Inhibitors of neutrophil recruitment identified using transgenic zebrafish to screen a natural product library

Neutrophils constitute about 40–60% of circulating white blood cells in the human body and are the first line of cellular defence against foreign pathogens deployed by the innate immune system. Neutrophils migrate rapidly to sites of infection or injury in response to diverse chemo-attractants, including N-formyl-methionine-leucine-phenylalanine (fMLF), interleukin-8 and reactive oxygen species (Clark and Klebanoff, 1979; Ellett et al., 2011; Hattori et al., 2010; Lekstrom-Himes et al., 2005). As the insult is contained, the inflammatory response is resolved through reverse migration of neutrophils (Mathias et al., 2006) or by their apoptosis followed by macrophage engulfment (Bratton and Henson, 2011). Uncontrolled neutrophilic activity and continued recruitment of neutrophils to inflammatory sites can result in persistent inflammation, a feature of many human diseases, including chronic obstructive pulmonary disease (COPD) and cystic fibrosis (reviewed by Gernez et al., 2010). Inhaled bronchodilators and corticosteroids are prescribed for COPD, largely for symptom control, but these drugs are, for the most part, ineffective in controlling inflammation and halting disease progression (Morjaria et al., 2010). There is thus a pressing need for novel drugs that specifically target neutrophils.

The zebrafish has emerged as a powerful model not only for the study of vertebrate development but also for high-throughput drug discovery, facilitated by their small transparent embryos that can readily be arrayed into 96-well plates. The generation of transgenic zebrafish expressing neutrophil-specific GFP reporters has made possible the in vivo analysis of neutrophil behaviour (Renshaw et al., 2006; Mathias et al., 2006). Like their human counterparts, zebrafish neutrophils migrate towards the wound site in a PI3K-signalling-dependent manner (Yoo et al., 2010) and clear microbes by phagocytosis (Colucci-Guyon et al., 2011). The removal of neutrophils by reverse migration as well as by apoptosis and macrophage uptake has also been observed during inflammation resolution (Ellett et al., 2011; Loynes et al., 2010).

Here we describe the use of one such transgenic zebrafish neutrophil-specific reporter line in an in vivo screen of natural product extracts for inhibitors of neutrophil recruitment. We report the identification of two inhibitory compounds using this zebrafish model and their validation using an established in vitro mammalian neutrophil migration assay.

Our aim was to establish an assay for the rapid identification of highly effective inhibitors of neutrophil migration. To this end, we constructed an assay protocol based on rapid visual assessment of neutrophil recruitment. Targeted expression of GFP, using the myeloperoxidase (mpx) or lysozyme C (lyz) promoters (Renshaw et al., 2006; Hall et al., 2007), reveals the anterior yolk sac and posterior intermediate cell mass (ICM) of living zebrafish larvae to be the origin of neutrophils. In uninjured larvae, caudal neutrophils remain along the ventral side (Fig. 1A) but, following tailfin amputation, they rapidly migrate towards the wound site. By 3 hours after amputation, ~ten neutrophils are recruited to the distal portion of the remaining tailfin (Fig. 1B). To determine whether this assay could be used to screen for compounds that affect neutrophilic migration, rather than those that accelerate inflammation resolution (Ellett et al., 2011; Loynes et al., 2010), 2- to 3-day post-fertilisation (dpf) larvae were pre-incubated with the known PI3K inhibitor LY294002 and microtubule inhibitor nocodazole for 1, 3 or 5 hours prior to tailfin transection, and subsequently transferred into fresh drug-containing media for a further 3 hours before analysis of neutrophil recruitment to the wound site (see Materials and Methods). We found that both LY294002 (50 μM) and nocodazole (33 μM) were effective in inhibiting neutrophil migration (Fig. 1C,D,I) after 3 or 5 hours of pre-incubation. Because there were more toxic effects following 5 hours of pre-incubation, we used a 3-hour pre-incubation for the screening protocol.

