-dimensional collagen matrices simulated common barriers during neutrophil recruitment (13, 19, 20). The presented results suggest that a general increase in nuclear segmentation (Fig. 1B) did not increase migration of activated neutrophil populations through narrow pores between endothelial cells and collagen fibers. In both the transmigration and three-dimensional migration experiments, the migration efficiency of banded neutrophils did not differ from the migration efficiency of normally segmented neutrophils. The hypothesis that incomplete nuclear segmentation would preclude a lowering of the nuclear diameter in restrictive conditions was contradicted by the results of the girth analysis (Fig. 5C). Banded and normally segmented neutrophils had similar nuclear diameters during migration in restrictive collagen matrices. Restrictive pore sizes may induce similar deformation of the nucleus in both subsets because the circularity of banded and normally segmented nuclei did not prove significantly different (Fig. 5B). Similar nuclear diameter and deformation can explain the similar migration efficiencies of banded and normally segmented neutrophils observed in all experiments (Figs. 2–4).

Alternatively, nuclear segmentation does affect neutrophil migration, but this effect was masked in our experimental setup. First, the three subsets isolated from the experimental endotoxemia model do not differ only in nuclear segmentation. Other differences between the subsets may have affected neutrophil migration, counterbalancing the effect of nuclear segmentation. Second, the effect may only show in even more restrictive collagen matrices than were used in this study. Third, the CD16^low and CD62L^low populations did not correspond to 100% pure populations of neutrophils with a banded or hypersegmented nucleus. We cannot exclude the possibility that more restrictive conditions and/or 100% pure populations would have yielded different results. However, we believe the sorted CD16^low and CD62L^low populations do correspond to functional subsets. As both the endotoxemia model and the gating strategy for sorting the CD16/CD62L subsets were the same as in previous studies (3, 5–7), the results of our migration analyses can be correlated to the functional subsets previously described. Hence, we conclude that the neutrophils with the highest antibacterial capacity, the CD16^low neutrophils (3), were not hampered by inefficient migration in vitro. Whether the in vitro migrational behavior of the neutrophil subsets isolated during experimental endotoxemia can be extrapolated to in vivo recruitment and to other conditions of acute inflammation remains to be confirmed.

Although the suggestion that neutrophils with a band-shaped nucleus pass through narrow pores as easily as neutrophils with a segmented nucleus may be counterintuitive, it does agree with the clinical phenotype of patients with the Pelger–Huët anomaly. In these patients, segmentation of the neutrophil nucleus does not occur during differentiation because of mutations in the lamin B receptor. The neutrophils of homozygous individuals have a completely round nucleus, whereas the neutrophils of heterozygous individuals mostly have bilobed nuclei (29). Despite all their neutrophils being hyposegmented, Pelger–Huët anomaly patients do not present with an increased rate of infections (30, 31). Although early studies found reduced chemotaxis of Pelger–Huët anomaly neutrophils in skin windows in vivo and through filter papers with 3- or 5-μm diameter pores in vitro (15, 32), later studies did not reproduce this defect. Chemotaxis in under-agarose assays, through filters with 2- or 3-μm pores, and in skin windows has been described as normal for human and canine Pelger–Huët anomaly neutrophils (33–35). More recently, the hyposegmented morphology was reproduced in the HL-60 neutrophil-like cell line by targeted mutation of the lamin B receptor. The mutated HL-60 cells showed normal migration through 5-μm diameter pores in a microfluidic device and through 3- or 8-μm diameter pores in a Transwell system (36). Together with the absence of clinical symptoms in Pelger–Huët anomaly patients, this supports the idea that nuclear segmentation is not required for efficient neutrophil migration.

Our results are also supported by a recent paper by Barzilai et al. (37) showing that neutrophils and T cells from peripheral blood take the same time to completely cross an endothelial monolayer. While migrating through endothelial gaps but also while migrating on flat surfaces, both cell types surprisingly protrude nuclear lobes into the leading pseudopod. Whereas neutrophils used the existing nuclear lobes, T cells quickly generated them de novo (37). In addition, the pores generated in the endothelial layer during either T cell or neutrophil transmigration were of equal size. These results suggest that independent of the shape, the leukocyte nucleus will always be compressed to obtain a certain diameter compatible with transendothelial migration.

This degree of compression may, however, not be sufficient when pore size decreases further (e.g., in dense extracellular matrix) because at the physical limits of migration differences between neutrophils and T cells were reported (10). Neutrophils could migrate through smaller pores than T cells, and this difference was attributed to nuclear diameter (10). However, as the authors also suggested, nuclear diameter may not only be dictated by shape but also by size and rigidity of the nucleus (10). These factors probably differ between T cells and neutrophils. Because the current paper shows that the role of nuclear shape in migration through very narrow pores seems negligible, size and rigidity of the nucleus may be more important factors in determining nuclear diameter during migration. Because neutrophils do not express lamin A/C proteins (10), the factors determining nuclear rigidity are currently unknown.

