Endotoxin contamination of engineered nanomaterials: Overcoming the hurdles associated with endotoxin testing

Abstract

Nanomaterials are highly susceptible to endotoxin contamination due their large surface-to-volume ratios and endotoxins propensity to associate readily to hydrophobic and cationic surfaces. Additionally, the stability of endotoxin ensures it cannot be removed efficiently through conventional sterilization techniques such as autoclaving and ionizing radiation. In recent times, the true significance of this hurdle has come to light with multiple reports from the United States Nanotechnology Characterization Laboratory, in particular, along with our own experiences of endotoxin testing from multiple Horizon 2020-funded projects which highlight the importance of this issue for the clinical translation of nanomaterials. Herein, we provide an overview on the topic of endotoxin contamination of nanomaterials intended for biomedical applications.

This article is categorized under:

• Therapeutic Approaches and Drug Discovery > Emerging Technologies
• Toxicology and Regulatory Issues in Nanomedicine > Regulatory and Policy Issues in Nanomedicine

Graphical Abstract

Endotoxin binding to cationic and hydrophobic surfaces on nanomaterials.

1 INTRODUCTION

First identified in 1892 by Richard Pfeiffer, endotoxin was described as a heat-stable product of Vibrio cholerae that is released from the bacteria following its thermal destruction and displays inherent toxic potential (Pfeiffer, 1892). It is now well established that endotoxin—also known as lipopolysaccharide or LPS; although endotoxin refers to the crude toxin, while LPS describes a purer endotoxin in the absence of additional microbial fragments (Thorn, 2001)—is a structural component of the cell membrane of gram-negative bacteria that induces toxic effects by way of host inflammatory responses (Rosadini & Kagan, 2017). It has been estimated that a single bacterial cell can harbor approximately 3.5 million LPS molecules (Rietschel et al., 1994). Following bacterial cell death, and during growth and division, endotoxin is released into the local environment; hence, this molecule is ubiquitously found on surface tops, glassware, lab reagents, and in the air (Can et al., 2013; Gorbet & Sefton, 2005; Heinrich et al., 2003). Therefore, owing to its thermal stability and natural abundance, endotoxin is a common contaminant in nanomaterials and a significant hurdle to their early pre-clinical evaluation (Crist et al., 2013). Minute amounts (picogram to nanogram) of endotoxin can elicit immunostimulatory effects in vitro and in vivo, and so this contaminant is a significant safety hurdle that is commonly overlooked until the latter stages of nanomaterial preclinical evaluation, at which point the endotoxin may have contributed to safety issues and/or evoked false positive/negative results in a variety of in vitro and in vivo tests (Hannon et al., 2019; Yang Li, Fujita, & Boraschi, 2017). Furthermore, once bound to nanomaterials, endotoxin is difficult to remove by conventional sterilization techniques and so this hurdle must be considered early into preclinical assessment to avoid late failures (Hirayama & Sakata, 2002; Sandle, 2013). Here, we provide an overview of the literature surrounding endotoxin structure, bioactivity and its potential to contaminate nanomaterials, providing a challenging obstacle to their early preclinical progression.

2 ENDOTOXIN: CHEMICAL AND BIOLOGICAL OVERVIEW

2.1 Endotoxin structure

Endotoxin is a large (10–20 kDa for monomers and ≥1000 kDa for micelles [Hirayama & Sakata, 2002]), essential component of the outer membrane of gram-negative bacteria. It consists of a highly conserved intracellular lipid A anchor, an extracellular polysaccharide antigen termed “O-antigen,” and an interconnected oligosaccharide core (illustrated in Figure 1; Schletter et al., 1995). The O-antigen structure is highly variable even within strains of bacteria, containing repeating oligosaccharide units of two to six sugar monomers. As its name suggests, it is responsible for the antigenicity of the toxin, with the presence and order of sugar residues determining its immune specificity (Heine et al., 2001; Sampath, 2018). Indeed, not all endotoxin structures contain this polysaccharide backbone, and structures that lack an O-antigen are defined as “rough,” while intact backbones are classified as “smooth” (Seydel et al., 2000). By contrast, the lipid A region is less variable, and is responsible for the bioactivity of endotoxin. The degree of phosphorylation on the disaccharide backbone, along with the number of fatty acids play a significant role in the potency of lipid A (Rietschel et al., 1994; Schromm et al., 1998; Steimle et al., 2016).

