Overview of publications relevant for regulatory risk assessment
About 52% (246) and 38% (58) of the original 468 and 152 literature search results for metal-containing NPs and nanofibers, respectively, were excluded because either the study was not in the scope of this review, the full-text access was not available, or results were already covered by the other search terms. The substance (S) score was finally evaluated in 137 in vitro studies and 85 in vivo studies on metal-containing NPs, and in 57 in vitro studies and 37 in vivo studies on nanofibers, which fitted the scope of the present review.
The quantity of studies on nanofibers or metal-containing NPs evaluated in each step of the quality assessment is displayed in Fig. 1. Within all groups, by far the largest number of studies were rejected due to an incomplete PC characterization (S score) of the tested NM. For metal-containing NPs, a slightly larger proportion of studies (69% and 65% for in vitro and in vivo, respectively) were rejected on this basis compared to nanofibers (61% and 51%, respectively). According to the GUIDEnano reliability evaluation (K score), 6% and 10% of the in vitro and in vivo studies on metal-containing NPs, respectively, were deemed unreliable, whereas all except one (2%) nanofiber studies were considered reliable. The differences may be due to the considerably smaller number of publications on nanofibers, but as the evaluations for each type of NM were conducted by a different evaluator, subjective interpretation of the evaluation criteria cannot be entirely ruled out. The proportion of in vitro studies, which passed all stages of the evaluation was similar for both metal-containing NPs and nanofibers, 15% and 12%, respectively. In case of the in vivo studies, 18% of the metal-containing NPs and 33% of the nanofiber publications were finally accepted.
Results of the quality evaluation based on the GUIDEnano quality approach
The S score was the most limiting of the evaluation steps (Fig. 1). Acceptable NM characterization was only available in 31% and 39% of the in vitro studies, and 35% and 49% of the in vivo studies, for metal-containing NPs and nanofibers, respectively. The information was available either in the publications themselves or other clearly indicated sources. Failing to fulfill the acceptable substance characterization mostly stemmed from failing at least one of the ‘red questions’ (obligatory criteria), and less frequently from failing to fulfill the minimum points required in the scoring. Up to 65% and 28% of the in vitro studies, and 58% and 29% of the in vivo studies, for metal-containing NPs and nanofibers, respectively, failed at least one of the red questions. This is in line with the findings by Fernández-Cruz et al. [16], who found insufficient characterization of the tested NMs to be the principal weakness in the toxicity studies evaluated during the development of the GUIDEnano approach.
The most common shortcomings in the substance characterization were failures to provide the purity of the NM and the size distribution during the exposure, which is also in agreement with the GUIDEnano’s evaluation [16]. In many papers which had an acceptable S score in this evaluation, the purity of a well-identified commercial NM could be found from the supplier's website. This was deemed acceptable in the evaluation, although it should be noted that information on websites often becomes unavailable over time, and quite detailed information (e.g., a product and a lot number) is often required to identify the exact product from the supplier's website. Other reasons for failing this question included previously published characterization data that was either mis-referenced or the previous publication was not available from typical sources. Purity with respect to the nanofibers was a much more obtainable criterion compared with metal-containing NPs. The characterization of carbon nanotubes (CNTs) is fundamentally linked to their carbon purity and is an extremely common feature of their published PC feature set. The question of purity becomes especially pertinent when discriminating e.g., between CNTs NM400 and NM401 which are extremely alike with the exception of higher metal impurity content in the NM400, an artifact of the catalysis. Nevertheless, it is highly recommended to provide a summary of the material characteristics provided by previous publications, the substance supplier’s website, or the certificate of analysis in the same publication with the genotoxicity test results.
As concerns the assessment of the size distribution during the exposure, dynamic light scattering (DLS) is one of the most used methods for characterizing NPs in the exposure medium. However, it may give an inaccurate size distribution for non-spherical materials. Transmission electron microscopy (TEM) is a more suitable method for accurate measurements of the particle size for nanofibers, although it does not provide a good estimation of the size distribution in the sample. In some conditions, particle size in medium changes with concentration [44]. Although measuring particle size from exposure medium in all the tested doses is not required by the S score nor used as a criterion in our evaluation, a dose-dependent particle size may contribute to a better interpretation of results.
