On the origin of low‐density neutrophils

Abstract Here we elaborate on the origin of low(er)‐density neutrophils (LDNs) to better understand the variation found in literature. Supplemented with original data, we test the hypothesis that buoyant density of neutrophils is characterized by a spectrum that as a whole shifts to a lower density after activation. Both the 20% highest density (HDNs) and 20% lowest density (LDNs) neutrophils from healthy donors were isolated by Percoll of different densities. Using this method we found that LDNs were significantly better in T‐cell suppression and bacterial containment than their 20% highest density counterparts. We found no statistically relevant differences in neutrophil survival or bacterial phagocytosis. Stimulation of healthy donor neutrophils with N‐formyl‐methionyl‐leucyl‐phenylalanine induced LDNs co‐segregating with peripheral blood mononuclear cells after Ficoll separation. These in vitro induced LDNs showed increased activation markers compared to HDNs and were comparable to the activation markers found on the LDN fraction seen in patients with chronic inflammatory conditions such as present in cancer patients. This all fits with the hypothesis that the density of neutrophils is distributed in a spectrum partially coupled to maturation. Additionally a shift in this spectrum can be induced by in vitro stimulation or by activation in disease.

existing data, found in the literature, with new original experimental data in order to better understand the origin and functional properties of LDNs in health and disease.

DENSITY IS A SPECTRUM IN NEUTROPHILS
In 1983, Pember et al. showed that heterogeneity of density with a bellshaped distribution exists within the healthy human circulating neutrophil pool. 4 In addition, in this study density and mean cell volume were found to be connected. A similar distribution of different densities was also found in elicited and nonelicited murine neutrophils, and in neutrophils in allergic human subjects. 5,6 More recently, Herteman et al. studied LDNs in asthmatic and healthy horses and found that, although their numbers varied between health and disease, the functional characteristics were similar regardless of health status. 7 So, LDNs in asthmatic and healthy horses both had an increased capacity to produce neutrophil extracellular traps, and were similarly more sensitive to activation by phorbol-12-myristate-13-acetate. 7 LDNs in both conditions had more N-formyl-methionyl-leucyl-phenylalanine (fMLF) receptors and were smaller in size, which according to the authors made them intrinsically different. LDNs were described in several studies in healthy volunteers as well. 8,9,10 Besides a distribution of buoyant densities, other studies have shown functional heterogeneity within the circulating neutrophil pool of healthy humans. 11,12,13 For instance, in the studies of Eggleton et al.
a similar Gaussian-shaped profile on the basis of different cell surface electrical charge was found. 12,13 The presence of heterogeneity in the density and functionality in healthy neutrophils prompted us to test the hypothesis that a spectrum of neutrophil densities associated with different functionalities are already present in the circulation of healthy humans.
Unstimulated neutrophils from healthy volunteers were centrifuged over different Percoll densities between 1.079 and 1.083 g/ml.
Similarly, as in the study by Pember et al., 4 the number of cells found in the upper ring increased when higher Percoll densities were used for centrifugation, but not all cells shifted up (Fig. 1A). Some donor variation was present but in all donors more cells stayed in the ring of Percoll with a high density (1.083 g/ml) compared to the Percoll with a low density (1.079 g/ml; Fig. 1A). This proved that a natural spectrum in densities was present in the circulating neutrophil pool, even in healthy individuals. Although neutrophil function is known to change with age, 14 we did not find a correlation between total numbers of LDNs or HDNs and the donors age (Supporting Information   Fig. S1). Next, we tested whether the 10-20% of the lowest density neutrophils (LDNs, typically found on top of Percoll gradient with a density of 1.081 g/ml) had different functional properties compared to the 10-20% of the highest density neutrophils (HDNs, typically found on the bottom of Percoll gradient with a density of 1.083 g/ml).
All neutrophils combined were used as a control because they also contain the cells with the intermediate densities.
First, a containment assay was used to compare LDN and HDN functionality, as described by us before. 15 This assay mimics the in vivo situation and tests several neutrophil functions: chemotaxis, phagocytosis, bacterial killing, and neutrophil survival. The delay in GFPlabeled bacterial outgrowth (lag time) serves as a measure for the antibacterial capacity of the different neutrophils. There was a clear difference in the containment capacity of Staphylococcus aureus of neutrophils in favor of the 20% lowest density neutrophils (Fig. 1B). The extent of the advantage of the LDNs compared to the HDNs varied between the experiments but was always in favor of the LDNs. The containment capacity of the total neutrophil fractions showed a lag time comparable to LDNs. This implies that HDNs are less capable of containing bacteria compared to the other neutrophils in the density spectrum. To rule out the overrepresentation of eosinophils in the 20% highest density fraction, the experiment was repeated after sorting out eosinophils by flow cytometry cell sorting, after which the same defect in bacterial containment capacity was seen (data not shown). Knowing that phagocytosis was comparable between the fractions (Supporting Information Fig. S2A-C), the enhanced containment of bacteria by LDN was suggested to be an effect of either bacterial killing or neutrophil survival. To differentiate between these possibilities a survival assay was performed, which did not show any difference between the different fractions (Supporting Information Fig. S2D). Therefore, it can be argued that the advantage seen in this bacterial containment assay is due to different capacity in bacterial killing.
Next, we tested putative differences in the capability to suppress proliferation of lymphocytes. It has been previously shown that different neutrophil phenotypes have the capability to suppress lymphocyte proliferation, 16,17 which has also been linked to LDNs found in a variety of acute and chronic conditions. 18,19,20,21 Indeed, LDNs isolated from healthy donors, showed a better capability to suppress lymphocyte proliferation after 4 d, especially when they are compared with the HDNs (Fig. 1C-E). Interestingly, the division index-which takes all cells into account-is clearly different (Fig. 1D), whereas the proliferation index ( Fig. 1E)-which only takes the dividing cells into account-is not. It is therefore possible that when lymphocytes have "escaped" the suppression, they do not seem to be sensitive for neutrophil-mediated suppression anymore. So, from the data mentioned earlier, it can be concluded that a property to suppress proliferation of lymphocytes is linked to the intrinsic buoyant density of neutrophils from healthy donors.