FIGURE 1

Schematic representation of endotoxin derived from Escherichia coli O111:B4. Endotoxin derives from the outer membrane of gram-negative bacteria. The bioactive lipid A region is embedded in the cell membrane, acting as an anchor for the polysaccharide backbone. The O-antigen region consists of repeating oligosaccharide units which varies in structure between bacterial species. The red bars on endotoxin depict negatively charged phosphate groups. The endotoxin structure provided herein is an illustration of the chemical structure of E. coli 0111:B4 (Magalhães et al., 2007)

2.2 Endotoxin bioactivity

LPS-binding protein (LBP) is primarily made by hepatocytes and can be found in the circulation at concentrations of 3–7 μg/ml in humans; however, its concentration dramatically increases (more than 10-fold) during host infection (Prucha et al., 2003; Tobias et al., 1999). LBP can bind with high affinity to the lipid A region of endotoxin, dissociating aggregates to monomers and facilitate complexation with serum or membrane bound CD14. Depending on its origin, CD14 results in the transfer and recognition of endotoxin by the Toll-like receptor-4 (TLR4) and MD-2 receptor complex on different cells (Ryu et al., 2017; Tsukamoto et al., 2018). CD14 containing a glycosylphosphorylinositol tail is membrane bound to myeloid cells (predominantly monocytes and macrophages) which results in their activation, while serum-derived CD14 lacks this membrane anchor but can result in the activation of cells lacking membrane bound CD14 such as epithelial, endothelial and smooth muscle cells (Marcos et al., 2010; Pålsson-McDermott & O'Neill, 2004; Tobias et al., 1995; Tobias et al., 1999). Following cellular recognition, TLR4 signaling pathway becomes activated and recruits downstream adaptor molecules resulting in proinflammatory cytokine and Type I interferon expression via the TIRAP-MYD88 and TRAM-TRIF pathways, respectively (Bryant et al., 2010; Lu et al., 2008). While TIRAP-MYD88 signaling is initiated at the plasma membrane following TLR4 activation, the entire TLR4/MD2 receptor complex must be first endocytosed for subsequent TRAM-TRIF signaling induction (Figure 2) (Granucci & Zanoni, 2013; Kagan et al., 2008). Alternatively, aside from extracellular recognition of endotoxin, cytosolic endotoxin resulting from internalized bacteria (or contaminated nanomaterials) can also be recognized by intracellular inflammatory caspase 4 (and murine equivalent caspase 11) via their caspase recruitment domains and their subsequent activation can induce pyroptosis and trigger NLRP3 inflammasome activation-releasing proinflammatory cytokines IL-1β and IL-18 (Zamyatina & Heine, 2020).

FIGURE 2

Endotoxin signaling and downstream effects. Endotoxin introduced into the circulation by nanomaterials can activate the TLR4 signaling and stimulate the immune system. LBP can bind directly to endotoxin and facilitate its association to serum or membrane bound CD14, which, in turn, transfers endotoxin to the TLR4/MD2 receptor complex. Here, TIRAP-MYD88 adaptor protein signaling is activated at the plasma membrane, resulting in pro-inflammatory cytokine expression. Additionally, the TLR4/MD2 receptor complex can become endocytosed which subsequently activates TRAM-TRIF signaling, leading to Type I interferon expression

While the purpose of these responses are to overcome bacterial infection in the host, an overstimulation of these pathways via excessive exposure to endotoxin in circulation can result in a potent inflammatory response, which—when severe, in cases like endotoxemia—can lead to septic shock, organ damage, and even death (Danner et al., 1991; Zivot & Hoffman, 1995). Clinical investigations in humans have determined a bolus injection of endotoxin between 2 and 4 ng/kg can consistently induce transient systemic inflammation (marked by elevated temperature, white blood cell count, and heart rate), while concentrations as low as 0.06–0.2 ng/kg exert a sharp, temporary induction of plasma proinflammatory cytokines (Bahador & Cross, 2007; Calvano & Coyle, 2012; Fullerton et al., 2016; Taudorf et al., 2007). Moreover, exposure to endotoxin via inhalation is common and can occur through smoking, use of aerosols or occupation/environmental exposure (Verena Liebers et al., 2020; V. Liebers et al., 2008; Shamsollahi et al., 2019). Studies in humans have shown that concentrations of 2–5 μg is enough to induce detectable increases in neutrophils in sputum and blood, while concentrations of 30–100 μg have been shown to mildly compromise lung function and produce fever-like symptoms (Janssen et al., 2013; Thorn, 2001; Zielen et al., 2015).