Another red question with which we encountered some shortcomings concerned the shape of the primary particle, which was not always well described for the metal-containing NPs. Ideally this information should be clearly given, also for spherical or amorphous NMs. In this evaluation, however, we accepted TEM images as proof of particle shape.
After evaluating the completeness of material characterization, a total of 64 in vitro studies (42 and 22 metal-containing NPs and nanofibres publications, respectively) and 48 in vivo studies (30 and 18 metal-containing NPs and nanofibres publications, respectively) were further assessed for the quality.
The reliability of the studies was evaluated by the K score (Fig. 1). For nanofibers, only one in vitro study was considered unreliable due to failing at least one of the essential criteria. In the case of metal-containing NPs, most of the studies with a complete material characterization—81% in vitro and 74% in vivo—were also considered reliable. Failing at least one of the ‘red questions’ was the reason for not succeeding in passing the K score in all the cases. In the in vitro studies, the most common shortcomings were failures to clearly describe the source of the test system, and the study endpoints and methods. Among the in vivo studies, the most critical shortcoming was the lack of positive controls. Our findings comfort with those of Fernández-Cruz et al. [16] who concluded that, in general, peer-reviewed publications complied with the majority of the questions included in the K score. Nevertheless, and as mentioned before for material characterization, also for the description of methods it is generally not advisable to refer to sources that may change or become unavailable over time. Instead, all relevant details should be cited in the same publication with the genotoxicity data to ensure accessibility.
Results of the assay-specific evaluation
A total of 55 in vitro studies (34 and 21 metal-containing NPs and nanofibres publications, respectively) and 39 in vivo studies (21 and 18 metal-containing NPs and nanofibres publications, respectively) were considered reliable in the K score-based evaluation. These studies were further assessed according to the assay-specific criteria detailed in Tables 2 and 3.
In vitro assays
From the total of 55 in vitro publications, the number of reported MN, CA and gene mutation assays were 29, 5 and 6 for metal-containing NPs, and 14, 7 and 2 for nanofibers, respectively. In some of the publications, more than one type of assay was used. Figure 2 summarizes the proportion of studies that successfully fulfilled the assay-specific criteria listed in Table 2. About half of the studies performed with the MN and gene mutations assays complied with the criteria. However, the proportion was much lower when using the CA assay.
A breakdown of the in vitro assay quality evaluation results according to the obligatory criteria is shown in Fig. 3. The most common shortcoming for all types of materials leading to exclusion among in vitro genotoxicity tests was the lacking or inadequate concurrent cytotoxicity measurement. In addition, studies on nanofibers had a markedly larger number of shortcomings concerning the treatment schedule, study design and adherence to sample sizes or number of replicates or experiments recommended by the OECD guidelines compared to studies on metal-containing NPs.
The purpose of a concurrent cytotoxicity measurement is to ascertain a relevant dose range for the genotoxicity analysis. According to the OECD guidelines, a minimum of three (MN and CA assays), or four (gene mutation tests) doses should be tested up to a highest test concentration aiming at 50–60% cytotoxicity in case of MN and CA assays, and 10–20% survival in case of the gene mutation tests (criteria 9, 15 and 21 of Table 2). Genotoxicity analysis at excessively toxic concentrations could lead to false positive outcomes [5, 24]. On the other hand, testing only doses with too low toxicity may prevent or underestimate the detection of the genotoxic potential of the NMs [5, 45, 46]. A concurrent cytotoxicity assessment using a recommended cytotoxicity parameter based on cell proliferation is required (as specified in the criteria 6, 13 and 19 of Table 2). Assessing cytotoxicity in the same experiment as the genotoxicity predictors is especially important when testing NMs as some of them exhibit poor repeatability between dispersions.
The treatment schedule could be considered clearly inadequate for detection of direct genotoxicity in 20% and up to 70% of the evaluated genotoxicity tests for metal-containing NPs and nanofibers, respectively (Fig. 3). A much larger proportion of the studies did not describe the exposure time relative to the cell cycle or include any information about the length of the cell cycle in the chosen model system. In this evaluation the treatment time was considered acceptable if it was in the range of normal average cell cycle length (15–24 h). The test result, however, could also be considered acceptable by expert judgement regardless of the inadequate treatment time in case of clearly positive results.
