Background
Air pollution epidemiology has traditionally focused on cardiovascular and respiratory outcomes. These adverse associations have been extended to show the acceleration of cognitive decline of elderly community-based populations [
1‐
5] and neurodevelopmental impairments of children [
6,
7]. The causes of cognitive impairment are being analyzed in rodent and cell models, which implicate neuroinflammatory responses to urban air pollutants [
8‐
11]. Specifically, we and others observed that the ultrafine size class of air pollution PM0.2 (<0.2 μm diameter) activated microglia and induced TNFα and IL-1, among other inflammatory responses [
10,
12‐
14]. This evidence supports findings of increased microglial activation and white matter hyperintensities in small postmortem samples of children from a highly polluted Mexican city [
7,
15] and in the association of white matter loss in older human adults in an MRI analysis of the WHIMS cohort of US women [
16].
We focus on traffic-derived ultrafine PM, which consistently shows higher toxicity than larger PM in vivo and in vitro [
17,
18], in neonatal rodents. Artificial ultrafine PM is rapidly transported after inhalation into the brain via the olfactory pathway [
19,
20]. Within the ultrafine PM, we examined a subfraction eluted from filters into aqueous suspension for its neurotoxicity and pro-inflammatory activity [
10,
11]. This subfraction is designated as nano-sized PM (nPM) to distinguish it from the total ultrafine PM and is depleted in black carbon and water-insoluble organics (Table
1) [
10]. The nPM fraction is highly active in vitro and in vivo after re-aerosolization, with free radical EPR signals that persisted
>30 days after initial collection. Notably, ozone and other gaseous pollutants with epidemiological cognitive associations [
2,
21] are absent from filter-collected nPM. In rodent cell models, nPM has both direct and indirect effects on neuronal viability and neurite outgrowth [
10]. Because TNFα is induced by chronic inhalation of ultrafine PM [
8,
22] and because TNFα can alter neurite outgrowth [
23‐
25], we further evaluated the role of TNFα in rapid brain responses to nPM. We first investigated the olfactory epithelium (OE), since little is known about the initial cellular responses of the olfactory gateway to urban traffic-derived ultrafine (or, equivalently) nPM. Based on the precedent of ex vivo OE incubation from OE biopsies [
26,
27], we developed an ex vivo model for incubation of the intact OE within neonatal mouse nasal cavities with nPM suspensions. In addition, using glia and neurons derived from neonatal cerebral cortex, we analyzed mechanisms by which nPM-induced TNFα inhibits neurite outgrowth.
Table 1
Composition of nPM
Black carbon | 13 | 1 | 7 |
Organic carbon, water soluble | 32 | 34 | 100 |
Organic carbon, water insoluble | | | |
Hopanes-steranes | 0.012 | 0.001 | 8.5 |
Organic acids | 0.097 | 0.009 | 9 |
Polyaromatic hydrocarbons | 0.02 | Not detected | 0 |
Metals (Cu, Fe, Ni, V) | | | >90 |
Methods
nPM collection and transfer into aqueous suspension
Nano-sized particulate matter (nPM; <0.2 μm in diameter) was collected on Teflon filters by a High-Volume Ultrafine Particle (HVUP) Sampler [
28] at 400 l/min flow in urban Los Angeles, downwind from the local I-110 Freeway [
10]. These samples are a mix of fresh ambient PM, mostly from vehicular traffic emissions and secondary aerosols [
29,
30]. The nPM samples were collected continuously during July–Sept. 2010 and Nov. 2011–Feb. 2012; these pooled samples approximate the annual average composition of nPM near the I-110 corridor [
31]. The filter-trapped dried nPM were eluted by sonication into deionized water. The nPM comprise 20 % by mass of ambient PM2.5. Water-soluble metals and organic compounds were efficiently transferred (Table
1). Relative to the total filter-trapped ultrafines (PM0.2), the nPM subfraction eluted into aqueous phases is depleted in black carbon and water-insoluble organic compounds. nPM suspensions (350 μg/ml) were stored at −20 °C. For controls of nPM extracts, fresh sterile filters were sham-extracted.
Animals
C57BL/6J mice were purchased from The Jackson Laboratory (Sacramento, CA, USA) for breeding and pregnant Sprague Dawley rats from Harlan Labs (Livermore, CA, USA). Animals were maintained following NIH guidelines, approved by the USC Institutional Animal Care and Use Committee (IACUC). Animals were euthanized by cervical dislocation after anesthesia by isoflurane or CO2.
