The USEPA 610 PAHs mix, a mixture of 10 PAHs (10 μg/mL in acetonitrile) including fluoranthene (Flu), pyrene (Pyr), benz[a]anthracene (BaA), chrysene (Chr), benzo[b]fluoranthene (BbF), benzo[k]fluoranthene (BkF), benzo[a]pyrene (BaP), dibenz[a,h]anthracene (DBA), benzo[ghi]perylene (BghiP) and indeno[1,2,3-cd]pyrene (IDP) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Three internal standards for PAHs, pyrene-d₁₀ (Pyr-d₁₀), benzo[a]anthracene-d₁₂ (BaA-d₁₂) and benzo[a]pyrene-d₁₂ (BaP-d₁₂) were purchased from Wako Pure Chemicals (Osaka, Japan). 1,6-Dinitropyrene (1,6-DNP), 1,3-dinitropyrene (1,3-DNP), 1,8-dinitropyrene (1,8-DNP), 2-nitroanthracene (2-NA), 9-nitroanthracene (9-NA), 9-nitrophenathrene (9-NPh), 2-nitrofluorene (2-NF), 2-nitrofluoranthrene (2-NFR), 3-nitrofluoranthrene (3-NFR), 1-nitropyrene (1-NP), 7-nitrobenz[a]anthracene (7-NBaA), 6-nitrochrysene (6-NC) and 6-nitrobenz[a]pyrene (6-NBaP) were purchased from AccuStandard, Inc. (New Haven, CT, USA) (100 μg/mL in toluene). 1-Nitrofluoranthrene (1-NFR), 2-nitropyrene (2-NP), 1-nitroperylene (1-NPer) and 3-nitroperylene (3-NPer) were supplied from Chiron AS (Trondheim, Norway) (0.1 mg/mL in toluene), and 4-nitropyrene (4-NP) was from Tokyo Chemical Industry (Tokyo, Japan). 6-Nitrochrysene-d₁₁ (6NC-d₁₁) was for an internal standard for NPAH analysis and was purchased from Cambridge Isotope Lab. Inc. (Andover, MA, USA). Chemical structures of the analytes and their abbreviations are presented in Fig. 1. All solvents and other chemicals used were of analytical-reagent or HPLC grade. Water was obtained from a Milli-Q water purification system (Millipore, Bedford, MA, USA). The standard reference material of urban dust (SRM 1649b) was purchased from National Institute of Standards and Technology (NIST, Gaithersburg, MD, USA). The SRM sample is an atmospheric particulate material collected in an urban site with certified concentrations of some PAHs and NPAHs (NIST, 2016; Albinet et al., 2014; Schantz et al., 2012).

Fig. 1. 
Structures of the 10 PAHs and 18 NPAHs analyzed in this study.