Using the tailfin transection assay, we screened a library of 1040 crude fungal extracts from the MerLion collection, a diverse assortment of natural products obtained from a variety of sources. From this library, we identified 35 candidate samples that met our predetermined criteria and further validation of these 35 samples confirmed that two of them produce consistent inhibitory effects (see supplementary material Table S1). The first of these, XF06-5B03, completely inhibited all neutrophil migration to the wound site (Fig. 1E). This extract was derived from an ascomycete of the genus Sphaeropsidales, isolated from a soil sample collected in Singapore. The second extract, XF06-2A10, derived from the Aspergillus genus, similarly inhibited neutrophil recruitment (Fig. 1G).

A major challenge posed by screening a natural product library is the identification of the active component in the biological mixture. To identify the active component(s) identified by the tailfin assay, each extract was fractionated by high performance liquid chromatography (HPLC) into 38 fractions. Remarkably, only a single fraction of XF06-5B03 produced the same response as the crude extract. Liquid-chromatography–mass-spectrometry (LC-MS) analysis of the active fraction identified a chemical structure corresponding to a compound within the MerLion Pharmaceuticals purified natural product compound library. This compound, termed antibiotic PF1052 (CAS No. 147317-15-5), is a tetramic acid (Fig. 1J) first described by Meiji Seika Kaisha, Ltd as having antimicrobial properties (Sasaki et al., 1992). A small amount of the pure compound was obtained for further evaluation and confirmed as being the active component within the fraction. We also purchased PF1052 from an independent source (Enzo Life Sciences) and verified its activity as a highly effective neutrophil migration inhibitor, with an effective concentration as low as 2 μM (Fig. 1F,I).

Fractionation and LC-MS analysis of the second extract, XF06-2A10, revealed its active component to be sterigmatocystin (Fig. 1L). Pure sterigmatocystin powder purchased from Sigma-Aldrich reproduced the effect of XF06-2A10 at 50 μM, confirming its identity as the active component (Fig. 1H,I).

PF1052 works as antibiotic on bacteria in vitro at 2.3 mM (Koyama et al., 2005). We tested whether this concentration was comparable to that within the embryos by performing LC-MS analysis of zebrafish larvae pre-treated for 3 hours. Using LC-MS analysis, we found that absorption of PF1502 by zebrafish larvae was rapid, the concentration reaching 219 μM by 3 hours and 301 μM by 6 hours, with a concomitant fall in media concentration from 2 μM to 0.67 and 0.44 μM, respectively (Table 1). Larvae incubated in 50 μM sterigmatocystin had concentrations of 870 and 704 μM by 3 and 6 hours, respectively. The concentration of PF1052 is much lower than the published effective antibiotic concentration, suggesting a different mechanism for its neutrophil migration inhibitory effects.

We next asked whether the compounds identified using the zebrafish assay could also inhibit the migration of mammalian neutrophils using an EZ-TAXIScan chemotaxis device, in which a stable chemo-attractant gradient is formed in a 260-μm-wide channel (Fig. 2A). Freshly purified mouse neutrophils (treated with DMSO as control) migrated robustly up the gradient (see supplementary material Movie 1). By contrast, fMLF (a formyl peptide)-induced chemotaxis of neutrophils exposed to 10 μM PF1052 or 100 μM sterigmatocystin was severely compromised, with most neutrophils losing motility and polarity. Although a few cells managed to migrate out, they showed a lack of directionality and slow migration up the chemo-attractant gradient (Fig. 2B,C; also see supplementary material Movies 2 and 3). The effect was concentration dependent, with 20 μM PF1052 and 200 μM sterigmatocystin having a stronger effect (Fig. 2A; also see supplementary material Movies 4 and 5).

Because sterigmatocystin is both hepatotoxic and carcinogenic, we excluded it from further analysis and focused exclusively on PF1052. To investigate whether the effect of PF1052 on migration is specific to neutrophils, we monitored macrophage migration to a tailfin wound in PF1052-treated larvae and found no difference between treated and control animals (Fig. 3A,B). This is consistent with the inhibitory activity of PF1052 being specific to neutrophils rather than a general effect on innate immune cell migration.

To confirm that PF1052 acts to block the recruitment rather than the ontogeny of neutrophils, whole-body neutrophil counts were performed in the presence of the active compound or vehicle-only control. No differences in neutrophil number were observed following PF1052 treatment and the expected developmental increase in neutrophil number still occurred in the presence of PF1052 (Fig. 3C). We also tested the effect of PF1052 on the viability of human neutrophils in vitro. Human neutrophils cultured with or without a range of doses of PF1052 were assessed morphologically for apoptosis. At lower concentrations (200 nM and 2 μM) PF1052 suppressed the apoptosis of isolated human neutrophils; at higher concentrations, however, this suppressive effect was lost (Fig. 3D).