The breakdown of the in vitro genotoxicity assay quality evaluation results according to the non-obligatory criteria is shown in Fig. 4. The main shortcoming for these criteria was the lack of confirmation of cellular uptake. As mentioned above, in order to come in contact with the genetic material, NMs should be internalized by cells. In this evaluation we did not consider assessment of particle uptake a mandatory requirement but used it as supporting information for the purpose of compensating other shortcomings and weighing the relevance of negative test results. An example of a well performed investigation of cellular uptake in the test system is the study by Di Bucchianico et al. [47]. These authors tested the time and dose dependence of Ni and NiO NPs uptake by inductively coupled plasma mass spectrometry (ICP-MS) and confirmed the presence of Ni in particle form at the last analyzed time point by TEM. They found that the uptake was rapid and dose dependent in the chosen test system. However, the same cannot be assumed for other cell lines, NMs, or methods of dispersion, and thus, especially in the case of negative genotoxicity test results, it is important that the study describes the capability of the cell line to internalize the material at the same time points and in the same conditions in which the genotoxicity test was performed.
Almost half of the assays for metal-containing NPs that included a concurrent cytotoxicity measurement failed to choose an appropriate dose range. In addition, in one fourth of the assays showing a positive outcome, the positive result was associated to excessive toxic doses (Fig. 4). As explained above, covering an appropriate dose range is critical for a proper interpretation of the genotoxicity outcomes. Some of the evaluated studies did only analyze non-toxic or very low cytotoxicity doses. On the other hand, others did not include enough low toxicity doses, which may have enabled a more reliable interpretation of the positive test results. We acknowledge fulfilling the regulatory requirements for an acceptable dose range with NMs may be challenging due to methodological limitations. At high doses heavy agglomeration of the test material may occur, making the test material effectively non-nano, or microscopical analysis can be hindered by material agglomerates covering the cell surface. However, in this case a clear description of the limitations and justification for the tested dose range should be given.
In vivo assays
From the total of 39 in vivo publications, the number of reported MN, CA, comet and gene mutation (Pig-a) assays were 16, 7, 20 and 3 for metal-containing NPs, and 6, 2, 14 and 0 for nanofibers, respectively. In most of the publications, more than one type of assay was used. Figure 5 summarizes the proportion of studies that successfully fulfilled the assay-specific criteria listed in Table 3. All the studies performed with the CA assay, as well as most of those using the MN assay, complied with the criteria. However, almost half of the studies using the comet assay did not. On the other hand, none of the studies involving the erythrocyte Pig-a gene mutation assay were acceptable.
A breakdown of the in vivo genotoxicity assay quality evaluation results according to the obligatory criteria is shown in Fig. 6. The most common shortcoming was a sample size that was smaller than recommended by the current OECD guidelines (sample size was revised in the 2014 version), or a draft of such guideline in case of the Pig-a assay. However, in the case of MN and CA assays, the sample size often followed the 1997 version of the OECD guidelines. In this case the test could be considered acceptable by expert judgement if the negative controls reached sufficient levels to enable reliable analysis and the test result was positive. A similar approach was applied regarding the concurrent toxicity measurement, where the only complaint was the cell number analyzed for the cytotoxicity parameter. Rare events such as chromosomal aberrations often require analysis of a larger number of metaphases (200 cells per animal) compared to what is recommended by the older (1997) version of the OECD guideline (100 cells per animal), however the data was in most cases useful even if the older version of the guideline was followed. A design with too low sample size may, however, pose a problem with interpretation of negative or weak positive results as lack of statistical power compromises the sensitivity of the test [48].
Another shortcoming concerned to the treatment and sampling schedules. A justification of a treatment schedule should optimally rely on toxicokinetic studies, which confirm the presence of the test material in the target organ at a given time point and, on the same time, take into account the transient nature of the measured phenomenon. A study using unjustified treatment schedules could be acceptable only if the test result is positive. However, in this evaluation we accepted studies that followed the OECD recommendations, although their suitability for NMs can be sometimes questionable as it may take a longer time for NMs to reach the target organ compared with soluble chemicals [5]. Especially in the case of comet assay, which is based on DNA damage that is usually repaired within hours, the bio-persistence of the nanomaterial should be confirmed, if samples are collected later than recommended.