A schematic diagram of the HPLC-FL system used for the simultaneous determination of PAHs and NPAHs is shown in Fig. 2. The system consists of 4 LC20AD pumps (Pump 1-4), a SIL-20AC auto sample injector, a degasser (DGU-20A5), a CMB-20A system controller and an integrator (LCsolution software), a CTO-20 AC column oven, a six-port switching valve, and a RF-20Axs fluorescence detector (All from Shimadzu, Kyoto, Japan). The injected sample was eluted through a clean-up column (Cosmosil, 5NPE, 150×4.6 mm i.d., 5 μm, Nacalai Tesque, Kyoto, Japan) with its guard column 1 (10×4.6 mm i.d.) and then NPAHs were reduced to their amino-derivatives by using a reduction column (NPpak-RS, 10×4.6 mm i.d., JASCO, Tokyo, Japan) at 80°C (0-15.4 min, switching valve position A). The mobile phase for the clean-up and reduction columns was ethanol/acetate buffer (pH 5.5) (95/5, v/v) at a flow rate of 0.2 mL/min. A fraction of the amino-derivatives and unchanged PAHs eluted from the reduction column with the mobile phase was mixed with 30 mM ascorbic acid through the guard column 2 (Asahipak ODP-50G-6A, 10×6.0 mm i.d., 5 μm, Shodex, Tokyo, Japan) at a flow rate of 1.6 mL/min and was then trapped on the concentration column (Spheri-5 RP-18, 30×4.6 mm i.d. 5 μm, Perkin Elmer, MA, USA) with a switching time of 15.4-33.0 min (position B). The concentrated fraction was passed through two separation columns (Inertsil ODS-P, 250×4.6 mm i.d., 5 μm, GL Sciences, Tokyo, Japan) with their guard column 3 (10×4.6 mm i.d.) in tandem (33.0-120.0 min, position A). All columns except the reduction column were maintained at 20°C. A gradient elution of the separation columns was performed using 10 mM imidazole buffer (pH 7.6) as eluent A and acetonitrile as eluent B. The gradient conditions (B concentration and flow rate) for the separation of amino-derivatives of NPAHs and unchanged PAHs were as follows: 0-15.4 min (B conc. 20%, 0.5 mL/min), 15.4-33.0 min (B conc. 65%, 0.5-0.8 mL/min), 33.0-65.0 min (B conc. 65-80%, 0.8-1.0 mL/min), 65.1-86.4 min (B conc. 80-100%, 1.0-1.8 mL/min), 86.5-120.0 min (B conc. 100%, 1.8 mL/min). Finally, the separated analytes were detected with their optimum excitation (Ex.) and emission (Em.) wavelengths by the dual-channel FL detector. The optimum wavelength used for each PAHs and NPAHs were as follows: Channel 1: 0-52.5 min (Ex. 369 nm, Em. 442 nm; 1,6-DNP), 52.5-56.1 min (Ex. 345 nm, Em. 430 nm; 9-NPh), 56.1-59.2 min (Ex. 260 nm, Em. 490 nm; 9NA), 59.2-63.2 min (Ex. 265 nm, Em. 488 nm; 2-NA, 2-NFR), 63.2-66.5 min (Ex. 360 nm, Em. 430 nm; 4-NP, 1-NP), 66.5-69.5 min (Ex. 338 nm, Em. 438 nm; 2-NP), 69.5-75.2 min (Ex. 300 nm, Em. 475 nm; 7-NBaA), 75.2-80.35 min (Ex. 227 nm, Em. 540 nm; 1-NPer), 80.35-82.75 min (Ex. 331 nm, Em. 391 nm; Pyr-d₁₀, Pyr), 82.75-84.3 min (Ex. 420 nm, Em. 475 nm; 6-NBaP), 84.3-88.7 min (Ex. 227 nm, Em. 540 nm; 3-NPer), 88.7-91.0 min (Ex. 265 nm, Em. 381 nm; Chr), 91.0-103.5 min (Ex. 295 nm, Em. 420 nm; BbF, BkF), 103.5-120 min (Ex. 268 nm, Em. 397 nm; DBA); Channel 2: 0-52.5 min (Ex. 395 nm, Em. 454 nm; 1,8-DNP, 1,3-DNP), 52.5-56.2 min (Ex. 285 nm, Em. 370 nm; 2NF), 56.2-63.0 min (Ex. 273 nm, Em. 437 nm; 1-NFR), 63.0-66.5 min (Ex. 300 nm, Em. 530 nm; 3-NFR), 66.5-78.0 min (Ex. 273 nm, Em. 437 nm; 2-NP, 6-NC-d₁₁, 6-NC), 78.0-82.75 min (Ex. 289 nm, Em. 450 nm; Flu), 82.75-86.0 min (Ex. 283 nm, Em. 513 nm; 6NBaP), 86.0-112.6 min (Ex. 264 nm, Em. 407 nm; BaA-d₁₂, BaA, BaP-d₁₂, BaP, BghiP), 112.6-120 min (Ex. 300 nm, Em. 500 nm; IDP).

Fig. 2. 
Schematic diagram of the developed HPLC-FL method.