Using time-lapse imaging, we investigated how PF1052 blocks neutrophil migration to the wound site. In control larvae, treated only with DMSO, neutrophils began migrating towards the wound 10 minutes after ventral fin resection. During migration, the cells actively extended pseudopodia to sense the direction of the wound site (Fig. 4A,C). By 40 minutes, ~eight neutrophils had been recruited to the site of injury (red star in Fig. 4A). By contrast, no neutrophils were recruited to the wound site in PF1052-treated embryos. They appeared static during the 40 minutes of imaging, remaining rounded and barely able to form any pseudopodia (Fig. 4B,D).

The role of PI3K signalling in regulating F-actin polymerisation to promote neutrophil migration is well established. During migration, PI3K is localised to the leading edge of neutrophils, catalysing the phosphorylation of PtdIns(4,5)P₂ (PIP2) to generate PtdIns(3,4,5)P₃ (PIP3), which binds the PH domain of AKT to activate F-actin polymerisation at the cell front. To investigate whether PF1052 inhibits PI3K directly, we generated a transgenic line expressing an in vivo sensor of PIP3 (see Materials and Methods) to probe PI3K subcellular localisation (Fig. 5A). In Tg(lyz:PHAkt-EGFP)ⁱ²⁷⁷ larvae, fluorescent signal accumulated at the leading edge of neutrophils, usually in the pseudopodia (Fig. 5B). When two pseudopodia extended from a single neutrophil during chemotaxis, the one with higher levels of PHAkt-EGFP signal predicted the direction of migration. In Tg(lyz:PHAkt-EGFP) larvae treated with the PI3K inhibitor LY294002, accumulation of fluorescent signal was lost at the leading edge, but instead was dispersed throughout the cell body (Fig. 5C). By contrast, in larvae treated with XF06-B03 or PF1052, EGFP signal accumulated at the periphery of the neutrophils, despite their loss of polarity (Fig. 5D,E).

For a more robust quantification of the effect of PF1052 on neutrophil polarity, we devised a measurement termed the ‘polarity index’ that reflects the difference in EGFP intensity at the leading edge of the cell compared with the trailing edge, as well as the total fluorescence within the cell. 3-dpf Tg(lyz:PHAkt-EGFP)ⁱ²⁷⁷ larvae were pre-incubated with either 2 μM PF1052, 50 μM LY294002 or DMSO control for 3 hours, followed by tailfin transection. Individual neutrophils were imaged and the fluorescence intensity was quantified along a longitudinal transection of each cell to calculate the polarity index (as described in Materials and Methods). The majority of neutrophils from DMSO-control-treated larvae appeared polarised, with the most intense EGFP expression in pseudopodia at their leading edge and only a faint EGFP signal within the cell body (Fig. 5F). Consistent with our previous observations, neutrophils from LY294002- or PF1052-treated larvae did not have a defined leading edge (Fig. 5G,H), which was reflected in their significantly reduced polarity index compared with the DMSO control (Fig. 5I).

The pattern of EGFP signal observed in neutrophils from PF1052-treated larvae indicated that this compound acts independently of PIP3 generation at the cell membrane. We therefore tested whether PF1052 prevented the phosphorylation of AKT by inhibition of activating kinases at the membrane. To investigate this, we examined expression levels of phosphorylated AKT (p-AKT) following PF1052 treatment in human neutrophils using western blotting techniques. We found no change in p-AKT expression in neutrophils following either 4 or 8 hours culture with PF1052 (Fig. 6A–C). These data, along with our in vivo observations in the zebrafish, suggest that PF1052 acts independently of the PI3K-Akt signalling pathway to disrupt leading-edge specification and induce its inhibitory effect on neutrophil migration.

