Control of cytokine-driven eosinophil migratory behavior by TGF-beta-induced protein (TGFBI) and periostin

Periostin, which is induced by interleukin (IL)-13, is an extracellular matrix (ECM) protein that supports αMβ2 integrin-mediated adhesion and migration of IL-5-stimulated eosinophils. Transforming growth factor (TGF)-β-induced protein (TGFBI) is a widely expressed periostin paralog known to support monocyte adhesion. Our objective was to compare eosinophil adhesion and migration on TGFBI and periostin in the presence of IL-5-family cytokines. Eosinophil adhesion after 1 h and random motility over 20 h in the presence of various concentrations of IL-5, IL-3, or granulocyte macrophage-colony stimulating factor (GM-CSF) were quantified in wells coated with various concentrations of TGFBI or periostin. Results were compared to video microscopy of eosinophils. Cytokine-stimulated eosinophils adhered equivalently well to TGFBI or periostin in a coating concentration-dependent manner. Adhesion was blocked by anti-αMβ2 and stimulated at the lowest concentration by GM-CSF. In the motility assay, periostin was more potent than TGFBI, the coating-concentration effect was bimodal, and IL-3 was the most potent cytokine. Video microscopy revealed that under the optimal coating condition of 5 μg/ml periostin, most eosinophils migrated persistently and were polarized and acorn-shaped with a ruffling forward edge and granules gathered together, in front of the nucleus. On 10 μg/ml periostin or TGFBI, more eosinophils adopted a flattened pancake morphology with dispersed granules and nuclear lobes, and slower migration. Conversion between acorn and pancake morphologies were observed. We conclude that TGFBI or periostin supports two modes of migration by IL-5 family cytokine-activated eosinophils. The rapid mode is favored by intermediate protein coatings and the slower by higher coating concentrations. We speculate that eosinophils move by haptotaxis up a gradient of adhesive ECM protein and then slow down to surveil the tissue.

Transforming growth factor (TGF) β-induced protein (TGFBI, βig-h3) is a paralog of periostin and, like periostin, contains an amino (N)-terminal cysteine-rich sequence and four fasciclin-1 (FAS1) modules [1]. The disulfide pattern of TGFBI was recently determined [8]. Intermodule bridges between the cysteine-rich sequence and the second FAS1 module and between the first two FAS1 modules indicate that the N-terminal cysteine-rich sequence and the first two FAS1 modules adopt a complicated fold, possibly leaving the third and fourth FAS1 modules exposed [8]. This model should be applicable to periostin, which contains the same set of cysteine residues [1]. Unlike periostin, TGFBI lacks an alternatively spliced tail carboxy (C) terminal to the fourth FAS1 module [1]. TGFBI is an ECM protein that is widely expressed, including in the lung, and is induced by TGF-β in bronchial myofibroblasts and other fibroblasts [1,9,10]. Genetic deletion of TGFBI is not lethal but has multiple effects, including defective development of alveolar structure and function in mice [11]. Although studies comparing and contrasting activities of TGFBI and periostin head-to-head are sparse, TGFBI and periostin were recently found to act similarly in the heart in affecting fibrosis and disease responsiveness; however, TGFBI is seemingly not necessary in the heart after myocardial infarction injury and is fully compensated by the more prominently expressed effector periostin [12].
TGFBI is known to support α M β 2 -integrin-mediated adhesion of monocytes [13]. We asked whether TGFBI adsorbed in a microtiter plate, like adsorbed periostin, supports eosinophil adhesion and migration induced by IL-5 family cytokines in short-term adhesion and long-term bead-clearing migration assays, which model adhesion to and migration on ECM molecules. To complement the bead-clearing assay, we examined the migratory behavior of eosinophils on TGFBI or periostin adsorbed on glass by video differential interference contrast (DIC) microscopy. We report that TGFBI and periostin are both adhesive ligands for eosinophil α M β 2 integrin that may be important for eosinophil recruitment and retention in ECM, and that IL-5 family cytokines acting on eosinophils have different potencies for adhesion and migration and potentially play different roles at various times after eosinophil activation. We observed two morphologies and modes of migration of activated eosinophils on TGFBI or periostin that are pertinent to the migration results: rapidly-migrating polarized, acorn-shaped cells, which tended to be most frequent on intermediate periostin coating, and slower flatter, pancake-shaped cells most frequent on high protein coating.