For column evaluation, a simple HPLC-FL system was used to determine the retention times of the PAHs and NPAHs on the candidate clean-up columns. The system consisted of two HPLC pumps for gradient elution, two column ovens for the tested column, the reduction column and FL detector. Eight reversed-phase columns, 2 octadecyl (5C₁₈-MS-II and 5C₁₈-AR-II)-, phenylethyl (PE)-, pentafluorophenyl (PFP)-, nitrophenylethyl (NPE)-, naphthylethyl (πNAP)-, pyrenylethyl (PYE)- and pentabromobenzyl (PBr)-bonded silica columns (Cosmosil columns, 150×4.6 mm i.d., 5 μm, all from Nacalai tesque, Kyoto, Japan) were examined (Table 1). A gradient elution using water (eluent A) and methanol (eluent B) was carried out (B, 70-100% liner gradient for 60 min, 100% isocratic after 60 min) at a flow rate of 1.0 mL/min. To evaluate the performance of the columns, the retention factor (k) was considered to be a factor for parameter calculation. The switching index can be difiened by Eq. (1).

where k_max, k_min and k_mean are the maximum, minimum and mean of retention factors (k) of all target compounds, respectively. Using the index, the overlap and length of elution times for each column were evaluated.

PM_2.5 samples were collected on a quartz fiber filter (2500QAT-UP, Pall Life Sciences, Ann Arbor, MI, USA) by a high-volume air sampler (Model HV-700F, Shibata Sci. Tech., Saitama, Japan) with an impaction plate for a 50% cutoff point of 2.5 μm (PM_2.5) for 24 h at a flow rate of 1000 L/min. The PM_2.5 samples were collected at an urban site in Kanazawa, Japan from November 5 to 17, 2016 and they were analyzed to determine the accuracy and precision of the developed method. The filters were stored at -20°C until analysis. The commercially available urban dust (SRM 1649b) was from atmospheric particulate material collected in the Washington, DC area in 1976 and 1977 (NIST, 2016). After the addition of internal standards, a mixture of Pyr-d₁₀, BaA-d₁₂ and BaP-d₁₂ (25, 12 and 13 ng, respectively) for PAH quantification and 6-NC-d₁₁ (1.8 pg) for NPAH quantification, the samples were ultrasonically extracted with dichloromethane (DCM) for 15 min. One-fourth of the PM_2.5 filters was cut into small pieces and extracted with 75 mL of DCM. The SRM samples (3 mg of the powder) were extracted with 5 mL of DCM. The extraction procedure was repeated 3 times. After adding 30 μL of dimethylsulfoxide (DMSO) to the extract, the DCM in the extract was completely evaporated. The resulting DMSO solution was mixed with 270 μL of ethanol. Finally, the solution was filtered through a centrifugal filter (Ultrafree-MC, 0.45 μm Millipore) and then an aliquot (100 μL) of the solution was injected into the developed HPLC-FL system.

The developed method was validated with calibration curves, the limit of detection (LOD), the limit of quantification (LOQ), precision and accuracy. The concentrations of PAHs and NPAHs were quantified from the peak area ratios of the analytes to the deuterated internal standards. The internal standards used for determining PAHs and NPAHs were as follows: 6-NC-d₁₁ for all of NPAHs, Pyr-d₁₀ for Pyr and Flu, BaA-d₁₂ for BaA, Chr, BbF, and BkF, and BaP-d₁₂ for BaP, DBA, BghiP, and IDP. Calibration curves of each PAHs and NPAHs were prepared by plotting five points (n=5) between the lowest concentration of 0.01-1 μg/L for PAHs and 0.005-5 μg/L for NPAHs, and at the highest concentration of 500 μg/L for PAHs and 10 or 100 μg/L for NPAHs. The LOD and LOQ were determined from lowest concentration at which the signal-to-noise (S/N) ratio was higher than 3 and 10, respectively, with precision less than 15% through the entire treatment of spiked blank samples. The intra-day and inter-day accuracy and precision were examined by repetitive determination of the PM_2.5 samples spiked with standards at a constant concentration for each analyte, together with non-spiked samples. The accuracy was expressed as the ratio of the quantified concentration to that of the known concentration of the spiked analyte. The precision was calculated as the relative standard deviation (SRD, %) of the replicates. To evaluate the intra-day precision, the spiked samples and non-spiked samples were prepared four times per day. The inter-day precision was determined using independent experiments repeated on four consecutive days. All extracts were analyzed on the same day as the samples were extracted.

