Research ArticleNanomedicine

Aurora kinase inhibitor nanoparticles target tumors with favorable therapeutic index in vivo

See allHide authors and affiliations

Science Translational Medicine  10 Feb 2016:
Vol. 8, Issue 325, pp. 325ra17
DOI: 10.1126/scitranslmed.aad2355

Accurin nanoparticles dutifully deliver drug

A class of drugs, called kinase inhibitors, could stop cancer in its tracks…if only these drugs could reach the tumors, stay for a while, and not be toxic. Hypothesizing that a nanoparticle formulation would solve the inhibitors’ woes, Ashton and colleagues investigated several different compositions of so-called Accurins—polymeric particles that encapsulate charged drugs through ion pairing. An Aurora B kinase, once formulated in Accurins, demonstrated a much-improved therapeutic index and preclinical efficacy compared with its parent molecule, when administered to rats and mice bearing human tumors. The Accurins allowed for sustained release of the drug over days, and did not have the same blood toxicity seen with the parent drug. A phase 1 trial is the next step for this nanomedicine, and additional preclinical studies will reveal whether such nanoformulations can improve the tolerability and efficacy of the broader class of molecularly targeted cancer therapeutics, including cell cycle inhibitors.


Efforts to apply nanotechnology in cancer have focused almost exclusively on the delivery of cytotoxic drugs to improve therapeutic index. There has been little consideration of molecularly targeted agents, in particular kinase inhibitors, which can also present considerable therapeutic index limitations. We describe the development of Accurin polymeric nanoparticles that encapsulate the clinical candidate AZD2811, an Aurora B kinase inhibitor, using an ion pairing approach. Accurins increase biodistribution to tumor sites and provide extended release of encapsulated drug payloads. AZD2811 nanoparticles containing pharmaceutically acceptable organic acids as ion pairing agents displayed continuous drug release for more than 1 week in vitro and a corresponding extended pharmacodynamic reduction of tumor phosphorylated histone H3 levels in vivo for up to 96 hours after a single administration. A specific AZD2811 nanoparticle formulation profile showed accumulation and retention in tumors with minimal impact on bone marrow pathology, and resulted in lower toxicity and increased efficacy in multiple tumor models at half the dose intensity of AZD1152, a water-soluble prodrug of AZD2811. These studies demonstrate that AZD2811 can be formulated in nanoparticles using ion pairing agents to give improved efficacy and tolerability in preclinical models with less frequent dosing. Accurins specifically, and nanotechnology in general, can increase the therapeutic index of molecularly targeted agents, including kinase inhibitors targeting cell cycle and oncogenic signal transduction pathways, which have to date proved toxic in humans.


Investment in the application of nanoparticle technology to develop cancer therapeutics is growing rapidly, with six products approved and many more in clinical development (1). By increasing the drug concentration at tumor sites relative to healthy tissue, nanoparticle formulations have the potential to improve both efficacy and safety, thereby enabling promising treatments otherwise limited by narrow therapeutic index (2)—that is, unacceptable toxicity at doses required for efficacy. To date, efforts to develop nanoparticle therapies for cancer have used established cytotoxic drugs as payloads, resulting in products like liposomal doxorubicin (DOXIL) and albumin-bound paclitaxel (Abraxane), which overcome toxicities associated with the conventional dosage forms. In particular, liposomal doxorubicin significantly reduces the cardiotoxicity that limits the cumulative dose of doxorubicin (3, 4), and the improved safety profile of albumin-bound paclitaxel enables a >50% increase in the dose of paclitaxel that can be safely administered (4). Other nanoformulated cytotoxic drugs are in clinical development, including BIND-014 (BIND Therapeutics Inc.), which contains docetaxel and is targeted to prostate-specific membrane antigen (5), and NC-6004 (NanoCarrier Co. Ltd.), a micellar nanoparticle containing cisplatin derivatives (6). In contrast, the application of nanotechnology to molecularly targeted agents (MTAs), such as antimitotic agents that inhibit the cell cycle and kinase inhibitors that interfere with oncogenic signal transduction pathways, has received little attention. Although MTAs are generally expected to be more tolerable than cytotoxic agents, many exert on- and off-target toxicities that preclude their administration at doses required for efficacy.