Eosinophils
Eosinophils were obtained and purified from heparinized blood of donors with allergy and with or without mild allergic asthma by negative selection using a cocktail of anti-CD16, anti-CD14, anti-CD3, and anti-glycophorin beads as before [4]. Subjects with prescriptions for low doses of inhaled corticosteroids (ICS) did not use their ICS on the day of the blood draw. The purity and viability of eosinophils were ! 98% [4]. The studies were approved by the University of Wisconsin-Madison Health Sciences Institutional Review Board (protocol No. 2013-1570. Informed written consent was obtained from each subject before participation.

TGFBI and periostin
Human TGFBI (Uniprot identifier No. Q15582) complementary DNA (cDNA) was obtained by reverse transcription of RNA isolated from MG-63 cells using M-MLV reverse transcriptase (Promega, Madison, WI, USA) and a TGFBI-specific oligonucleotide (5' tgc att cct cct gta gtg c 3') coding for a sequence at the beginning of the 3' untranslated region. The cDNA was used as template for polymerase chain reaction (PCR) and cloned into pAcGP67.coco [14] using standard molecular biology techniques. The complete insert sequence was verified before expression. TGFBI was produced as a secreted protein by insect cells using a baculovirus system and purified as described [14]. PN-S, the shortest splice variant of human periostin containing the N-terminal cysteine-rich sequence and the four FAS1 modules but lacking variably spliced Cterminal sequences encoded by exons 17, 18, 19, and 21 (Uniprot No. Q15063-7), was produced in insect cells as before [3,4,15]. Recombinant human TGFBI and longest periostin variant (PN-L) (Uniprot No. Q15063-1) [3,4], expressed by mouse NSO cells, were purchased from R&D Systems (Minneapolis, MN, USA).

Cell adhesion assay
Eosinophil adhesion for 1 h was quantified using an eosinophil peroxidase assay as described [3], with the following modifications. After eosinophils were resuspended at 2 x 10 6 /ml in Roswell Park Memorial Institute (RPMI) medium with 0.2% bovine serum albumin (BSA), cells were equilibrated for 1 h at 37˚C [16] before dilution and addition to coated wells. Further, the colored eosinophil peroxidase product was measured in a SpectraMax M5 plate reader (Molecular Devices, Sunnyvale, CA, USA).

Cell motility assay
Motility of eosinophils, equilibrated as in the cell adhesion assay, was assessed in a bead clearing assay that ran for 20 h with wells viewed in an Eclipse Ti inverted microscope (Nikon, Melville, NY, USA) and images acquired and quantified as before [4].

Differential interference contrast (DIC) video microscopy
Glass bottom culture 35 mm diameter Petri dishes with 14 mm glass diameter No. 1.0 coverslip (MatTek, Ashland, MA, USA) were used. The glass area was coated with 350 μl protein (5 or 10 μg/ml) in Tris-buffered saline (TBS), pH 7.4, for 2 h at 37˚C, washed with TBS, blocked with neat fetal bovine serum (FBS) for 30 minutes at 37˚C, and washed with TBS. Eosinophils, resuspended at 2 x 10 6 /ml in RPMI with 10% FBS and equilibrated as above, were diluted in the same medium to 2 x 10 5 /ml, and 2 ml of this cell suspension was added to the dish. Two μl of IL-5 stock solution (50 μg/ml) was added, yielding a final concentration of 50 ng/ml. Cells were viewed in a DMi8 inverted wide-field microscope (Leica, Buffalo Grove, IL, USA) with a motorized stage and Tokei Hit temperature-and CO 2 -controlled chamber using a 63x oil immersion objective, at the University of Wisconsin-Madison Optical Imaging Core facility. Five minute videos with 1 frame/s were acquired up to 35 minutes after addition of IL-5, and videos and individual images were exported, using Leica Application Suite X (LASX) software. Cells defined as having an acorn-shaped morphology were polarized with a ruffling leading edge and pseudopods, the nuclear lobes at the rear, and the granules closely gathered together between the forward edge and the nucleus and moving as a unit in a coordinated manner. Cells defined as having a "pancake-shaped" morphology were flatter, more spread with dispersed granules and nuclear lobes. Cells defined as unactivated were round, not polarized, not flat, and not spread. Cell area was quantified using Fiji software (http://fiji.sc/Fiji).