ACTIVATION OF NEUTROPHILS IN VITRO LEADS TO A SHIFT OF DENSITY
Both human and murine neutrophils change density when activated by endotoxin-activated serum or fMLF in vitro. 4 Interestingly, this change in density still resulted in a distribution of densities with a bell shape. 4 This finding suggests that activation of neutrophils leads to a change of density for all cells, but only neutrophils that are in the lower density of the Gaussian distribution prior to activation will Other proof that lower density granulocytes are an effect of activation was shown when neutrophils were incubated with Mycobacterium tuberculosis in vitro. 22 Because LDNs were also found in patients infected with Mycobacterium tuberculosis, it is very plausible that a direct interaction between neutrophils and the bacterium leads to a decrease in density. 22 Further complexity is introduced when taking into account that neutrophil density gradually decreases when the processing of neutrophils is delayed for more than 6 h. 23 This "spontaneous" density shift of neutrophils in vitro is likely to contribute to the wide variety of LDN numbers found between different experimental setups.
To reproduce that in vitro activation shifts the buoyant spectrum of freshly isolated neutrophils from healthy volunteers, we stimulated neutrophils with fMLF and platelet activating factor (PAF) in a . Data are depicted as median with IQR. Data of (A) and (C) were analyzed using Friedman test without correction for multiple comparison. Data of (B) was analyzed using a Mann-Whitney test markers (Fig. 2C-G). However, there was a remarkable difference in activation markers between LDNs and HDNs. LDNs showed a higher expression of CD66b, CD11b, CD35, and CD16 and a lower CD62L expression after 1 nM fMLF stimulation (Fig. 2C-G). This is in contrast with unstimulated neutrophils of different densities, which did not show a difference in activation markers between LDNs and HDNs (Supporting Information Fig. S3). A higher concentration of fMLF (0.1 M) also led to an increase of activation markers in both density fractions, but the differences between LDNs and HDNs were less pronounced (Supporting Information Fig. S4). This could mean that during homeostasis, certain neutrophils are more responsive to a lower concentration (1 nM) of fMLF with a subsequent shift in density of only these cells. However, because all neutrophils show a Ca 2+ influx with this low concentration of fMLF (Fig. 2H), another possible explanation could be that all cells shift in density but, as described in Section 2, only the cells on the lower end of the spectrum end up in the PBMC fraction.