3 ENDOTOXIN AND NANOMATERIAL CONTAMINATION

3.1 Endotoxin is a major hurdle for the translation of nanomaterials

Nanomaterials are particularly vulnerable to endotoxin contamination due to its available phosphate groups and hydrophobic lipid sites, along with nanomaterials high surface-to-volume ratios (Figure 3). Additionally, the multistep processes required for general nanomaterial development increase the risk of endotoxin being incorporated into the end-product through environmental contamination and/or the use of reagents or other chemical components that are not endotoxin-free. Removing it is also incredibly difficult, with common sterilization techniques such as autoclaving and ionization radiation proving ineffective (Yang Li, Fujita, & Boraschi, 2017). Under the National Cancer Institute, the United States (US) Nanotechnology Characterization Laboratory previously reported that more than a third of the nanomaterials they tested over a 1-year period failed early preclinical assessment due to endotoxin levels that did not satisfy regulatory requirements (26 out of 75) (Crist et al., 2013). Likewise, our group has provided endotoxin contamination assessment for nanomaterials associated with many Horizon 2020-funded projects including the European Union Nanomedicine Characterization Laboratory (EUNCL; Grant number: 654190), NoCanTher (Grant number: 685795), REFINE (Grant number: 761104), SAFE-N-MEDTECH (Grant number: 814607), along with other academic collaborations, and out of 29 nanomaterials tested up until now, 12 have had endotoxin levels above regulatory requirements that prevented advancement into preclinical assessment (see Table S1 for further details on these nanomaterials).

FIGURE 3

Endotoxin binding to cationic and hydrophobic surfaces on nanomaterials. (a) Nanomaterials with cationic surfaces are particularly susceptible to endotoxin binding due to their negatively charged phosphate groups (illustrated with red bars). Endotoxin can also form micelles in aqueous environments at high concentrations due to their hydrophobic lipids and hydrophilic polysaccharides, which too can interact electrostatically via their available phosphate groups. (b) The lipid A structure on endotoxin can hydrophobically bind to lipophilic surfaces on nanomaterials. Endotoxin contamination is not exclusive to cationic and hydrophobic nanomaterials. Endotoxin can be incorporated into nanomaterials at any point during synthesis or handling through environmental contamination or through the use of reagents or other chemical components that are not endotoxin-free

3.2 Regulatory requirements for endotoxin contamination

Nanomaterials utilized for biomedical applications have been approved as both medical devices and pharmaceutical drug products, and the endotoxin requirements vary between each. It is important to note that in the regulatory setting, endotoxin levels are expressed as endotoxin units, or EU, to account for the varying potencies of endotoxin derived from various bacteria (Morrison, 1983; Pearson et al., 1985). In the research setting, it is common to approximate 1 EU to 100 picograms of endotoxin (Dobrovolskaia et al., 2014). For medical devices that do not come into contact with cerebral spinal fluid (CSF), their limits are 0.5 EU/ml at no more than 20 EU/device. If contact with CSF does occur, the limits are 0.06 EU/ml at no more than 2.15 EU/device. Medicinal products that avoid the CSF have a limit of 5 EU/kg/h, whereas contact with CSF results in a limit of 0.2 EU/kg/h (FDA, 2012; USP, 2011) (Table 1).

TABLE 1. Summary of endotoxin requirements for medical devices and drugs

Product	Contact with CSF	No contact with CSF
Medical device	0.06 EU/ml at 2.15 EU/device	0.5 EU/ml at 20 EU/device
Drug	0.2 EU/kg/h	5 EU/kg/h

• Abbreviations: CSF, cerebral spinal fluid; EU, endotoxin units.

3.3 Endotoxin contamination interferes with nanomaterials safety and efficacy assessment

Endotoxin may contribute false positive/negative results in nanomaterials safety and efficacy testing if its levels are not established prior to undertaking these assessments (Y. Li & Boraschi, 2016; Yang Li, Fujita, & Boraschi, 2017). Our recent review covering the immunotoxic effects of cancer nanomedicines found that, of the 63 papers detailing immunological effects of various nanomaterials reviewed in-text, only 8 included any endotoxin assessment; suggesting this contaminant is commonly overlooked until the latter stages of preclinical assessment (Hannon et al., 2019). Importantly, mice are approximately one million times less sensitive than humans to LPS treatment, therefore preclinical studies may give an illusion of safety (Seemann et al., 2017; Warren et al., 2010). By contrast, nanomaterials are also capable of exacerbating the immunostimulatory effects of endotoxin (Dobrovolskaia et al., 2012; Inoue et al., 2006; Ko et al., 2018); accordingly, safety studies with contaminated nanomaterials are capable of generating misleading positive and negative results.