The two-dimensional HPLC system consists of clean-up, reduction, column-switching, separation and FL detection steps (Fig. 2). A high percentage of ethanol is necessary for the reduction step using the Pt/Rh column and acetonitrile decreases the reduction efficiency (Hayakawa et al., 2001). On the other hand, acetonitrile is an effective solvent for clear separation of all targets including similar isomers (Tang et al., 2005a) and to decrease column pressure. To eliminate opposing solvent effects, the two-dimensional system was able to switch solvent source to minimize the solvent effects of the 1st dimension. The clean-up column can be incorporated into the 1st dimension for a partial purification to remove hydrophilic matrix in a sample through the column-switching based on differences in the elution times from the column between the analytes and the sample matrix. Since PAHs and NPAHs have a wide range of hydrophobicities, ODS columns have little effect on removing substances that may interfere with the detection of analytes and require a long time for the elution of all analytes. This is especially true for low abundance NPAHs and consequently, the CL system required laborious pretreatments such as washing sample extracts with sodium hydroxide and sulfuric acid (Tang et al., 2003). Furthermore, our previous CL method required 58 min to trap the analytes on the concentration column and then 108 min to separate only NPAHs (total 166 min) (Tang et al., 2005a). The retention characteristics of 8 reversed-phase columns were evaluated to find a more effective column than the conventional ODS column. The characteristics of the stationary phases are listed in Table 1. The ideal characteristics of a clean-up column are a short elution band and large retention factors for all analytes. A shorter band can decrease the loading time to the concentration column and strong retention can increase the specificity of clean-up column. To satisfy these conditions, analytes with low logP values such as DNPs need to be retained with other interactions in addition to hydrophobic interaction, whereas the retention of the analytes with high logP values, such as IDP, needs to be suppressed.

The retention times of PAHs and NPAHs on each column were determined with methanol : water as the mobile phase, because of limitations in the reduction column and incompatibility between acetonitrile and π-π interaction (Snyder et al., 2004). The k values of the PAHs and NPAHs are listed in Table 2. The hydrophobicity of NPE, PFP and PBr phases is much smaller than that of ODS phases (5C₁₈-MS-II and 5C₁₈-AR-II) and similar to that of the PE phase (Kimata et al., 1992). Nevertheless, mean k values of the three phases for NPAHs were comparable to or higher than those of the ODS phases. A non-substituted PAH (Pyr), nitro-substituted PAHs (1-, 2- and 4-NPs) and dinitro-substituted PAHs (1,3-, 1,6- and 1,8-DNPs) were eluted from the 6 tested columns in the following order: Pyr<NPs<DNPs, showing a reversal of the elution order in the ODS columns. The strong retention of NPAHs on NPE and PFP phases compared to the PE phase indicated the presence of strong dipole-dipole interactions (Kimata et al., 1992). In particular, PAHs and NPAHs were strongly retained on PYE and PBr phases, indicating strong dispersive interactions between the aromatic species and the stationary phases in addition to their hydrophobic and π-π interactions (Turowski et al., 2001).

To evaluate the length of the elution band and the distribution of retention times, switching indexes were calculated for each column (Table 2). A small value indicates a short elution band and strong retention of the analytes. Although the PE column showed the smallest switching index (0.93), the retention of the analytes on the column was considerably weaker than the other columns. Taking into consideration the separation of the analytes from a hydrophilic matrix, the NPE column (switching index: 1.00) was selected as the clean-up column for the switching column system. Finally, ethanol/acetate buffer (pH 5.5, 95/5, v/v) at a flow rate of 0.2 mL/min was used for the clean-up column as the mobile phase, taking into account the conditions required for the reduction step (Hayakawa et al., 2001). The performance of the NPE column was maintained under the conditions because the switching index showed 1.05, the same value as observed in the conditions for the column evaluation. All the analytes were retained for over 15 min and eluted for 17.6 min (switching time: 15.4-33.0 min).