The breakdown of the in vivo genotoxicity assay quality evaluation results according to the non-obligatory criteria is shown in Fig. 7. The main shortcoming for these criteria was the failure of confirming the presence of the material in the target tissue that, as commented in the previous paragraph, is necessary for the correct interpretation of negative outcomes. For instance, in the case of metal-containing NPs, only 3 out of 21 studies evaluated the biodistribution in all the target tissues where genotoxicity was measured, whereas 13 included some biodistribution data, but not all target tissues were measured, and 5 included no biodistribution data. Although 17 out of 21 studies included assessment of systemic genotoxicity, accumulation in bone marrow was measured in only 3 studies. In 9 out of these 17 studies, peripheral blood was used as an indicator of systemic distribution. However, as bone marrow is the target organ for the currently accepted tests for systemic genotoxicity and there may be significantly less NM available in the bone marrow compared to peripheral blood, measuring accumulation in blood may not be sufficient. In one of the studies that did consider the biodistribution in bone marrow, oral exposure to different sizes of Ag NPs led to minimal silver accumulation in the blood and especially in the bone marrow compared to other organs [49]. Unfortunately, as pointed out before, an appropriate TG for assessing the toxicokinetics of NMs is still in development [39].
Regarding the use of an adequate route of exposure, most of the studies were considered as fulfilling this criterion. Although the route of exposure should be chosen based on the realistic human exposure, it is worth noting that, according to the recommendations of the European Chemicals Agency for the safety assessment of NMs, studies are recommended to be performed via the respiratory route [36]. Only two studies, one by inhalation and the other by intratracheal instillation, explored the respiratory exposure to NMs in the case of metal-containing NPs, whereas a total of 11 studies (one by inhalation) did it in the case of nanofibers.
Outcomes from the qualified publications
For the metal-containing NPs, 20 out of 34 in vitro publications (59%) and 15 out of 21 in vivo papers (71%) that fitted the requirements of the GUIDEnano quality assessment, also passed the assay-specific criteria put forward by the authors of the present manuscript. With respect to the nanofiber publications, the corresponding numbers were 7 out of 21 in vitro (33%) and 12 out of 18 in vivo (71%) papers. This elevated success rate observed in vivo may therefore be a result of the stricter test guidance which applies to toxicology testing in animal models as opposed to in vitro cultures. When considering the total number of papers evaluated in this study (Fig. 1), only 20 (15%) of the 137 in vitro publications and 15 (18%) of the 85 in vivo publications were considered of acceptable quality from a regulatory perspective for the metal-containing NPs. In the case of nanofibers, the corresponding numbers were 7 out of 57 evaluated in vitro publications (12%), and 12 out of 37 in vivo publications (32%).
The qualified publications covered a broad range of metal-containing NPs, including only 1–2 publications per chemical composition except for TiO2 and Ag NPs, for which more results were available. Conversely, the majority of the nanofiber studies concerned single- or multi-walled carbon nanotubes (SWCNT and MWCNT), and only one publication among those assessing other fibers (including nanocellulose, carbon nanofibers, imogolite, europium nanorods, cotton fibers, and graphene nanoribbons) passed the quality assessment. Most of the test results were positive, this may reflect true NM-induced genotoxicity, but may also be the result of publication bias, as negative results are usually more difficult to get published. Furthermore, more restrictive quality criteria were applied to studies reporting negative results, as shown in the previous sections (e.g., more justification for treatment schedule, cellular uptake or biodistribution is required in case of negative results).