We have successfully used a transgenic zebrafish model in what we believe to be the first in vivo screen for naturally occurring inhibitors of neutrophil migration. Through subfractionation and LC-MS analysis, we identified the antibiotic PF1052 and sterigmatocystin as the active components of the two fungal extracts showing inhibitory activity. When applied to mouse neutrophils, both compounds identified in the zebrafish screen showed a dose-dependent inhibitory effect on migration, with an inhibition of pseudopod formation similar to that observed in their zebrafish counterparts. These findings further validate the utility of the Tg(mpx:GFP)ⁱ¹¹⁴ zebrafish as a model system for the analysis of human neutrophil pathobiology and illustrate the efficacy of such phenotypic assay systems for screening complex natural product extract mixtures.

Because sterigmatocystin is reported to have carcinogenic properties (Ellett et al., 2011), we chose to focus our attention on PF1052, the therapeutic potential of which has not previously been explored. PF1052 was initially isolated from fermentation broth of the fungus Phoma sp. and shown to have antimicrobial activity (Takahashi et al., 1996). It is structurally closely related to spylidone, also isolated from Phoma sp. fermentation broth, but unlike spylidone, PF1052 has no effect on lipid droplet accumulation in macrophages (Koyama et al., 2005). In this study, we found that PF1052 had no inhibitory effect on macrophage migration.

The effect of PF1052 on neutrophils was distinct from that of the PI3K inhibitor LY294002; neutrophils treated with the latter retain their ability to form thin pseudopods and PHAkt-GFP localises to the centre of the cell (Yoo et al., 2010). By contrast, treatment with PF1052 had no effect on Akt phosphorylation or PHAkt-EGFP localisation but caused loss of polarisation and pseudopod formation, reflected in the significantly reduced neutrophil ‘polarity index’. These effects are similar to those caused by a dominant-negative form of human Rac2D57N or morpholino-mediated knockdown of Rac2 (Deng et al., 2011). There are also similarities to the phenotype of mouse neutrophils mutant for SHIP, a Src homology 2 (SH2)-domain-containing inositol-5-phosphatase, which hydrolyses PI(3,4,5)P3 to PI(3,4)P2. SHIP was believed to coordinate with PI3K to determine the localisation of PIP3 and PIP2, thus controlling neutrophil polarity and motility (Nishio et al., 2007). In zebrafish, however, knockdown of SHIP leads to an increase in neutrophil motility and infiltration to the wound, whereas overexpression of SHIP impairs such migration (Lam et al., 2012). Although further analysis is needed to resolve the disparity between these two data sets, we are inclined to argue against an inhibitory effect of PF1052 on SHIP, based on the findings in zebrafish. Additional studies will be required to identify the specific target(s) of PF1052 in neutrophils.

Here, we demonstrate a feasible in vivo assay for the discovery of compounds with anti-inflammatory actions from natural compound libraries. Testing over 40 compounds with suitable positive and negative controls can easily be accomplished in a day by a single operator. This screen has identified a novel and highly effective neutrophil migration inhibitor, PF1052, with the potential to inform our understanding of neutrophil chemotaxis. We have found that this compound acts independently of the PI3K and AKT enzymes and might either act downstream of these molecules or act via other pathways. Identification of the molecular target has the potential to reveal a novel target pathway for future drug discovery programmes.

Adult fish were maintained on a 14-hour light/10-hour dark cycle at 28°C in the AVA certificated IMCB Zebrafish Facility (Singapore). The Tg(mpx:GFP)ⁱ¹¹⁴ line (Renshaw et al., 2006) was used for neutrophil assays. To investigate macrophage recruitment, we generated a Tg(mpeg1:Gal4.VP-16)^sh256 line as previously described (Ellett et al., 2011) and crossed this to the Tg(UAS:Kaede)^s1999t line (Davison et al., 2007) to generate double-transgenic larvae expressing green fluorescent Kaede protein under the macrophage-specific mpeg1 promoter.

The 6.6 kb of lysozyme C promoter (Hall et al., 2007) was cloned into a gateway vector p5E-MCS (Tol2kit). The PH domain of Akt was amplified from zebrafish cDNA by PCR using forward primer 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTCCACCATGGGCATGA-ACGAGATCAGCGTCGT-3′ and reverse primer 5′-GGGGACCAC -TTTGTACAAGAAAGCTGGGTCACCCAGCAGCTTTAAGTAGTCAA-3′, and the PCR product was recombined with pDONR 221 (Invitrogen) to produce pME-PHAkt. The final lyz:PHAkt-EGFP construct was created by recombination of p5E-lysC, pME-PHAkt, p3E-EGFP and pDestTol2pA2 (Tol2kit) using LR Clonase II plus (Invitrogen). The DNA construct was co-injected with Tol2 mRNA into one-cell-stage embryos to produce the transgenic line.