LPS has also been evaluated in clinical studies for its potent immune stimulatory effects in cancer. A number of trials have assessed its potential to induce antitumor immunity following the work of William B. Coley and others in the late 19th century which noted spontaneous remissions in cancer patients who contracted bacterial infections (Coley, 1893; Kienle, 2012). While the moderate antitumor activity observed in clinical trials has not yet outweighed its systemic toxic effects (Boushehri & Lamprecht, 2019; Engelhardt et al., 1995; Lundin & Checkoway, 2009; Otto et al., 1996), many preclinical studies in various cancers have demonstrated positive results (Han et al., 2017; Luo et al., 2004). Conversely, LPS has also been shown in multiple studies to enhance the metastatic potential of cancer; hence, nanomaterials containing endotoxin are capable of eliciting both false positive and negative results in cancer efficacy studies also (Jain et al., 2019; S. Li, Xu, et al., 2015).

4 DETECTING ENDOTOXIN CONTAMINATION—ESTABLISHED METHODS

In order to evaluate the levels of endotoxin in nanomaterials, four methods are currently accepted in Europe and US Pharmacopeia's: the rabbit pyrogen test (RPT), Limulus amoebocyte lysate (LAL) assay, monocyte activation test (MAT) and recombinant factor C (rFC) assay (EDQM, 2020; Franco et al., 2018; USP, 2020).

4.1 Rabbit pyrogen test

The RPT was the first regulatory-approved technique for pyrogen assessment of injectables and was implemented into multiple pharmacopeia standards in the 1940s. This model can identify-but not discriminate between-endotoxin and non-endotoxin-based pyrogens through the induction of fever (Schindler et al., 2009; Vipond et al., 2016). Herein, the test sample is intravenously injected into the ear vein of rabbits and rectal temperature measurements are taken every 30 min for 3 h to compare against temperatures taken prior to sample administration. Individual and cumulative rises in body temperature are used to determine positive responses, with a sensitivity down to 0.5 EU/ml (Borton & Coleman, 2018; Fennrich et al., 2016; Vipond et al., 2016). Notably, different pharmacopeia's have their own, individual experimental requirements and acceptance criteria for the RPT and have been reviewed and compared elsewhere (Dobrovolskaia & McNeil, 2016; Du et al., 2011; Franco et al., 2018; Hoffmann et al., 2005).

4.2 Limulus amoebocyte lysate assay

The LAL assay—otherwise known as the bacterial endotoxin test—entered pharmacopeia standards in the 1980s and proved a highly sensitive (down to 0.001 EU/ml), less costly and more ethical alternative to the RBT that quantifies levels of endotoxin specifically—with the exception of beta-glucans (cell wall components of bacteria and fungi) that can interfere with LAL assays if an inhibitory buffer is not used (Fennrich et al., 2016; Yang Li, Fujita, & Boraschi, 2017; Neun & Dobrovolskaia, 2019). The lysate used in LAL assays derives from the blood of a horseshoe crab—Limulus polyphemus—which clots at a rate dependent on the concentration of endotoxin (Young et al., 1972). Following endotoxin exposure, hemocytes aggregate and undergo degranulation, resulting in the release of clotting factors into the hemolymph. Factor C, a serine protease zymogen, is highly sensitive to endotoxin and becomes activated, initiating the clotting cascade (Iwanaga, 2007). Based on this physiological mechanism, three universally accepted assays have been developed that rely on three different endpoints: turbidity, colorimetric, and gel clotting (Dullah & Ongkudon, 2017; Franco et al., 2018; ISO, 2010; USP, 2011).