In contrast to our initial hypothesis, the hypersegmented neutrophils migrated slower than banded neutrophils in 8.0 mg/ml collagen matrices. Clearly, nuclear shape is not the sole determinant of migrational behavior, and other mechanisms also differed between the neutrophil subsets. The decreased migration speed of the hypersegmented neutrophils coincided with a decreased circularity of the cell (Fig. 5D). Extreme uropod elongation was observed in part of the hypersegmented population (Supplemental Video 4). These results suggested the reduced migration speed may result from a delayed release of the rear of the cell. Significant elongation of murine leukocytes has been correlated with delayed extravasation in vivo (38). Notably, uropod elongation during extravasation also coincided with the release of microparticles, vesicles originating from the plasma membrane (38). Data from our collagen matrices suggested that the increased elongation of the hypersegmented neutrophils may be correlated to a higher release of such membrane vesicles (Fig. 5F). Notably, a cluster of proteins that showed increasing expression levels going from banded to segmented to hypersegmented neutrophils was significantly enriched for proteins annotated to membrane vesicle, extracellular vesicle, and exosome components (6). Vesicles released by human or murine neutrophils can act as messengers to other neutrophils or other cells (39, 40). Therefore, it would be interesting to examine the content of the observed vesicles in follow-up studies.

In the Transwell experiments, the transmigration speed of the three subsets did not differ, but the hypersegmented neutrophils reached a higher total percentage of transmigration. This observation could not be explained by the timing of pipetting because the banded subset was always added to the Transwell insert first and the hypersegmented subset was always added last. As the transmigration speed was very high (Fig. 2A, 2B), the neutrophils had possibly already started transmigrating in the short time period between placing the cells into the plate and starting the plate reader measurement. The hill slope of the hypersegmented subset (Fig. 2B) might have been underestimated if the initial curve segment missing from the analysis was relatively steep. This may explain why the hill slope does not show a significant difference between the neutrophil subsets, whereas the starting and plateau percentages of transmigration do.

Even if the Transwell experiments suggested that hypersegmented neutrophils have a slight migratory advantage, the migration experiments in very dense collagen matrices suggested that hypersegmented neutrophils have a slight migratory disadvantage. This discrepancy may result from the way migration is classified in the two types of experiments. In the transmigration experiments, a neutrophil is considered to be transmigrated when the fluorescent cell staining is located below the FluoroBlok insert membrane. However, the suggested delay in rear release could also play a role in the transmigration experiments. When a hypersegmented neutrophil would linger at the underside of the insert membrane with its rear still trapped in the narrow pore, the majority of its cell body, and thus the fluorescence, would be classified as transmigrated. However, in our collagen matrices, this same neutrophil would be slow in pursuing its further migration path. The idea that hypersegmented neutrophils got trapped during crossing of the insert membrane was supported by the fact that removal of the Transwell insert also removed part of the fluorescent signal from the well (Supplemental Fig. 3). The decrease in fluorescence was much larger for the hypersegmented neutrophils than for the other subsets, suggesting that more neutrophils were trapped in the insert after 60 min of transmigration. Thus, a higher transmigration rate of hypersegmented neutrophils may not translate to a migratory advantage for hypersegmented neutrophils during neutrophil recruitment, similar to elongated leukocytes in vivo (38).

Interestingly, transendothelial migration specifically induced upregulation of CD11b and CD35 expression and downregulation of CD62L expression (Fig. 2H). Directly after isolation from the blood, those same markers had clearly different expression levels on hypersegmented versus normally segmented neutrophils (5). Because the phenotype of hypersegmented neutrophils resembles the phenotype of transmigrated segmented neutrophils, in vivo the phenotype of hypersegmented neutrophils possibly diverges during or after (reverse) transendothelial migration. In contrast, the hypersegmentation of the nucleus found in CD62L^low neutrophils does not seem to result from a migration history. Nuclear segmentation was not induced by neutrophil migration through narrow pores because it was not altered after transendothelial migration (Fig. 2E) or after 2 h of migration in collagen matrices (data not shown).

In conclusion, the functional neutrophil subset recognized by low CD16 expression did not show decreased migration efficiency in in vitro models of transendothelial and interstitial migration. The migration speed but not persistence of the CD62L^low neutrophil subset was only slightly reduced in very dense collagen matrices but not in less dense matrices or during transendothelial migration. It seems that nuclear segmentation or lack thereof does not significantly contribute to neutrophil migrational efficiency, although it could not be excluded that other differences between the subsets counterbalanced an effect. Interestingly, CD16^low neutrophils, considered as an immature form, seem just as capable of fast migration through narrow pores as CD16^high mature neutrophils. Together with their superior antibacterial capacity (3), this argues against considering CD16^low cells as functionally immature cells. At the banded stage, neutrophils already contain all three types of granules (41) and can produce reactive oxygen species (7). It is tempting to speculate that CD16^low neutrophils are a functional subset that is only to be released in certain emergency conditions, and we aim to pursue this hypothesis in future research.

We thank Roger van Groenendael and Guus Leijte for recruiting the volunteers and performing the LPS challenges, Na Chen and Leo Houben for help with performing experiments, Jeroen van Velzen and Pien van der Burght for help with the FACS machines, the Hubrecht Imaging Center and Laboratory of Translational Immunology Imaging Core for help with microscopy, Willem-Jan Pannekoek for the gift of HUVECs, and Johan de Rooij for the gift of rat tail collagen.

The authors have no financial conflicts of interest.