After column switching, 10 PAHs and 18 amino-derivatives of NPAHs were separated on the separation columns which consist of 2 polymeric-type ODS columns (4.6 mm i.d. ×250 each, 5 μm) in tandem. Fig. 3 shows typical chromatograms of a standard mixture of target PAHs, NPAHs and deuterated standards, which were all well separated by gradient elution. The reduced NPAHs were eluted from the columns faster than non-substituted PAHs. Three deuterated PAHs (Pyr-d₁₀, BaA-d₁₂, BaP-d₁₂) and the amino-derivative of 6-NC-d₁₁ were separated from the non-deuterated compounds with sufficient resolution (R_s>2.85). In general, stable isotope-labeled compounds are excellent internal standards for mass spectrometric detection, but not for optical detection methods such as FL detection. However, deuterated PAHs can be separated from the non-deuterated analogues with baseline resolution on polymeric-type ODS columns and have nearly the same fluorescence characteristics (Toriba et al., 2003). Furthermore, we have successfully applied deuterated PAHs, 1-NP and hydroxylated PAHs to HPLCFL methods for environmental and biological samples (Ohno et al., 2009; Toriba et al., 2007). Total analytical time for simultaneously determining PAHs and NPAHs was 120 min including 33 min for clean-up and reduction steps in the 1st dimension. After the 1st dimension, the 2nd dimension required 52 min and 82 min to separate NPAHs (elution time: 45.0-85.0 min) and PAHs (elution time: 79.0-115.0 min), respectively. The analysis time for NPAHs was reduced by half compared to the HPLC-CL method (Tang et al., 2005a) and was followed by PAH analysis.

Fig. 3. 
Representative standard chromatograms of PAHs and NPAHs measured by the developed HPLC-FL method. Injected amounts: Channel 1; 1,6-DNP, 750 pg; 9-NPh, 3 ng; 9-NA, 3 ng; 2-NA, 500 pg; 2-NFR, 1 ng; 4-NP, 12.5; 1-NP, 500 pg; 7-NBaA, 1 ng; 1-Nper, 10 ng; Pry-d₁₀, 11 ng; Pyr, 1 ng; 3-NPer, 10 ng; Chr, 1 ng; BbF, 1 ng; BkF, 1 ng; DBA, 1 ng; Channel 2; 1,8-DNP, 1 ng; 1,3-DNP, 1 ng; 2-NF, 500 pg; 1-NFR, 1 ng; 3-NFR, 5 ng; 2-NP, 4 ng; 6-NC-d₁₁ 6 ng; 6-NC, 1 ng; Flu, 1 ng; 6-NBaP, 6 ng; BaA-d₁₂, 6 ng; BaA, 1 ng; BaP-d₁₂, 7 ng; BaP, 1 ng; BghiP, 1 ng; IDP, 1 ng.

Moreover, we focused on the separation of nitrofluoranthrenes and nitropyrenes which display similar chromatographic behavior on reversed-phase columns. Specifically, 2-NFR and 2-NP are well-known secondary products formed via atmospheric reactions of their parent PAHs and have been frequently found in PM samples (Tang et al., 2014; Hayakawa et al., 2011). It is essential to separate these two compounds from the other detectable isomers (1-NFR, 3-NFR, 1-NP and 4-NP) in real samples. The 6 compounds were successfully separated on the 2 ODS columns in tandem which required a high number of theoretical plates (Fig. 3). Dual-channel fluorescence detection was used to avoid frequent changes of the detection wavelengths or separate co-eluted or closely eluting peaks. The two-channel program was set based on optimal excitation and emission wavelengths of the analytes (Šoustek et al., 2008; Cvačka et al., 1998). Remarkably, 6-NC and 7-NBaA were successfully separated by their specific excitation and emission wavelengths, not by the separation columns as they had identical retention times (Fig. 3). The signal arising from one analyte was not detected on the other channel. These results suggest that the proposed HPLC-FL system is suitable for the identification of all target compounds.

The concentrations of the analytes which establish S/N ratio of 3:1 and 10:1 were considered to be the LODs and LOQs, respectively. The LODs and LOQs for PAHs and NPAHs are listed in Table 3, indicating that the developed method is sensitive enough to measure the analytes in the PM samples (1.1 pg/m³ for 3-NPer which showed the lowest sensitivity). Compared to previous HPLC-CL method (Tang et al., 2005a), the LODs of 2-NF, 9-NPhe, 2-NA, 1-NP, 2-NP 7-NBaA and 6-NC were substantially improved, whereas the sensitivities of 9-NA, 1- and 3-NPer were decreased. The sensitivities of the PAHs were higher than those previously reported because of instrumental improvement (Toriba et al., 2003). Calibration curves were constructed from the peak area ratio of the analyte to deuterated internal standards and good linearity (r²>0.9999 for PAHs and >0.99 for NPAHs) was observed for all calibration curves, as shown in Table 3.