4.3 Monocyte activation test

While the LAL assay has played an influential role in the decline of rabbit numbers used for pyrogen testing, this test is not without its own ethical dilemmas as 10–30% of horseshoe crabs do not survive the bleeding procedure, and the surviving fraction can experience behavioral and physiological alterations (Anderson et al., 2013). Consequently, animal-free methods have been sought to overcome these issues and the most successful of these to-date is the MAT. Initially reported in the 1990s, the MAT measures changes in proinflammatory cytokine expression (IL-6 and IL-1β, in particular) in whole blood or isolated human blood cells following 6–24 h exposure of a test sample, with a sensitivity of 0.03 EU/ml (Borton & Coleman, 2018; Fennrich et al., 2016; T. Hartung & Wendel, 1995). The development of MAT over the last 25 years has been summarized comprehensively in a review by Thomas Hartung, the holder of initial patents on the test. Following an international validation study in 2005, and European Pharmacopeia acceptance in 2010, the MAT subsequently received backing by European Medicines Agency (EMA) and the Food and Drug Administration (FDA) in the last decade as a potential non-animal alternative for pyrogen testing (T. Hartung, 2021). Notably, rabbit use in pyrogen testing has dramatically declined since 2010 with about 170,000 used per year from 2008–2011, to a total of 35,172 in 2017 (Busquet et al., 2020; Thomas Hartung, 2015).

4.4 Recombinant Factor C assay

Identification of Factor C as the key initiator of the endotoxin-induced activation of the coagulation cascade led to the development of a recombinant form (rFC) at the turn of the century to overcome the ethical issues with bleeding horseshoe crabs (Bolden & Smith, 2017). Additionally, while the LAL assay can experience interference from beta-glucans, the rFC test only measures the interaction between endotoxin and Factor C in isolation and so overcomes false-positive results related to other blood components (Factor G) (Ding & Ho, 2001; Grallert et al., 2011). Based on this mechanism, a number of kits have become commercially available—PyroGene, EndoLISA, and EndoZYME—that measure fluorescence resulting from the endotoxin-activated rFC interacting with a fluorogenic substrate (excitation at 380 nm and emission at 440 nm). A sensitivity of 0.005 EU/ml can be achieved with these assays (Piehler et al., 2020). Multiple reports have validated the rFC assay against the LAL assay (Abate et al., 2017; Bolden & Smith, 2017; Maloney et al., 2018; Marius et al., 2020; Piehler et al., 2020), and in 2018 the FDA approved the first drug that utilized this assay for its endotoxin detection (USP, 2020). Moreover, both Europe and US Pharmacopeia have recently released reports endorsing its use (EDQM, 2020; USP, 2020); hence, it is expected EMA approval will follow.

5 DETECTING ENDOTOXIN CONTAMINATION—ALTERNATIVE METHODS

On top of the regulatory accepted methods for endotoxin detection in nanomaterials, several alternative approaches have also been successfully utilized in the literature and may prove valuable substitutes following successful validation in the future. Some notable examples of these are highlighted next.

5.1 High performance liquid chromatography and mass spectrometry

One method described in 2019 successfully utilized high performance liquid chromatography paired with mass spectrometry (HPLC-MS) to identify endotoxin in nanomaterials with a lower sensitivity range between 0.03 and 0.7 EU/ml (Giannakou et al., 2019). This method takes inspiration from the airborne detection of endotoxin achieved via gas chromatography and mass spectrometry (Mielniczuk et al., 1993), along with a similar HPLC-MS approach used for detecting plasma levels of endotoxin (Pais de Barros et al., 2015). Here, 2-hydroxy fatty acids and 3-hydroxy fatty acids were detected in nanomaterials as these are the key active components of the lipid A region of endotoxin. The HPLC-MS method was compared with the chromogenic LAL assay for a number of nanomaterials (liposomes, dendrimers, cerium dioxide, titanium dioxide, and iron oxide nanoparticles) and could outperform the LAL assay in cases of interference due to pH, optical overlap and lipid-based nanoparticles, while displaying reasonably comparable results to the LAL assay when successful measurements could be achieved for both methods.

5.2 Macrophage activation test

The macrophage activation test is a refinement on the traditional, regulatory-accepted MAT where macrophages are isolated from human peripheral blood mononuclear cells to achieve a more specific response. This test has been evaluated using five graphene-based materials (GBM) at concentrations confirmed to be non-cytotoxic and compared against the chromogenic and gel clot LAL assays (Mukherjee et al., 2016). The primary macrophages were treated with the GBM in the presence and absence of the endotoxin inhibitor polymyxin B sulfate (PMB) for 24 h and changes in expression to TNF-α were subsequently measured. Again, benefits could be observed with this technique over the chromogenic LAL assay where optical interference was induced with the GBM and could be overcome with this cellular assay. Moreover, a sensitivity of 0.005 EU/ml could be achieved with this method, hence, outperforming the traditional MAT in this regard.