One class of MTAs that has been especially limited by toxicity is kinase inhibitors that target the cell cycle, such as inhibitors of polo-like kinases, Aurora kinases, and cyclin-dependent kinases. In particular, Aurora kinase inhibitors have shown disappointing single-agent activity at tolerable doses in the clinic, particularly in solid tumors (7), and have proven difficult to combine with cytotoxics or other MTAs owing to overlapping, enhanced, or unexpected toxicities (8). Aurora B kinase plays a pivotal role in cell cycle progression by controlling cytokinesis, chromosome biorientation, and segregation through regulation of the spindle checkpoint (9, 10). Inhibiting Aurora B kinase ultimately initiates mitotic catastrophe and cellular apoptosis (11).

AZD2811 (previously known as AZD1152 hydroxyquinazoline pyrazol anilide; AZD1152-hQPA; Fig. 1A) is a potent and specific small-molecule Aurora B kinase inhibitor (12). Its water-soluble dihydrogen phosphate prodrug, AZD1152, has been tested in clinical trials in various tumors, including acute myeloid leukemia (AML). In a randomized phase 2 trial, AZD1152 administered as a 7-day infusion led to a significant improvement in the complete response rate compared to standard of care (13). Efficacy, however, was associated with major toxicities in those patients, including myelosuppression. Despite this clinical proof of concept, the toxicity profile and requirement for continuous intravenous infusion together limit the broader utility of AZD1152 in humans. Similar hurdles have confronted other MTAs targeting the cell cycle, including other Aurora kinase inhibitors as well as inhibitors of polo-like kinase and kinesin spindle protein, and as a result, many have been discontinued (14).

Fig. 1. Characterization of AZD2811 nanoparticle formulations in vitro and in vivo.

(A) Chemical structures of AZD2811 and AZD1152. (B) Schematic diagram of nanoparticle containing AZD2811 and ion pairing agents. (C and D) Cumulative in vitro release of AZD2811 measured over time for Accurins A to F. Data are averages of duplicate measurements. All duplicates differed by less than 5%. (E) Plasma concentration of total AZD2811 over time following a single intravenous injection of Accurins B or E (25 mg/kg) in nude rats. The plasma concentration of AZD2811 after intravenous administration of an AZD1152 bolus (25 mg/kg) is shown as a reference. Data are means ± SD (n > 4).

To enable a safe and effective Aurora B kinase–targeted therapy, we have encapsulated AZD2811 in polymeric nanoparticles termed Accurins. Accurins are composed of block copolymers of poly-d,l-lactide (PLA) and poly(ethylene glycol) (PEG)—both clinically validated biomaterials with a long history of safe use in humans (15). Accurins accumulate in tumors, increasing the concentration and duration of exposure of cancer cells to the therapeutic payload (5). To optimize the AZD2811 load and duration of tumor exposure, we developed an in situ ion pairing approach in which organic acid counterions were used to increase encapsulation efficiency and decrease the release rate of AZD2811. From a panel of nanoparticles with diverse characteristics, we identified a formulation profile that could deliver active drug for more than 1 week, resulting in prolonged target inhibition in tumor tissue together with improved preclinical efficacy and therapeutic index over the AZD1152 prodrug in several animal models. In situ ion pairing in Accurin nanoparticles can be applied to other drugs with ionizable functional groups, and has the potential to enable nanomedicine with profoundly differentiated pharmacological properties.