Nasal cavity ex vivo incubation
P3 mice (both sexes) were anesthetized and decapitated; the nasal bone was removed to reveal the nasal cavity. The entire nasal cavity including the snout intact was removed in the gross. Nasal cavities were incubated with 12 μg/ml nPM in artificial cerebral spinal fluid (CSF) for 2 h/37 °C. After incubation, the OE was peeled from the nasal cavity for quantitative polymerase chain reaction (qPCR) or immunohistochemistry. Mice were chosen for these experiments because their smaller size facilitates slide preparation and obviates decalcification.
Cell culture
Mixed glia were originated from the cerebral cortex of postnatal day 3 (P3) rats (both sexes). Primary glia were grown in Dulbecco’s modified Eagle’s medium/Ham’s F12 50/50 Mix (DMEM F12 50/50) supplemented with 10 % fetal bovine serum (FBS) and 1 %
l-glutamine in a humidified incubator (37 °C/5 % CO
2) [
32]. After culture for 2.5 weeks, their composition was 3:1 astrocytes:microglia. Microglia were isolated by shaking for 4 h/37 °C. Embryonic day 18 (E18) rat cortical neurons were originated at 15,000 neurons/cm
2 on poly-
d-lysine-coated coverslips in DMEM supplemented with B27 (Invitrogen, Grand Island, NY).
For in vitro exposure, mixed glia were trypsinized and replated in six-well plates at 1 × 10
6 cells/well and grown overnight. Secondary cultures of mixed glia were treated with nPM aqueous suspensions (12 μg/ml) diluted in neuronal media for 24 h before assay. This dose consistently induced glial TNFα and IL-1α messenger RNA (mRNA) [
10]. The resulting conditioned media (CM) was collected and centrifuged (10,000
g/10 min) to remove residual cells. For small interfering RNA (siRNA) experiments, mixed glia were treated with siRNA (Silencer Negative Control No. 1 siRNA, AM4611; Ambion, Austin, TX) or TNFα siRNA (AM16708, Ambion). Scrambled and TNFα siRNAs were mixed with a siPORT NeoFX transfection agent (Ambion) to 50 nM. Mixed glia were grown for 24 h post transfection and then treated with nPM or vehicle before plating onto E18 neurons. Immunoneutralization of TNFα used 20 μg/ml antibody (MAB510; R&D Systems, Minneapolis, MN); TNF receptor activity was inhibited by TNFR1/2 blocking peptide (E-20, L-20; SCBT, Dallas, TX) at 5 μg/ml before CM application. Rats were used for in vitro experiments, following our prior studies [
10] and the better yields of microglia than from mice.
Quantitative polymerase chain reaction
Total cellular RNA was extracted using TRI reagent (Sigma, St. Louis, MO). cDNA was prepared from 1 μg of RNA by Superscript III RT kit (Invitrogen, Carlsbad, CA) and analyzed by qPCR with appropriate primers for both mouse and rat for Ct (threshold cycle) values. Genes examined by qPCR include TNFα (forward: 5′ CGTCAGCCGATTTGCTATCT 3′; reverse: 5′ CGGACTCCGCAAAGTCTAAG 3′) (CT range 26–30), Iba1 (forward: 5′ CCTGATTGGAGGTGGATGTCAC 3′; reverse: 5′ GGCTCACGACTGTTTCTTTTTTCC 3′) (CT range 25–26), IL-1α (forward: 5′ TCGGGAGGAGACGACTCTAA 3′; reverse: 5′ GTGCACCCGACTTTGTTCTT 3′) (CT range 29–31), GFAP (forward: 5′ CCAAGCCAAACACGAAGCTAA 3′; reverse: 5′ AGGAATGGTGATGCGGTTTTC 3′) (CT range 30–31), iNOS (forward: 5′ CATTGGAAGTGAAGCGTTTCG 3′; reverse: 5′ CAGCTGGGCTGTACAAACCTT 3′) (CT range 27–29), TNFR1 (forward: 5′ GGGCACCTTTACGGCTTCC 3′; reverse: 5′ GGTTCTCCTTACAGCCACACA 3′) (CT range 22–23), TNFR2 (forward: 5′ CAGGTTGTCTTGACACCCTAC 3′ reverse: 5′ GCACAGCACATCTGAGCCT 3′) (CT range 25–26), βIII-tubulin (forward: 5′ CGCACGACATCTAGGACTGA 3′; reverse: 5′ TGAGGCCTCCTCTCACAAGT 3′) (CT range 19–20), and rGAPDH (forward: 5′ AGACAGCCGCATCTTCTTGT 3′; reverse: 5′ CTTGCCGTGGGTAGAGTCAT 3′) (CT range 16–17). Data were normalized to GAPDH and quantified as ΔΔCt.