Statistical analysis
Student's t test was used to compare data between two groups. This was complemented for some experiments by analysis of variance (ANOVA) with Dunnett's or Tukey's multiple comparisons post test in order to compare data among groups. P 0.05 was considered significant. Analyses were performed using Prism (GraphPad, San Diego, CA, USA).

TGFBI supports eosinophil adhesion with similar coating concentration dependence as periostin variants
In order to determine whether TGFBI supports eosinophil adhesion and compare it to periostin, assays were performed in microtiter plates in which wells were coated with different concentrations of recombinant human TGFBI commercially available from R&D Systems or produced in insect cells in our laboratory, the longest human periostin splice variant purchased from R&D (PN-L), or the shortest human periostin splice variant (PN-S) produced in insect cells. Wells were post-coated with neat fetal bovine serum (FBS). Control wells were coated only with FBS.
R&D and insect cell TGFBI supported adhesion at 1 h of eosinophils stimulated with IL-5 (10 ng/ml) compared to the FBS control in a manner dependent on the coating concentration ( Fig 1A). R&D PN-L or insect cell PN-S also supported adhesion of IL-5-stimulated eosinophils in a coating concentration-dependent manner (Fig 1A), as we found previously [3]. The coating concentration dependences of R&D TGFBI, baculovirus TGFBI, PN-L, and PN-S were similar, with no significant difference among the four proteins at each coating concentration ( Fig 1A).

GM-CSF stimulates eosinophil adhesion to TGFBI with a different doseresponse curve than does IL-5 or IL-3
The stimulating effect of IL-5 on eosinophil adhesion to TGFBI at 1 h was dose dependent ( Fig  1C). The IL-5 family cytokines IL-3 and GM-CSF also stimulated adhesion to TGFBI (Fig 1C). IL-3 stimulated adhesion with a similar dose-response curve as did IL-5 ( Fig 1C). The doseresponse curve of GM-CSF was shifted left compared to those of IL-5 or IL-3, with GM-CSF being clearly active down to 0.1 ng/ml; thus, GM-CSF was active at a lower concentration than IL-5 or IL-3 ( Fig 1C). These curves were similar to those for cytokine-stimulated eosinophil adhesion to periostin, on which GM-CSF was also active down to 0.1 ng/ml [3].

TGFBI supports motility of eosinophils but is less effective than periostin
The effect of TGFBI on surface-dependent motility (haptokinesis) of eosinophils over 20 h was assessed with a microbead monolayer assay [3,4]. After protein coating and blocking, microbeads were added at a concentration that, after centrifugation of the plate, forms a nearly confluent "lawn" of beads. In Fig 2A, a positive example with PN-L coated at 5 μg/ml, white areas cleared of microbeads represent paths of migration in contrast to the dark bead monolayer background. Fig 2B shows a negative example with a control well only coated with neat FBS. Motility was quantified as cleared area of tracks using Fiji software [4]. Immobilized TGFBI supported random motility of IL-5-stimulated eosinophils ( Fig 2D). However, the motility on TGFBI in the presence of 10 ng/ml IL-5 was variable among donors (compare the three donors in Fig 2D with a mean of 7% of area cleared and the three donors in Fig 2C with 1% under the same experimental conditions). In a comparison using the same donors, TGFBI was less effective in supporting motility than PN-L or PN-S in the presence of 10 ng/ml IL-5 (Fig 2C). At a 10 μg/ml coating concentration, PN-S or PN-L were more effective in supporting eosinophil motility than TGFBI (Fig 2C). Further, at 5 μg/ml, PN-L was more effective than PN-S (Fig 2C). Motility on PN-L was maximal at this concentration (5 μg/ml) (Fig 2C), as we have found before [3].