The microtubule inhibitor nocodazole (M1404), PI3K inhibitor LY294002 (L9908), sterigmatocystin (S3255) and Tricaine (E10521) were ordered from Sigma-Aldrich (St Louis, MO). PF1052 (ALX-380-147) was purchased from Enzo Life Sciences (Farmingdale, NY).

Crude fungal extracts were prepared by fermenting fungal strains in a variety of growth media and extracting the freeze-dried products of fermentation using methanol, followed by removal of solvent by rotary and centrifugal evaporation. 1 mg of each dried crude extract was dispensed per well of a 96-well microtitre plate.

The fungal extract was prepared at a concentration of 10 mg/ml in DMSO and centrifuged at 15,000 g for 2 minutes. An aliquot of extract (up to 800 μg or 80 μl) was analysed by HPLC [Thermos Hypersil BDS C18 4.6×150 mm, 5 μm, solvents: (A) H₂O + 0.1% formic acid, (B) MeCN + 0.1% formic acid; gradient: 0 minutes 0% B, 22 minutes; 100% B, 32 minutes; 100% B; flow: 1 ml/minute; 30°C). The eluent from the DAD (Agilent 1100 with UV detection at 210, 254, 354 and 480 nm) was split in a 1:10 ratio between an Esquire 3000 ion trap mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany) and a fraction collector configured to collect into a 96-well microtitre plate (0.79 minutes/well). A total of 38 wells were collected (0-30 minutes). The 96-well microtitre plate was dried using the Genevac HT-8 centrifugal evaporator (8-hour cycle) at 40°C. Wells 39 and 40 were used as controls, which contained 125 μg and 250 μg of crude extract, respectively. After complete evaporation of the solvent, materials in the wells were analysed for activity against the tailfin resection assay as described below. Active fractions were analysed by MS and MS-MS, and data matched against MerLion’s Esquire compound library, containing mass spectra records of 2496 compounds that have been analysed under the same conditions.

Zebrafish larvae treated with test compound were mixed with glass beads and 0.1 M of KH₂PO₄ buffer (150 μl), and smashed with TOMY Micro smash MS-100 (4500 rpm, 30 seconds ×4). After adding acetonitrile (350 μl), the mixture was shaken for 30 seconds ×2, and centrifuged at 20,000 g, 4°C, for 10 minutes for two cycles. Supernatant was separated for LC-MS analysis, which was carried out on a Waters 2795 Separations Module equipped with a Waters 2996 Photodiode Array (PDA) detector and micromass Quattro micro mass spectrometer.

Tg(mpx:GFP)ⁱ¹¹⁴ transgenic zebrafish larvae (2-3 dpf) were first exposed to extracts in a 96-well plate format, with three larvae per well, containing 200 μl of fish water, 1 mM of Tris pH 7.4, 1% DMSO and 10 μg fungus extract, at 28°C. Three hours later, the larvae were anaesthetised with Tricaine and their tailfins amputated using sharp needles to induce the migration of neutrophils. After injury, the zebrafish larvae were transferred to fresh extract solution for 3 hours before visual inspection using a fluorescence stereomicroscope. The number of neutrophils recruited to the wound site was counted. The effect of PF1052 on macrophage migration was assessed in the same manner, using the Tg(mpeg:Gal4;UAS:Kaede)^sh256 line.