Nanoparticle formulation strategy

We sought to identify AZD2811 nanoparticle formulations with efficacy and tolerability characteristics superior to AZD1152, with the ultimate goal of clinical testing. We hypothesized that the concentration-time profile of tumor exposure to AZD2811 would be a key determinant of efficacy and differentiation from conventional forms of AZD1152, and that this profile would strongly depend on the kinetics of release of AZD2811 from the nanoparticles. Similarly, we anticipated that the incidence and severity of bone marrow toxicity, which limits the dose of conventional AZD1152 and other Aurora kinase inhibitors, would also depend on release kinetics. We therefore generated a series of nanoparticle formulations spanning a range of in vitro release rates using commercially available materials suitable for use in clinical products, including organic acids to enhance the organic solvent solubility of AZD2811 (Fig. 1B), which contains an aliphatic tertiary amine (16), through ion pairing. Hydrophobic ion pairs may be less prone to rapid release from polymeric nanoparticles under physiological conditions than the free base by virtue of increased size and hydrophobicity.

Nanoparticle formulation development

We first determined the solubility of AZD2811 in benzyl alcohol and ethyl acetate, solvents comprising the organic phase of nanoemulsions used in the manufacture of Accurin nanoparticles. AZD2811 was only moderately soluble in benzyl alcohol and sparingly soluble in ethyl acetate (table S1). These results were not optimal because lower organic phase solubility of the active pharmaceutical ingredient (API) limits the maximum theoretical drug loading and increases the solvent volume and tank size requirements for large-scale manufacturing. AZD2811 was highly soluble in dimethylformamide and dimethyl sulfoxide (DMSO) (polar aprotic solvents which can be used as organic phase co-solvents), and the solubility in benzyl alcohol was increased slightly by addition of water (table S1). AZD2811 solubility in benzyl alcohol was increased to a much greater extent by addition of mono- and divalent carboxylic and sulfonic organic acids, with increases in solubility ranging from 4- to 10-fold compared to benzyl alcohol alone. With the exception of oxalic acid, all solutions containing organic acids and AZD2811 were stable to precipitation overnight under ambient conditions. For certain acids, addition of ethyl acetate to AZD2811/acid solutions resulted in immediate precipitation. In most cases, however, the solutions were stable enough to manufacture nanoparticles.

Two nanoparticle formulation approaches were evaluated that differed mainly with respect to the organic phase used in the nanoemulsion process: (i) benzyl alcohol with co-solvents (table S2A), and (ii) benzyl alcohol/ethyl acetate mixtures with organic acid counterions of varying size, hydrophobicity, and pKa (where Ka is the acid dissociation constant) (table S2B). In addition to the organic phase composition, key variables included the ratio of AZD2811 to remaining nanoparticle components and the solids concentration of the organic phase before emulsification. All nanoparticle formulations contained a PLA-PEG diblock copolymer (approximate number-average molecular weight, 16 kD for PLA and 5 kD for PEG), which is commercially available in quantities and quality suitable for use in the manufacture of a clinical material.

Nanoparticles were characterized against target ranges for particle size (80 to 130 nm), drug loading (≥5% w/w), and release kinetics (time to 50% release = 2 to ≥72 hours). Characteristics of AZD2811 nanoparticles produced with benzyl alcohol alone or combined with DMSO or water are in table S2. In all cases, nanoparticles formed at low solids concentration were close to the target size range, but only nanoparticles produced from a benzyl alcohol/water organic phase achieved the target AZD2811 load. Although the increased solubility of AZD2811 in benzyl alcohol/water or benzyl alcohol/DMSO mixtures allowed higher organic phase solids concentrations, increasing the solids concentration resulted in particle diameters above the target range. Moreover, all nanoparticle formulations displayed high levels (>10%) of in vitro release of the drug immediately after hydration at 37°C and release rates at the faster end or outside of the target range (time to 50% release ≤ 2 hours).