TiO2 nanoparticles
Table S4 (Additional file 2: Table S4) summarizes the in vitro studies performed with different types of TiO2 NPs. The results were not consistent as different types of NPs and cell systems were used. TiO2 P25 AEROXIDE (also known as JRC NM-105), a 15–30 nm TiO2 anatase/rutile that was used as a benchmark NP in many studies, produced inconsistent results in different cell lines. P25 AEROXIDE induced a statistically significant increase in MN at 20, 50 and 100 µg/ml compared with the untreated cells, together with a significant dose–response, in bronchial epithelial BEAS-2B cells, but only when using serum-containing medium [44]. Interestingly, these authors also found that the genotoxicity results were highly dependent on the quality of dispersion. P25 AEROXIDE also induced a significant increase of MN in mouse Balb/3T3 fibroblasts, but only at the lowest tested concentration (10 µg/cm2) [50]. Conversely, the same material did not increase the frequency of MN in human TK6 lymphoblasts and human lymphocytes [51]. However, cellular internalization was not confirmed in the latter study. Di Bucchianico et al. [52] tested three types of TiO2 with different particle size and crystalline structure. Both 5–8 nm anatase (NM-100) and 22–28 nm rutile induced a significant increase of MN at 1 and 1–5 µg/ml, respectively. However, a larger 50–150 nm anatase particle (NM-103), which has a tendency of forming large aggregates, produced a negative result regardless of confirmed particle uptake. However, as recognized by the authors, cellular uptake was only assessed by flow cytometry side scatter, making impossible to distinguish whether the particles have been internalized or attached on the surface of the cells. Significant increase of MN was also induced by 50 nm TiO2 anatase in the epidermoid carcinoma cell line A431 [53]. Negative results were obtained with the in vitro cytokinesis-blocked NM (CBMN) assay in Caco-2 cells when testing 20–60 nm TiO2 anatase [54]; however, the study did not include an assessment of cellular uptake.
One study investigated the mutagenicity of < 25 nm TiO2 anatase by using the CA assay after 24, 48 and 72 h culture of human lymphocytes [55]. A significant increase of aberrations, together with a significant dose–response, was reported at 48 h culture.
TiO2 was the only metal-containing nanoparticle for which we found acceptable in vivo data for the respiratory route (Additional file 2: Table S5). Sprague–Dawley rats were repeatedly intratracheally instilled with P25 AEROXIDE, resulting in a significant increase of peripheral blood micronucleated erythrocytes at 35 days post-administration [56]. On the other hand, the inhalation exposure of male C57BL/6 J mice with 21 nm TiO2 anatase/brookite for 5 days, 4 h/d, also increased the frequency of micronuclei in peripheral blood erythrocytes [57]. Both studies also analyzed the induction of DNA damage by the comet assay. However, this assay was not considered acceptable in any of the studies due to small sample size and lacking cytotoxicity indicator. A third study with acceptable comet assay data found a significant increase of DNA damage in the liver cells of male Swiss albino mice after a 14-day repeated oral exposure to 10–100 mg/kg of 20–50 nm TiO2 anatase [58]. Interesting, the results were also positive when applying the enzymatic (Fpg) version of the comet assay, indicating oxidative DNA damage.
Based on the above results, certain forms of TiO2 NPs seem to have mutagenic potential in vivo. However, it is unclear whether these effects are only caused by secondary mechanisms of action, as the outcomes of the in vitro assays – which can only detect primary mechanisms [38]—are contradictory among particles and cellular systems. Interestingly, a recent scientific opinion of the European Food Safety Agency [59] concluded that a concern for genotoxicity of TiO2 particles that may be present in the food additive E 171 cannot be ruled out.
Ag nanoparticles
The in vitro assay quality evaluation yielded four acceptable publications on Ag, including data on pristine, citrate-coated and polyvinylpyrrolidone (PVP)-coated Ag NPs (Additional file 2: Table S6). The results appeared to depend more on the primary size than on the coating. In TK6 lymphoblasts, the MN assay results were positive for small (≤ 20 nm) Ag particles regardless of the coating, but negative or positive only at cytotoxic doses for larger 50–100 nm Ag particles [60, 61]. In L5178Y mouse lymphoma cells, on the other hand, only citrate-coated 20 nm Ag NPs gave a clear positive response, whereas with the PVP-coated and larger 50–100 nm particles, the statistically significant positive responses coincided with significant cytotoxicity [60]. In accordance with the previous, PVP-coated 42.5 nm particles also gave a negative result in bronchial epithelial BEAS-2B cells [62]. The mouse lymphoma assay results were either negative (pristine particles) or positive only at doses which also exhibit significant cytotoxicity (coated particles) [60, 63]. However, the two consistently negative studies did not confirm cellular uptake [62, 63].