ELISA
CM from nPM-treated glia was sampled after 24 h of exposure and analyzed for TNFα by solid phase sandwich ELISA (BD Biosciences, San Jose, CA).
Immunohistochemistry
The OE and olfactory bulb of P3 neonatal mice were fixed with 4 % paraformaldehyde in phosphate buffered saline (PBS) pH 7.4. Specimens were immersed in 10 % sucrose/PBS pH 7.4, then 30 % sucrose/PBS pH 7.4 at 4 °C, then embedded in optimal cutting temperature compound (OCT; Fisher Scientific, Waltham, MA) before transverse cryostat sectioning (18 μm). Antigen retrieval was performed by submerging slides in 10 mM sodium citrate buffer and microwaving for 3 min. Tissue was permeabilized with 1 % NP-40/PBS and blocked with 5 % BSA, then probed with antibodies specific for the Olfactory Marker Protein of olfactory sensory neurons (OMP 1:100; SCBT, Dallas, TX), βIII-tubulin (1:400; Sigma Chemical Co., St. Louis, MO), astrocytes (GFAP 1:400; Sigma), and microglia (Iba1 1:200, Wako). Immunofluorescence was visualized with Alexa Fluor 488 or 594 antibodies (1:400; Molecular Probes).
Microscopy
Fluorescent images were analyzed with a Nikon Eclipse TE300 microscope (Nikon, Melville, NY). One hundred neurons were selected from a distribution of nine images per coverslip for analysis.
Neurite outgrowth assays
After exposure to glial conditioned media, E18 neurons were fixed in 4 % paraformaldehyde and immunostained with anti-βIII-tubulin (1:400). Neurites were visualized by F-actin with Rhodamine phalloidin (1:50; Molecular Probes, Carlsbad, CA). Images were analyzed for neurite length, density, and number by NeuronJ of ImageJ software; soma size was determined by the Neurphology plugin of ImageJ. Only neurons with neurites fully visible were analyzed. Neurite density was assayed as total βIII-tubulin fluorescence after skeletonizing. Axons were identified as the longest neurite [
33].
Image analysis
The olfactory sensory neuron (OSN) dendritic layer of the OE was assessed by NeuronJ plugin of ImageJ in 20 evenly spaced regions in the nasal septum and ethmoturbinates. The dendritic layer thickness was defined as the distance between the OSN cell body and the outer edge of the sensory dendrites in the nasal cavity.
Statistical analysis
GraphPad Prism Version 5 (Graph Pad, La Jolla, CA) was used. Single and multiple comparisons used Student’s t test (unpaired) and ANOVA/Tukey’s multiple comparison post-test, respectively. Level of significance alpha = 0.05.
Discussion
These studies further document the role of glial TNFα in neuroinflammatory responses to air pollution PM that modify neuronal function. In particular, we studied nPM, which are a subfraction of urban PM2.5 (“
Methods” section) that epidemiological studies have associated with neurodevelopmental dysfunctions from pre- and early childhood exposure [
37,
38]. Rodent models include exposure of pregnant rats to nPM, which altered neonatal neuronal maturation [
39] and exposure of early postnatal mice to ultrafine PM, which caused ventriculomegaly and glial activation [
22]. For inflammatory responses, we focused on TNFα because of its consistent elevation in rodent models of air pollution [
8,
10,
40‐
42] as well as in postmortem human brains from a highly polluted megacity [
15]. In vitro activities of nPM include induction of TNFα in mixed glia from cerebral cortex and reduced neurotrophic support by the CM of mixed glia exposed to nPM [
10]. We also document the stability of nPM activity to induce TNFα, in which the dose response was nearly identical, despite collection from the same site on different years.
We hypothesized that glial TNFα was a mediator of these CM effects because TNFα in vitro inhibits neurite outgrowth [
24,
34] with growth cone collapse [
43] and inhibits astrocytic neurotrophic support [
44]. Before further analysis of cerebral cortex glia, we investigated if TNFα induction by air pollution PM extended to the OE which is the initial site of exposure of inhaled air pollutants from which olfactory neurons project into the brain. Importantly, besides the acute inflammatory responses of TNFα and macrophage activation, the OE expresses high levels of phase I and phase II detoxifying enzymes, e.g., cytochrome P450 (CYP) isoforms and glutathione S-transferases (GST) [
45,
46], which may mediate detoxifying environmental pollutants.