IL-3 stimulates eosinophil motility on TGFBI or periostin to a higher degree than does IL-5 or GM-CSF
The stimulating effect of IL-5 on eosinophil motility to TGFBI (Fig 2D) or PN-L (Fig 2E) over 20 h was dose-dependent. IL-3 and GM-CSF also stimulated motility on TGFBI ( Fig 2D) and PN-L ( Fig 2E). IL-3 stimulated motility up to a higher maximal level than did IL-5 or GM-CSF on TGFBI or PN-L, with IL-3 having a greater effect than IL-5 on both proteins (Fig 2D and  2E). In addition, we observed that in the absence of cytokine (i.e., at 0 ng/ml), there was no or essentially no motility on TGFBI (Fig 2C and 2D), whereas there was, as we have noted before [3], a low degree of motility on PN-L ( Fig 2E).

Video microscopy of eosinophil motility on TGFBI or periostin
Video microscopy with differential interference contrast optics was performed up to 35 minutes after addition of IL-5, to compare further the effects of adsorbed TGFBI and periostin, and of two different coating concentrations (5 and 10 μg/ml), on eosinophil motility, and to examine the morphology of migrating eosinophils and mode of migration. The majority of eosinophils migrating on TGFBI or periostin had a polarized morphology and were acornshaped (Fig 3A and S1-S4 Movies, particularly S1 Movie). The morphology was somewhat similar to eosinophils that have been activated by IL-5 in suspension and examined after cytospinning [16], except that the migrating cells had a ruffling leading edge with pseudopods ( Fig  3A and S1-S4 Movies). This morphology tended to be most frequent on periostin coated at 5 μg/ml (82% ± 7% of the cells, mean ± standard error of the mean [SEM], n = 5, at 5-10 minutes) (see S1 Movie). Migration of acorn-shaped IL-5-activated eosinophils was persistent and each coating concentration. (B) Wells were coated with R&D TGFBI at 5 μg/ml followed by post-coat with FBS. Eosinophils were preincubated with mAb (10 μg/ml) to the indicated integrin subunit. Mean ± SEM, n = 2 donors, ÃÃ P < 0.01, Ã P < 0.05 versus isotype control. ANOVA confirmed significant difference (P < 0.01) among treatments and Dunnett's post test confirmed significant difference (P < 0.01) of each of the mAbs versus isotype control. The number of adherent cells with isotype control was 27% ± 15%. (C) Wells were coated with R&D TGFBI at 5 μg/ml followed by post-coat with FBS. Mean ± SEM of two wells per treatment in one representative experiment (of two). In addition, similar dose response curves were obtained in two experiments in which wells were coated with 2 μg/ml TGFBI.
https://doi.org/10.1371/journal.pone.0201320.g002 acorn-shaped eosinophils on periostin coated at 5 μg/ml was 0.2 μm/s. Some cells had a flatter, more spread, and "pancake-shaped" morphology with dispersed granules and nuclear lobes (Fig 3B, S1-S4 Movies, particularly S2 and S4 Movies). Of the pancake-shaped cells, some migrated persistently albeit more slowly than the acorn-shaped cells, some migrated intermittently, and some did not migrate at all. The pancake-shaped morphology was most frequent on periostin or TGFBI coated at 10 μg/ml (Fig 3C). Pancake-shaped cells eosinophils had a mean area of 199 μm 2 , which was significantly higher than that of acorn-shaped cells (Fig 3D). In addition, a few cells were observed alternating between the acorn-and pancake-shaped morphologies, transitioning back from pancake-to acorn-shaped (also see S1-S4 Movies).