In the two in vivo studies on Ag, identified as acceptable in our evaluation, different sizes of pristine, PVP-coated and silica-coated Ag particles were tested by the peripheral blood MN test (Additional file 2: Table S7). Boudreau et al. found no systemic genotoxicity, although the accumulation of Ag NPs in bone marrow and blood after oral gavage was smaller compared to other organs [49]. Another study by Li et al. [64] found no genotoxicity after intravenous administration of PVP- and silica-coated Ag particles. However, acceptable data was limited and only available for systemic genotoxicity. As the in vitro studies show genotoxic potential may exist in some conditions, more in vivo evidence is needed.
Other metal-containing nanoparticles
With the exceptions of TiO2 and Ag NPs, mainly 1–2 studies per chemical composition were found acceptable in the assay quality evaluation. These studies reported positive genotoxicity results with Co3O4, Cu–Zn alloy, Fe3O4, Ni, NiO, ZnO, a variety of coated quantum dots (QD), W, and WC–Co in vitro (Additional file 2: Table S8) and with CeO2, Cr2O3, MgO, and MnO2, WO3, and Y2O3 in vivo (Additional file 2: Table S9). Only one in vivo study with Fe2O3 reported negative results in bone marrow CA assay and peripheral blood comet assay [65]. However, only small amounts of material were found in bone marrow compared to liver, spleen, kidney, and heart, and thus testing these other organs would have been warranted.
For ZnO we found two acceptable in vitro MN tests, both of which, however, had some limitations. Senapati et al. [66] observed cellular uptake and reported positive results when treating THP-1 monocytes for 3 h, and thus, the study was considered acceptable despite the short treatment time. Zijno et al. [54] also concluded positive results in intestinal epithelial Caco-2 cells, but as the MN test only gave a positive result at highly cytotoxic doses and dose response was not tested, the result should be interpreted with care. We found no acceptable in vivo studies with ZnO NPs.
NiO NPs were tested in one study with BEAS-2B bronchial epithelial cells in vitro [47] and one oral gavage study in vivo [67]. In both studies, NiO NPs exhibited genotoxicity. In the in vivo study, NiO NPs were found in all the tested organs of female albino Wistar rats, and both local and systemic genotoxicity was systematically observed in the comet assay of peripheral blood leucocytes, liver and kidney, in the erythrocyte MN assay, and in the bone marrow CA assay [67]. As for the rest of the in vivo studies on metal-containing NPs we found acceptable in this evaluation, there was no overlap with the NMs tested in vitro.
For tungsten oxide (WO3) and yttrium oxide (Y2O3) NPs, two acceptable studies with positive genotoxicity results for each material were found, all conducted by the same research group. These studies included a single oral exposure with female Wistar rats, and a 28-day repeated oral exposure study with both male and female Wistar rats. Comet and MN assays for liver and peripheral blood erythrocytes, respectively, were all positive in rats treated with a single dose of Y2O3, and both single and repeated dose of WO3 at the highest tested dose (1000 mg/kg body weight) [68,69,70]. In rats treated with a repeated dose of Y2O3, all tests were positive already at 120–480 mg/kg body weight [71].
Based on the above results, many of the assessed NPs seem to be able to induce genotoxicity by primary mechanisms. In the case of ZnO, the partial overlap between genotoxic and cytotoxic doses [72], may affect the outcome of the assays. On the other hand, an occupational exposure limit (OEL) was set for nickel compounds (including NiO) as it was considered that there is a mode-of-action based threshold for these genotoxic carcinogens [73]. However, this OEL is not applicable to nano-sized particles. Although most of the NMs showing primary genotoxicity are assumed to act though indirect mechanisms (not involving direct interaction with the DNA molecule) [3, 13, 37], which can have a thresholded response [74], it is unclear whether the cut off values are similar to those of their counterpart particles.
Nanofibers
As concerns the in vitro publications, each of the 7 papers which qualified was investigating CNTs as the test NM (Additional file 2: Table S10). From these 7 publications, 6 focused upon the in vitro CBMN assay. Interestingly, in each occasion of the in vitro CBMN assay being utilized, the test material induced a positive response for at least one concentration. For those publications, cellular uptake was confirmed in two studies, the first of which in 2013 by Manshian and colleagues [75]. Manshian et al. reported significant point mutations with the Hprt assay as well as primary genotoxicity in MCL-5 and BEAS-2B cells respectively, following exposure to three variants of SWCNTs. These CNTs differed in length primarily, ranging from 400 to 800 nm, 1 to 3 µm and 5 to 30 µm. These CNTs were exposed to the cells over a range of concentrations up to and including 100 µg/ml.