We developed an ex vivo model for the initial impact of air pollution on olfactory neurons, in which the neonatal mouse nose is incubated with aqueous suspensions of nPM. During ex vivo incubation with nPM, the neonatal OE showed rapid shrinkage of the OSN dendritic layer concurrently with induction of TNFα and macrophage activation in the OE. We hypothesized that olfactory neuron dendritic regression was driven by TNFα from macrophages in the OE. This is supported by another model of olfactory damage, where TNFα was shown to inhibit OE regeneration [
47]. We further tested this hypothesis with primary glial cultures from the neonatal mouse cerebral cortex as discussed below.
In rodent models, nPM cross from the nose into the brain by undefined transport processes which are presumed to include the projections of OSN axons that synapse in the main olfactory bulb [
19,
20]. Studies with different artificial ultrafine PM observed that inhaled [
19] or nasally instilled [
20] PM reached the forebrain and cerebellum as well as the OB within 24 h [
48]. The passage of nPM from the nares beyond the OB into the posterior brain structures gives a rationale for using cerebral cortex glia as an experimental model for direct nPM exposure. Although astrocyte cell bodies were not detected in the OE, there still may be a role of astrocytic TNFα in the OB which has deep neuronal projections caudally into the brain.
To develop our observations of OE dendritic shrinkage, we further analyzed mechanisms of neuronal responses to nPM with a model of primary cultures of mixed glia and neurons from the cerebral cortex. We extended our observation that CM from nPM-exposed mixed glia inhibited neurite outgrowth [
10] by resolving cell type contributions. In subcultures from mixed glia, microglia contributed 60 % of the TNFα in CM, consistent with the greater inhibition of neurite outgrowth by CM from microglia. Similarly, the microglial CM caused more inhibition of neurite outgrowth and neurite density than the astrocyte CM. A primary role of microglia in nPM responses is also consistent with the low abundance of GFAP-immunopositive cells or processes in the OE, especially during development [
49]. The precise mechanism of nPM uptake in cells is not well defined but could include phagocytosis [
50] as well as direct diffusion [
51].
The role of TNFα in neurite outgrowth inhibition was further defined by suppressing TNFα expression with siRNA, by immunoblockade of TNFα, and by TNFR1 blockade, all of which restored neurite outgrowth to control levels. The restoration of axonal length by TNFα immunoblockade is also consistent with enhanced axonal regeneration by TNFα blockade after injury [
34]. Because these conditions did not consistently alter the total number of neurites or neuronal perikaryal size, they define an experimental model for effects of nPM on neuronal plasticity without major cell damage that could be useful for efficient screening of neuroprotective agents.
Several mechanisms may mediate the glial-derived TNFα influences on neurite outgrowth. Although TNFα has both cytosolic and transmembrane forms, we would not expect a significant role for transmembrane TNFα because the nPM-CM has negligible cell membrane content. Notably, of the two defined TNFRs, only blockade of TNFR1 rescued the nPM-CM effect. This specificity is consistent with the 20-fold higher affinity of TNFR1 (K
a) to soluble TNFα vs TNFR2 [
52‐
54]. TNFR1 activation is associated with reduced neuronal differentiation, as well as apoptosis, whereas TNFR2 is associated with neuroprotection and survival [
55]. Blocking TNFR1 may have improved neurite outgrowth by diminishing growth cone collapse (Fig.
7c) through reduction of CM TNFα signaling. The small GTPase RhoA mediates the TNFα inhibition of neurite outgrowth [
24], but mechanisms from receptor signaling to neurite outgrowth inhibition are less defined. RhoA activation by TNFα can cause growth cone collapse and attenuate neurite outgrowth [
24,
34,
56], but this process has not been directly linked to TNFR1/2 signaling [
23].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
HC carried out the experiments and data analysis and drafted the manuscript. DAD assisted with neuronal cultures. SH contributed to the collection, extraction, and chemical characterization of the nPM samples. CS contributed the nPM samples and designed the nPM collection. TEM participated in the design of studies and manuscript editing. CEF conceived the study, guided the experimental design, and edited the drafts. All authors read and approved the final manuscript.