Discussion
The widely expressed ECM protein TGFBI, which is a paralog of periostin and induced by TGF-β in several cell types [1,9], is shown to support specific adhesion and migration of blood eosinophils stimulated by IL-5 family cytokines IL-5, IL-3, or GM-CSF. Eosinophil adhesion was sensitive to the coating concentration of TGFBI and concentration of the activating cytokine. Effects of coating concentration was similar for TGFBI and periostin, indicating that the adhesive site(s) reside(s) somewhere in the homologous region encompassing the N-terminal cysteine-rich sequence and the four FAS1 modules shared by the two proteins [1]. Eosinophil adhesion to TGFBI was mediated by α M β 2 integrin, for which the highly activated conformation, reported by mAb CBRM1/5, is known to be induced by IL-5 or IL-3 [17][18][19][20][21]. The dependence on α M β 2 for adhesion of eosinophils to TGFBI is consistent with the report that TGFBI supports monocyte adhesion in an α M β 2 -dependent manner [13]. The present adhesion results indicate that the paralogs TGFBI and periostin support adhesion of eosinophils by similar mechanisms. Immobilized TGFBI also significantly supported random motility of cytokinestimulated eosinophils. However, TGFBI was less effective than PN-L or PN-S, and PN-S was less effective than PN-L in supporting migration. These quantitative differences in motility on TGFBI, PN-S, and PN-L implicate the alternatively spliced sequences in the C-terminal tail of periostin [1] as being important for maximal migration. Thus, in addition to (an) adhesive site (s) in the major and conserved part of TGFBI and periostin, there appears to be another, promigratory, site in full-length periostin's tail.
GM-CSF was the most potent cytokine for adhesion to TGFBI at 1 h, being active at a lower concentration than IL-5 or IL-3. The potency may be related to the greater abundance of CS2RA (GM-CSF receptor α) in comparison to IL3RA (IL-3 receptor α) and IL5RA (IL-5 receptor α) subunits in blood eosinophils at baseline, as found by quantitative proteomic analysis [22]. IL-3 supported greater eosinophil motility over 20 h on TGFBI or periostin (PN-L) than IL-5 or GM-CSF. This is in line with the observation that IL-3 upregulates the cell-surface expression of and induces the high-activity conformation of α M β 2 integrin at 20 h to a significantly greater degree than does IL-5 [21]. The differential effect between the cytokines may in turn be a function of the fact that IL-3 upregulates the cell-surface level of its cognate IL-3 receptor α subunit (IL3RA), whereas IL-5 downregulates IL-5 receptor α (IL5RA) at 4-24 h [20,[23][24][25][26][27]. Overall, the greater effect by IL-3 at 20 h is consistent with the scenario that IL-3 induces stronger and prolonged signaling and translation in eosinophils than does IL-5 or GM-CSF [21,25,[28][29][30]. Our results hint at important differences among IL-5 family cytokines in their effects on eosinophil adhesion and migration at various times after eosinophil activation. IL-5, IL-3, and GM-CSF are present in bronchoalveolar lavage (BAL) fluid from subjects with mild allergic asthma and are increased after segmental lung antigen challenge [31]. IL-5 is produced by several cell types including T helper cell type 2 (TH 2 ) cells, mast cells, NK cells, and type 2 innate lymphoid cells; IL-3 is produced by, e.g., T cells, macrophages, stromal cells, NK cells, and mast cells; and GM-CSF by a variety of cells such as T cells, macrophages, endothelial cells, fibroblasts, and epithelial cells [32,33]. The greater effect by GM-CSF and IL-3 than IL-5 on eosinophil adhesion at 1 h and migration over 20 h, respectively, may motivate therapeutic targeting of GM-CSF or IL-3, or their receptors, in addition to or as an alternative to IL-5 or its receptor, in order to inhibit activation of eosinophils and their interactions with matrix in tissues in eosinophilic diseases.