5.3 TLR4 reporter assay

This approach utilizes a reporter cell line transfected with TLR4 and its co-receptor MD-2/CD14, along with secreted embryonic alkaline phosphatase (SEAP) reporter gene to measure endotoxin levels in titanium dioxide, silicon dioxide, silver, and calcium carbonate nanomaterials in comparison to the gel clot and chromogenic LAL assays (Smulders et al., 2012). Triggering of TLR4 signaling in these cells results in activation of transcription factors AP-1 and NFκB that induce expression of SEAP. A substrate is then added to the supernatant (QUANTI-blue) which undergoes a color change in the presence of this phosphatase that can be detected spectrophotometrically between 620 and 655 nm, correlating to endotoxin concentration (Sharma et al., 2019). In the abovementioned Smulders et al. study, the reporter cells were treated with each nanomaterial for 22 h (with and without PMB) at concentrations that did not elicit cytotoxic effects. The supernatant was then measured and endotoxin in the nanomaterials could be quantified successfully with optimal spike recoveries. The TLR4 reporter assay performed comparably to the chromogenic LAL assay with a sensitivity of 0.05 EU/ml and could overcome nanomaterial interference observed in the gel clot assay.

6 NANOMATERIAL INTERFERENCE WITH ENDOTOXIN CONTAMINATION ASSESSMENT

A variety of nanomaterials have proven capable of interfering with commonly used assays such as MTT, LDH, and ELISA's (Kroll et al., 2012). They have also shown a strong propensity to interfere with the endotoxin contamination assessment. These interferences manifest as either inhibitions (false negatives) or enhancements (false positives) in the assays. For this reason, the inhibition/enhancement control (IEC) is included in LAL, MAT, and rFC testing which involves spiking the nanomaterial with a known concentration of endotoxin and quantifying its recovery in the assay. If less than 50% of the endotoxin is recovered, the nanomaterial inhibits the assay; if more than 200% is reportedly recovered, the nanomaterial evokes an enhancement to the assay; a recovery between 50 and 200% the nanomaterial is deemed not to interfere (Dobrovolskaia, 2015; Dobrovolskaia et al., 2014; ISO, 2010).

Many reports have been published describing nanomaterial interference with LAL assay, particularly by the Nanotechnology Characterization Laboratory in the US. Notably, nanomaterials with high absorbance of wavelengths between 400 and 550 nm interfere with the chromogenic and rFC assays and nanoparticles with high turbidity interfere with the turbidimetric assay. Additionally, nanomaterials can bind to endotoxin directly, or components of the assay, leading to interference with multiple LAL assays (Dobrovolskaia et al., 2009; Kucki et al., 2013; Y. Li, Italiani, et al., 2015). One LAL assay has been suggested as not enough to accurately determine endotoxin levels in a nanomaterial due to potential assay interferences which may not be picked up with the IEC. Accordingly, a comparison between two LAL assays recommended; moreover, if there is greater than 25% difference between the two assays, the results should be confirmed with the RPT or MAT (Dobrovolskaia, 2015; Dobrovolskaia et al., 2009; Dobrovolskaia et al., 2010).

Nanomaterials that elicit proinflammatory or immunosuppressive responses, carry cytotoxic drugs, absorb light at the endpoint wavelength or bind to ELISA constituents may also interfere with the MAT. Moreover, the MAT does not quantify endotoxin levels directly but only the induction of proinflammatory cytokine expression as a marker for fever (Dobrovolskaia et al., 2014; Yang Li, Fujita, & Boraschi, 2017). The RPT, by contrast, overcomes many of the interference issues associated with in vitro assays; however, it is also not endotoxin specific, and nanomaterials that are inherently immune stimulatory, immunosuppressive, or carry cytotoxic drugs may interfere in this case also (Jin et al., 2018; Palma et al., 2017).