The extraction method of PAHs and NPAHs in PM samples involved ultrasonic extraction with DCM, evaporation and redissolving steps, and then the crude extract was directly injected to the HPLC system. The analysis of the crude extracts is difficult to perform with GC-MS. The recoveries of the deuterated internal standards (Pyr-d₁₀, BaA-d₁₂, BaP-d₁₂ and 6-NC-d₁₁) were 80-119% through the entire pretreatment. Known amounts of PAHs and NPAHs were added to PM_2.5 samples, and their quantification accuracy and precision were evaluated (Tables 4 and 5). The accuracies were 86-98% and 87-109% for the PAHs and NPAHs, respectively. The precision was good with a RSD of 15% or less for all analytes. Both intra- and inter-day accuracy and precision were satisfactory for the simultaneous determination of all PAHs and NPAHs in the PM_2.5 samples. Fig. 4 shows typical chromatograms resulting from a crude extract of the spiked PM_2.5 samples. The sample matrix did not interfere with the identification and quantification of the analytes in the chromatograms. The results suggest effective fractionation of the analytes from the sample matrix by the clean-up column and column-switching in the 1st dimension.

Fig. 4. 
Representative chromatograms resulting from a crude extract of the spiked PM_2.5 samples.

The concentrations of PAHs in the suburban PM_2.5 samples used for the validation study (the concentrations of non-spiked samples in Table 4) were slightly higher than those in a background site located in the Noto peninsula without any major sources near the station (Tang et al., 2014). In contrast, the PAH levels were lower than samples collected at a road side in Kanazawa city (Tang et al., 2005b) and the other Japanese cities (Hayakawa et al., 2016b; Naser et al., 2008). PAHs are commonly found in PM_2.5 fractions and PAH concentrations have been directly correlated with PM_2.5 levels (Naser et al., 2008). The PAH concentrations in this study were reasonable, with respect to the PM_2.5 levels of 7.7±2.9 μg/m³ measured during the sampling period. Among the NPAHs, 1,3-DNP, 2-NA and NPAHs having higher molecular weight were not detected in the PM_2.5 samples (Table 5). The concentration of 1-NP in this study was significantly lower than other sites, including the background site (Hayakawa et al., 2016b; Tang et al., 2014; Naser et al., 2008), indicating little emission source related to automobile exhaust around the site. The 1-NP contribution to NPAHs has substantially decreased throughout Japan in recent years, in conjunction with government regulations regarding automobile emissions (Kojima et al., 2010).

Using the developed HPLC-FL method, we also quantified PAHs and NPAHs in an urban dust standard reference material (SRM1649b). Fig. 5 shows representative chromatograms resulting from the crude extract of SRM 1649b. The obtained concentrations of 10 PAHs and 18 NPAHs were compared with certified concentrations by NIST and data reported from other studies (Table 6). The quantified PAH concentrations showed good agreement with the certified values in the range of 90-105% (NIST, 2016). Five NPAHs were highly consistent with the certified NPAH values (92-117%), except for 9-NA, 9-NPh and 3-NFR. Several compounds showed similar concentrations with other literature data (Albinet et al., 2014; Schantz et al., 2012). There were large variations in the reported concentrations of NPAHs from literature data (Table 6), and these differences could arise from all of the procedure of the analytical protocols employed. The NPAH concentrations determined by pressurized solvent extraction (PLE) reported by Schantz et al. (2012) were very close to those in the certificate. On the other hand, Albinet et al. (2014) used two extraction methods, PLE and the Quick Easy Cheap Effective Rugged and Safe (QuEChERS) approaches for the NPAHs in SRM1649b. They reported concentrations of numerous NPAHs which included analytes not reported in the certificate with several other being different from the certificate data and having variations between the extraction methods. The variation of the quantified data may be from the difference among general Soxhlet extraction, PLE and sonication.

Fig. 5. 
Representative chromatograms resulting from the crude extract of SRM 1649b.