LDNS IN DISEASE ARE HETEROGENEOUS
The appearance of neutrophils in the PBMC fraction after Ficoll-Paque centrifugation is described in several inflammatory diseases.
However, when different patient groups and different studies are compared, these cells are quite variable in both phenotype and functionality. 3,9,20,24,25 In the following paragraphs, the different subsets of LDNs are described, supplemented with our own data. A more extensive description of LDN in inflammatory diseases has been summarized recently in an excellent review by Scapini et al. 3

Are LDNs mainly immature neutrophils?
In the human bone marrow, a progressive increase of buoyant density was described for maturation of cells within the granulocyte lineage. 26,27 Banded and segmented neutrophils displayed a density above 1.080 g/ml, whereas promyelocytes and myelocytes exhibit a density below 1.080 g/ml. 27 Both acute and chronic inflammatory conditions often lead to an enhanced (emergency) granulopoiesis in the bone marrow 28,29 that will subsequently lead to the appearance of immature neutrophils in the circulation, often referred as a left shift. 30 Similarly, in a model of systemic inflammation, human experimental endotoxemia also leads to the appearance of banded neutrophils in the circulation. 31 Because neutrophil progenitors have a lower buoyant density, it is not surprising that in certain inflammatory conditions such as cancer or sepsis these immature neutrophils are found in the PBMC fraction. 19,32,33 To test whether density is linked to maturation of neutrophils in healthy volunteers, we used a multicolor flow cytometry analysis on unstimulated neutrophils after Percoll density centrifugation.
Although their numbers were small in healthy donors, a striking difference was found in the amount of CD16 dim /CD62L high banded neutrophils in the 20% lowest density neutrophils (median of 2.7% in LDNs vs. median of 0.24% in HDNs; Fig. 3A). 34 Next, we challenged healthy volunteers with LPS (2 ng/kg), as a model of controlled acute systemic inflammation as was described earlier. 35 After LPS more CD16 dim /CD62L high banded neutrophils were recruited to the blood. Significantly more neutrophils appeared on top of the lowest density gradient (1.077 g/ml) after LPS compared to the healthy volunteers that is depicted in Figure 1A (see also

Are LDNs mainly activated/degranulated neutrophils?
As was discussed in Section 3, mature neutrophils from healthy individuals decrease their density when activated in vitro, whereas mature neutrophils in inflammatory conditions can already be found in the PBMC fraction at steady state without additional in vitro activation.
These inflammatory LDNs often show signs of activation, especially when compared with neutrophils of healthy donors. 6,8,22,36 LDNs from patients with advanced adenocarcinoma in particular displayed higher expression of activation markers such as CD11b and CD66b compared to the normal density neutrophils from the same patients. 37 This difference in activation markers between LDNs and NDNs in advanced cancer is remarkably similar to the situation that is present after stimulation of healthy donor neutrophils with 1 nM fMLF (Fig. 2). It is therefore tempting to speculate that during (chronic) inflammation a subset of (mature) neutrophils is more susceptible to activation and that these mature activated neutrophils are the neutrophils found in the PBMC fraction in disease.
It is believed that this activation reflects degranulation in vivo. 3 However, it cannot be excluded that LDNs are not in vivo activated but rather primed cells that are more susceptible for in vitro manipulation.
Arguments for in vivo degranulation are provided by measuring the expression of certain granule markers on the surface of neutrophils, such as CD63 (azurophilic granules), CD66b (specific granules), gelatinase granules (CD11b), and secretory vesicles. 38,37 The extent of degranulation in vivo is poorly understood as the data of the actual content of the different granules and vesicles in neutrophils, for example, microscopy data, is scarce. Furthermore, existing electron microscopy data of the granular content of LDNs does not show signs of significant degranulation. 39 Another argument against degranulation as the cause for the change in buoyant density is the fact that the majority of the in vitro activated neutrophils revert to their original density after only a couple of hours (Supporting Information Fig. S5). 5 It is quite unlikely for neutrophils to regain granular content in this short time frame. An alternative mechanism could be a change in mean cell volume as a result of water uptake, regulated by aquaporin-9. 40 Changes to a higher forward scatter and a lower side scatter by flow cytometry, which is an argument often used to provide proof of degranulation, could also be caused by this phenomenon. 24,41 The activated phenotype of circulatory LDNs is perhaps acquired in the tissue. Although, this has not been directly shown, neutrophils in the tissue have a very similar phenotype compared to in vitro activated neutrophils. 42,43 This could mean that LDNs are cells that have reverse transmigrated from the endothelial membrane. Moreover, evidence has been provided that homing of neutrophils to the lung tissue is sufficient to adopt an activated phenotype, irrespective of inflammatory disease. 44