7 OVERCOMING INTERFERENCE

While diluting the nanomaterial can overcome interference in some instances, this is only possible to a certain point—the maximum valid dilution (MVD). The MVD is the maximum dilution at which endotoxin can still be detected in the assay—calculated using the concentration of the nanomaterial, its endotoxin limit, and the sensitivity of the assay (USP, 2011). If interference is maintained at this dilution, an alternative approach must be considered—either a change in the assay itself or a modification to the source of the interference, if possible. For instance, inorganic nanomaterials have displayed a strong propensity to interfere with the chromogenic assay due to high absorbances at 405 nm. In Figure 4 below, we describe a case involving iron oxide nanoparticles (IONP) from various projects. Here, a strong inhibition in the chromogenic assay is observed from a screen of five different IONP that could be overcome by shifting absorbance endpoint to 550 nm via the diazo reagent. This chromogenic modification has also proved efficient at removing interference for gold, silver and other IONP previously by Li et al. (Y. Li, Italiani, et al., 2015; Yang Li, Shi, et al., 2017). Inherently turbid nanomaterials such as liposomes and nanoemulsions are likely to interfere through enhancements in the turbidimetric LAL assay (Dobrovolskaia et al., 2014; Neun & Dobrovolskaia, 2011). If this interference remains up to the MVD for these materials, an alternative LAL assay should be used. Conversely, lipid-coated nanomaterials can also bind to the lipid moiety on endotoxin, leading to inhibitions in a variety of assays. This interference can be overcome, however, through thermal destruction of the lipid at 96°C for 15–30 min which prevents endotoxin binding while preserving the heat-stable endotoxin in the formulation (Neun & Dobrovolskaia, 2018). Furthermore, nanomaterials with cationic surfaces also elicit this inhibitory effect in many assays, as direct binding of phosphate groups on endotoxin to the positively charged functional groups on the nanomaterial (Sondhi et al., 2019). Such interference could be avoided by exposing cationic liposomes to 1% sodium dodecyl sulfate (diluted 1:1), which neutralized the positive charge and prevented direct binding of the spiked endotoxin, enabling an acceptable recovery to be achieved in this case (Neun & Dobrovolskaia, 2018). Interestingly, it has also been suggested that the use of endotoxin inhibitors (such as PMB) could be useful for identifying interference in the MAT. Nanomaterials that are inherently immune stimulative may evoke false positive effects in this assay, and PMB can be used to confirm this by neutralizing the effects of endotoxin in the assay; thereby, enabling the immune stimulative effects of the nanomaterial to be directly observed (Dobrovolskaia & McNeil, 2016; Mukherjee et al., 2016). These methods come with caveats, however, as chemical modification of the nanomaterial or endotoxin may unintentionally antagonize the function of the other, making it difficult to extrapolate meaningful results in these cases and should be used with caution.

FIGURE 4

Overcoming inhibitions with the chromogenic assay. Various iron oxide nanoparticles (IONP) were found to strongly inhibit the chromogenic assay with an absorption endpoint of 405 nm. Switching the endpoint to 550 nm via the addition of the diazo reagent successfully overcame this interference in each case. Characteristics of each nanoparticle (core, coating, functional group, and size) were provided by the suppliers. Red lines on graph depict the acceptable margins for spike-recovery (50–200%). Abbreviations: PEG, polyethylene glycol

Hence, the unique characteristics of nanomaterials make them particularly proficient at interfering with assays used for endotoxin detection. If diluting the nanomaterial is not successful for a particular assay, the assay can be changed, or the source of interference can be altered to address this issue. Some nanomaterial characteristics evoke predictable interferences in endotoxin assays, enabling considerations to be made early with regard to assay choice. In Table 2 below, we provide a summary of these predictable interferences and note considerations for each case.

TABLE 2. Common sources of nanomaterial-based interferences with endotoxin contamination assessment and some considerations for avoidance

Nanomaterial characteristic	Potential interferences	Considerations
Optical property: Absorbs highly between 400 and 550 nm.	Chromogenic LAL assay, rFC and MAT rely on these wavelengths as their endpoint.	Turbidimetric and gel clot LAL assays and the RPT may be more suitable.
Optical property: Inherently turbid.	Enhancement of turbidimetric LAL assay.	Chromogenic and gel clot LAL assays, rFC and MAT may be more suitable.
Surface chemistry: Lipid-based surface coating.	Direct binding to lipid on endotoxin; Inhibition of LAL, rFC and MAT assays.	Thermally destroy the lipid nanomaterial prior to testing (~100°C for 15–30 min).
Surface chemistry: Positive surface charge.	Direct binding to phosphate groups on endotoxin; Inhibition of LAL, rFC, and MAT assays.	Neutralized positive charge on nanomaterial with sodium dodecyl sulfate prior to testing. RPT may be more suitable.
Immune stimulation: Contains beta-glucans.	Enhancement of LAL assay cascade via Factor G.	Use glucan inhibitory buffer or use rFC assay.
Immune alteration: Inherent immune stimulation or suppression.	Inhibition or enhancement of MAT and RPT.	LAL and rFC assays may be more suitable.
Biological effects: Contains cytotoxic components.	Interference with MAT and RPT.	LAL and rFC assays may be more suitable.