The use of BEAS-2B cells to detect genotoxicity following CNT exposures in vitro was investigated in the work by Catalan et al. [76] and Louro and colleagues [77]. Catalan et al. [76], reported no statistically significant effects for the in vitro CBMN assay following 5, 10 and 100 µg/cm2 exposures. Louro and colleagues [77] conversely, showed significant data in the in vitro CBMN assay although in the human alveolar epithelial (A549) cell type. This did require fairly high concentrations of NM401 and NM402 however (with NM400 and NM403 inducing no genotoxicity), with the concentration range extending up to 150 µg/cm2 [77]. Tavares et al. [78], and Catalan et al.[55], both utilizing human blood-derived lymphocytes were both able to demonstrate significant DNA damage. Firstly, Tavares and colleagues showed that MWCNTs at just 15 µg/ml induced significant chromosomal damage using the in vitro CBMN assay, however this proved to be CNT-type dependent [78]. Catalan and colleagues reported that both SWCNTs and MWCNTs between 6 and 300 µg/ml were capable of inducing significant increase of CAs in human blood-derived lymphocytes. These effects appeared to be time dependent as the significant data findings were found following 48- and 72-h exposures [55].
For the in vivo publications which successfully passed the assay-specific criteria (Additional file 2: Table S11), only one publication focused on a material other than CNTs. Catalan et al. [79] investigated the effects of nano fibrillated cellulose (NFC) over a range of 10, 20, 40, 80, 200 µg/mouse, delivered via pharyngeal aspiration into female C57BI/6 mice. The authors then utilized the in vivo comet assay and the MN assay in bone marrow erythrocytes. From the investigation the authors reported dose-dependent accumulation of NFC in the bronchi and contained within macrophages. This corresponded to significant DNA damage in the comet assay, however negative findings were reported in the MN assay [79]. The remaining 11 in vivo publications which passed the assay-specific guidance all focused upon either SWCNTs or MWCNTs. From these publications, the in vivo comet assay proved to be a common technique employed to ascertain CNT genotoxicity. From the in vivo comet data, three publications reported no significant findings whereas 7 reported at least one significant dose which induced a genotoxic response. Of the three negative studies, Pothmann et al. [80], Christophersen et al. [81] and Honda et al. [82] all reported no significant results. All three studies utilized different rodent models, different doses of CNTs (two of which were MWCNTs), and two of the studies investigated the toxicokinetics. In the work by Pothmann and Honda and their respective teams they confirmed alveolar deposition of CNTs reaching the target tissue and still reported negative findings with the in vivo comet assay [80,81,82]. Where the in vivo MN assay was utilized to determine chromosome damage, five studies reported negative data whereas only two report significant dose response data. These results were published by Patlolla et al. in 2010 [83] and 2016 [84], who investigated the effects of MWCNTs and then SWCNTs on adult male Swiss-Webster mice via intraperitoneal exposures of 0.25, 0.5 and 0.75 mg/kg. The MN data from these studies reveal only the top two doses could induce significant chromosomal breakage, which was further supported in both cases by positive comet data and CAs [83, 84].
In conclusion from the qualified in vitro papers, we can deduce that nanofiber genotoxicity can be shown in several key cell lines at low and high concentrations, following both acute (24 h or less) and slightly longer exposures of 48- and 72-h. Furthermore, in the studies which utilized uptake as part of the methodology, cellular internalization had been confirmed using TEM. These findings however do not seem to discriminate between SWCNTs and MWCNTs in terms of mode of action inducing their genotoxic effects. It appears more likely that their geometry and fiber paradigm are the primary factors driving their genotoxicity, which has been supported both in the literature and the qualified publications of this study. Where the qualified in vivo publications were concerned the majority of the data indicated negative responses even at when high, but still adequate doses were used. Additionally, when the in vivo comet assay was performed, 11 of the studies reported the test material was present in the target tissue thus demonstrating that the toxicity (or lack of toxicity) was reliably reported. From the qualified studies, the data is not conclusive enough to strictly classify nanofibers (the majority of which are CNTs) as genotoxic.