The EZ-TAXIScan chamber (Effector Cell Institute, Tokyo, Japan) was assembled with a 260 μm wide × 4 μm thick silicon chip on a 2 mm untreated glass base, as described by the manufacturer, and filled with HBSS (with Ca²⁺ and Mg²⁺)/0.2% BSA. Murine neutrophils were isolated from bone marrows of 8-week-old male C57BL/6 mice (the Jackson Laboratories, Bar Harbor, ME). Drug-treated (or DMSO-treated) murine neutrophils (1 μl, 10×10⁶/ml) were added to the lower reservoir of each of the six channels and allowed to line up by removing 18 μl of buffer from the upper reservoir. HBSS (with Ca²⁺ and Mg²⁺)/0.2% BSA (15 μl with the appropriate pharmacological inhibitor) was then added to fill both reservoirs, which were maintained at 37°C. 1 μl of 1 μM chemoattractant (fMLF) was then added to the upper reservoir and neutrophil migration in each of the channels was captured sequentially every 30 seconds for 20 minutes using a 10× lens on a Discovery Screening System (Universal Imaging Corporation, Downingtown, PA). Drug PF1052 or sterigmatocystin was added directly to murine neutrophils (100 μl, 10×10⁶/ml) in HBSS (with Ca²⁺ and Mg²⁺)/0.2% BSA and incubated in a 37°C, 5% CO₂ chamber for 30 minutes before the chemotaxis assay.

The (x,y) coordinates of migrating neutrophils (i.e. neutrophils that cross >65 μm from the starting line) were tracked from sequential images using DIAS imaging software (Solltech, Oakdale, IA). Cell tracks were then realigned such that all the cells started from the same starting point (0,0) and plotted using Matlab (MathWorks, Natick, MA). Directionality (0 to 1) is defined as straight-line migration distance from the origin divided by the total migration length. Upward directionality (−1 to 1) is defined as straight-line distance migrated in the upward direction divided by total migration length. Migration speed (μm/minute) was calculated as the average of cell speeds (migration distance between the current frame and the previous frame divided by the time between sequential frames, 0.5 minutes) at each captured frame.

Neutrophils were isolated from healthy volunteers as previously described (Wardle et al., 2011), with approval of the South Sheffield Research Ethics Committee (reference number: STH13927). Neutrophils were cultured with DMSO, or 0.2, 2 or 20 μM PF1052 for 8 hours at 37°C, and apoptosis was assessed by morphology on cytospins stained with Quick-Diff (Gentaur, Brussels, Belgium).

Neutrophils were isolated and cultured as described above, for either 4 or 8 hours, with 2×10⁶ cells per condition. Cells were lysed as previously described (Wardle et al., 2011) and western blotting was performed according to standard protocols (Sambrook, 1989). Antibodies used were anti-phospho-Akt (Ser473) (Cell Signaling, Herts, UK) at 1:1000, or anti-β-actin (Sigma-Aldrich, Poole, UK) at 1:4000, with polyclonal goat anti-rabbit secondary antibody (Dako, UK) at 1:2000. To analyse changes in protein expression, densitometry was performed using ImageJ, as previously described (Wardle et al., 2011).

For time-lapse imaging, Tg(lyz:PHAkt-EGFP)ⁱ²⁷⁷ zebrafish larvae were anaesthetised with 0.02% Tricaine and embedded in 1% low melting agarose on a depression slide. Embedded larvae were then placed under a 40× water objective, immersed in water containing Tricaine. Images were captured every 15 seconds, using an Olympus spinning disc fluorescence microscope (BX61) or Olympus Confocal microscope (BX61). To assess the effects of PF1052 on neutrophil polarity, Tg(lyz:PHAkt-EGFP)ⁱ²⁷⁷ larvae were pre-incubated with either 2 μM PF1052, 50 μM LY294002 or DMSO control for 2 hours prior to tailfin transection and mounting. Images of individual neutrophils in the region between the injury site and the posterior blood island were captured using an UltraVIEWVoX spinning disk confocal imaging system (PerkinElmer Life and Analytical Sciences) with an inverted Olympus IX81 microscope, at 60× magnification with ten Z slices. Images were analysed in ImageJ, by drawing a transection through each neutrophil from the trailing edge towards the leading edge. In cases where there were no clear leading or trailing edges, the line was drawn through the most longitudinal section of the cell. A plot profile was generated to measure the fluorescence intensity per pixel along the length of the line and the mean intensities along sections of this line (as defined in Fig. 5A) were used to calculate the polarity index, with the equation:

Data were analysed in GraphPad Prism 5.0 using one-way ANOVA with appropriate post-test adjustment.

The authors thank the colleagues of P.W.I. lab, and Sun Tao, Li Zhi and Liu Huijuan in Tianjin for technical assistance.