• Abbreviations: LAL, Limulus amoebocyte lysate; MAT, monocyte activation test; rFC, Recombinant factor C; RPT, rabbit pyrogen test.

8 STEPS TO AVOIDING ENDOTOXIN CONTAMINATION IN NANOMATERIALS

There are no standards in place for endotoxin removal from nanomaterials. The variability in molecular weight and thermal stability of endotoxin make it extremely difficult to remove once bound to the materials surface, and conditions that are effective (temperatures 250°C for 30 min) compromise the structure of nanomaterials used in biomedical applications (Y. Li & Boraschi, 2016). Additionally, as conventional sterilization techniques such as autoclaving and irradiation have proven minimally effective, the best way to avoid contamination is by ensuring endotoxin-free synthesis (Afonin et al., 2011; Lebre et al., 2019; Yang Li, Fujita, & Boraschi, 2017; Smulders et al., 2012; Vallhov et al., 2006). Hereto, we provide an example we faced in a large European project-NoCanTher, where the scale up of a leading IONP was prevented due to endotoxin contamination above the regulatory requirements for a medical device (Figure 5). Upon further investigation, it was determined that both the iron oxide core and starch coating of the nanomaterial were contributing to the endotoxin levels in the final product. Accordingly, it is essential that all starting reagents are endotoxin-free to ensure endotoxin is not being introduced in the earliest phases of synthesis. Moreover, all glassware used should be heated to remove endotoxin prior to use, and any plasticware should be certified endotoxin free also. Finally, as much of the synthesis as possible such be undertaken in a laminar flow hood to limit environmental exposure. These recommendations have also been voiced by the US Nanotechnology Characterization Laboratory and other groups (Dobrovolskaia & McNeil, 2016; Leong et al., 2019; Y. Li & Boraschi, 2016; Mukherjee et al., 2016).

FIGURE 5

Endotoxin assessment of a leading IONP. Following interference in the chromogenic assay, the UV–Vis absorbance spectra of the nanomaterial showed that the IONP absorbed highly at the wavelengths used in the assay (0.91 AU at 405 nm). With wavelengths used in the diazo chromogenic assay, however, it absorbed considerably less (0.23 AU at 550 nm). Based on this, the diazo chromogenic assay was successfully used to overcome the interference and determine the endotoxin contamination in the nanomaterial. This concentration was deemed too high to fulfill the requirements of a medical device and so the nanomaterials constituents were tested to determine the source of contamination. The polymer coating around the nanomaterial was found to have a considerable amount of endotoxin, whereas the iron oxide core contributed less than half this concentration, and the diluent had levels of endotoxin that were undetectable by the assay. The transmission electron microscopy image scale bar represents 100 μm

9 CONCLUSION

Endotoxin is a potent immune stimulant and ubiquitously found on surfaces, glassware, laboratory reagents, and in the air. It is also a common contaminant of nanomaterials due to their large surface-to-volume ratios and endotoxins available phosphate and lipid groups. If nanomaterials are not assessed for their levels of endotoxin contamination prior to safety and efficacy testing, it may lead to false positive or false negative results. The extent of this issue has led to a lot of published works on guidelines for endotoxin detection of nanomaterials and overcoming interference with these assays. Limiting endotoxin contamination during synthesis and ensuring efficient contamination assessment in the future will reduce the number of late failures with nanomaterials and improve overall translation.

ACKNOWLEDGMENTS

The authors would like to thank the Irish Research Council (Grant ID: GOIPG/2016/1629), EUNCL (Grant agreement number: 654190), NoCanTher (Grant agreement number: 685795), SAFE-N-MEDTECH (Grant number: 814607) and REFINE (Grant agreement number: 761104). We would also like to acknowledge BioRender for the use of their software to construct Figures 1 and 2.

CONFLICT OF INTEREST

The authors have declared no conflicts of interest for this article.

AUTHOR CONTRIBUTIONS

Gary Hannon: Conceptualization; data curation; methodology; software; writing-original draft. Adriele Prina-Mello: Conceptualization; data curation; funding acquisition; investigation; project administration; resources; supervision; validation; writing-review & editing.

RELATED WIREs ARTICLES

Forming interdisciplinary expertise: One organization's journey on the road to translational nanomedicine

Nonclinical regulatory immunotoxicity testing of nanomedicinal products: Proposed strategy and possible pitfalls

Nanoparticle therapeutics: A personal perspective

Open Research

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.