Video microscopy revealed that periostin and TGFBI support two modes of migration by IL-5 family cytokine-activated eosinophils. The majority of migrating eosinophils had, at least during the first 35 minutes after activation with IL-5, a polarized acorn-shaped morphology with a ruffled forward edge and migrated rapidly and persistently with granules closely clustered together, in front of the nucleus. This morphology tended to be most frequent and thus appears to be favored on periostin (PN-L) coated at 5 μg/ml. Cells with the acorn-shaped morphology had a mean area of 108 μm 2 . In addition, this morphology is similar to that of newt (Taricha granulosa) eosinophils migrating on quartz coverslips in the presence of newt serum; the newt eosinophils were observed to be tear-drop shaped and polarized, crawling with an advancing lamellipod and the nucleus located near the other, rear end [34,35]. Some eosinophils in our experiments adopted a flattened pancake-shaped morphology with dispersed granules and nuclear lobes, and slower migration. This morphology was most frequent on a protein coating concentration of 10 μg/ml. Eosinophils with the pancake-shaped morphology had a mean area of 199 μm 2 , significantly greater than and almost twice as large as the acornshaped cells. A few cells converted between the two morphologies transitioning back from pancake-to acorn-shaped. We do not know how much time the eosinophils spend in either mode or morphology during a longer time, e.g., during 20 h, which was the incubation time in the bead-clearing motility assay. The acorn-shaped morphology appears to be somewhat similar to that described for migrating neutrophils [36], except that in the acorn-shaped migrating eosinophils, the nucleus is localized at the rear end of the cell, a localization similar to that in suspended acutely IL-5 activated eosinophils [16], whereas in migrating or suspended activated neutrophils the nucleus has a random localization [16,36].
In the lung, TGFBI appears to be produced by resident cells and is localized to the vascular and airway ECM [10]. TGF-β is synthesized by a large variety of cells, including epithelial cells, fibroblasts, and immune cells [32], is produced in various disease states, and is believed to contribute to tissue remodeling, e.g., in the airway in asthma [37][38][39][40]. Eosinophils themselves, which contain TGF-β1 [22], are believed to be one source of TGF-β in eosinophilic disorders [37]. Periostin is upregulated by IL-13 in bronchial epithelial cells and lung fibroblasts and deposited widely in the bronchi of subjects with asthma [1,[41][42][43][44]. IL-13, which is secreted by, e.g., TH 2 cells [2], acts on epithelial cells to cause basal secretion of periostin [42]. Periostin is not detected or detected at very low levels (pg/ml) in bronchoalveolar lavage from normal or asthmatic subjects, indicating that secretion of periostin into the bronchial lumen does not occur or is negligible [45]. In addition, periostin can be secreted by stromal cells stimulated by TGF-β and other cytokines and growth factors at sites of injury or inflammation [41,[46][47][48]. It should be emphasized, however, that little is known about how TGFBI or periostin is deposited in the ECM [1], and therefore caution should be taken in extrapolating studies done on interactions of eosinophils with coatings of these proteins to events that may take place in the lung. It should also be noted that nine splice variants of periostin have been identified, only two of which were tested by us. Eight splice variants have been detected in normal fetal lung, including the longest isoform and isoforms lacking 1-4 of exons 17, 18, 19, and 21, whereas five variants, not including the longest one, were found in adult lung [49,50]. In idiopathic pulmonary fibrosis (IPF), exon 21 was discovered to be more likely spliced out than in controls [51]. Nevertheless, the findings lead us to speculate that eosinophils move rapidly by haptotaxis up a gradient of adhesive ECM ligand in the form of periostin or possibly TGFBI and, when having reached a site with higher density of ligand, slow down to surveil and "decide" whether to reside quietly in the tissue or release effectors.