Are LDNs a distinct lineage of cells?
In SLE the presence of LDNs has been established by multiple studies. 25,45,39,46 These neutrophils are believed to have a contribution to the pathogenesis of SLE by an enhanced capability to produce type I IFNs and release neutrophil extracellular traps. 39,47 Moreover, it is suggested that these LDNs have an immature phenotype because of a premature release from the bone marrow, instigated by stimuli, which are produced in the course of the disease. 39 However, other studies have suggested that LDNs in SLE are a complete distinct subset of neutrophils, because they evidently differ from NDNs. 25,45,39 For example, in the study of Singh et al. an increased level of copy number alterations, losses of heterozygosity and microsatellite instability were found in LDNs and not in NDNs. 45 This made the authors suggest that LDNs are an abnormally developed lineage of neutrophils as a result of genomic damage. 45 Although, this is an interesting model, further research is needed to elucidate these findings.
Furthermore, it is still unknown if these findings are specific for SLE or that LDNs from other diseases share these characteristics in genetic variations.

CONCLUDING REMARKS
Here we found that circulating neutrophils in healthy individuals are heterogeneous regarding their buoyant density and that low density correlated with the increased responsiveness to fMLF, increased bacterial containment, and the capacity to suppress T-cell proliferation.
Although it is difficult to compare healthy neutrophils with neutrophils from patients with different inflammatory conditions, it is tempting to speculate that the cells present in the 20% lowest density fraction of our healthy donors are at least in part comparable with the immunomodulatory cells found in the PBMC fraction in inflammatory conditions such as SLE and cancer. Therefore, we believe that buoyant density is not a distinct feature in a subset of neutrophils found in disease, but rather a spectrum of different densities already found in healthy individuals. An inflammatory condition can subsequently lead to a decrease of the buoyant density of all neutrophils but only the cells on the lower end of the spectrum will shift to the PBMC fraction after Ficoll-Paque density centrifugation (Graphical Abstract).
Neutrophils with immunomodulatory capabilities are accompanied with an intrinsic lower buoyancy in both health and disease. One can only speculate what the underlying mechanism for the spectrum in different buoyant densities is during homeostasis. However, the shift of the spectrum during inflammation can be related to neutrophil maturation stage, activation status, and granulation. We do not exclude, however, that separate neutrophil lineages are being recruited and constitute part of the LDN fraction during inflammation. We encourage further studies to focus on mechanistic features that can explain a spectrum in buoyant densities in neutrophils rather than a subdivision of only two subsets (LDNs and NDNs).

Experimental endotoxemia model
The human experimental endotoxemia experiment was induced exactly as described before. 35

Neutrophil isolation
Human blood samples were collected in sodium heparin tubes (Vac-  Hereafter, the neutrophils were centrifuged over Ficoll-Paque as described earlier. The neutrophils that shifted to the "PBMC" layer were regarded as LDNs, the neutrophils that remained below were regarded as HDNs.

Containment assay
A neutrophil containment assay was used modified from Li et al. 48 and described by van Grinsven et al. 15 In short, a gel with fibrin scaffolds was formed to mimic a tissue-like environment. A total of 100 L of the fibrin gel was made by mixing 50 L of 5 × 10 5 neutrophils/ml in Hepes buffer with 50 L Hepes buffer containing 2 mg/ml fibrinogen (Sigma-Aldrich) with 40% human pooled serum (Sigma-Aldrich), 1 U/ml thrombin (Sigma-Aldrich), and 5 × 10 5 CFU GFP-expressing S.

Phagocytosis assay
Isolated LDNs and HDNs, and total neutrophils (as a control) were iso-  Fig. S6). For all experiments at least 10,000 events were recorded for data analysis.

Statistical analysis
Data was plotted and statistical analysis was preformed using Graph-