The effects of adding organic acids to AZD2811 nanoparticle formulations were investigated using trifluoroacetic acid (TFA), oleic acid, xinafoic acid, and dioctyl sulfosuccinic acid (DOSA)—all components of U.S. Food and Drug Administration–approved pharmaceutical products. TFA, oleic acid, and xinafoic acid enhanced the solubility of AZD2811 in benzyl alcohol by up to 10-fold (table S1). Separately, it was observed that stable solutions containing AZD2811 (more than 300 mg/ml) in benzyl alcohol could be prepared by adding one equivalent of DOSA. Characteristics of the resulting AZD2811 nanoparticles are in table S2. Addition of organic acids in quantities ranging from 0.7 to 2 molar equivalents resulted in marked increases in the AZD2811 load together with significant reductions in the initial in vitro release and overall release rate (increased time to 50% release). AZD2811 loading levels were above the target value in formulations where acid/API molar ratios were 1.0 or greater, and trended toward higher values as the molar ratio was increased. In vitro release kinetics of nanoparticles containing organic acids varied considerably depending on the quantity and type of acid used, with progressively decreasing release rates observed with oleic acid, xinafoic acid, TFA, and DOSA, respectively. In most cases, the initial drug release was less than 5%. In compositions including at least one equivalent of TFA or DOSA, times to 50% release were greater than 48 hours, corresponding to 25- to >100-fold decreases in the release rate compared to nanoparticles lacking organic acid additives.

In vitro release and in vivo pharmacokinetics of AZD2811 nanoparticles

On the basis of these results, nanoparticle formulations designated Accurins A to E (table S2) were selected and scaled up for preclinical studies. Accurin F, generated subsequently, mimicked the release profile of Accurin E with a higher drug load. These formulations collectively spanned a wide range of in vitro release rates, with times to 50% release ranging from 0.7 to >72 hours (table S2). The characterization data for the large-scale batches (Fig. 1) were generally consistent with the formulation screening results and, with the exception of the AZD2811 loading levels for Accurins C and D (4.7 and 4.1%, respectively), fell within the target ranges (Table 1). Accurins C, D, and E displayed noticeably higher initial in vitro release levels than the screening batches, but were suitable for in vivo evaluation. In vitro release kinetic profiles displayed extended release durations ranging from hours to several days (Fig. 1C). Accurin E exhibited continuous release for about 1 week (Fig. 1D).

Table 1. Characteristics of large-scale AZD2811 nanoparticle batches.
View this table:

The plasma pharmacokinetic (PK) profiles of Accurins B and E (versus the equivalent dose of AZD1152) were determined in vivo in rats by measuring the total (encapsulated plus released) AZD2811 concentration by liquid chromatography–tandem mass spectrometry (LC-MS/MS). AZD2811 plasma concentrations were about two orders of magnitude higher after nanoparticle administration compared to AZD1152, and were detectable for more than 20 days (Fig. 1E). Thus, Accurin nanoparticles are retained in the vascular compartment and cleared slowly, and AZD2811 release in rodents occurs at a slow and controlled rate.

Pharmacodynamic effects of AZD2811 nanoparticles

To test whether AZD2811 nanoparticles effectively inhibit Aurora B kinase, nude rats bearing human colorectal adenocarcinoma SW620 tumors were administered a single intravenous dose of AZD1152 (25 mg/kg) or Accurins B, C, D, and E containing a single dose equivalent of AZD2811 (25 mg/kg) (Fig. 2). SW620 is a robust xenograft model in mice and rats with known sensitivity to the prodrug AZD1152. As expected from a previous study (11), AZD1152 reduced the level of phosphorylated histone H3 (pHH3)—a marker of mitosis—in tumors within 6 hours. pHH3 levels recovered within 24 hours, which is consistent with the rapid clearance of AZD2811 (Fig. 2 and fig. S1).

Fig. 2. PDs of AZD2811 nanoparticles in nude rats bearing human tumors.

Nude rats bearing colon adenocarcinoma (SW620) xenografts were treated with AZD1152 or AZD2811 Accurin nanoparticles (25 mg/kg). Percent change in pHH3 was determined by fluorescence-activated cell sorting (FACS) analysis of disaggregated tumors. The percentage of pHH3+ cells is relative to 30% sucrose control and is represented as a mean ± SEM (n = 3 to 5).

The nanoparticle formulations displayed more prolonged pHH3 reductions than AZD1152, with the time course of pharmacodynamic (PD) response consistent with the in vitro release profile (Fig. 1C). Accurins B, C, and D induced reductions in pHH3 within 6 hours and maximal pHH3 reductions at 24 or 48 hours (Fig. 2). In contrast, the slower-releasing Accurin E displayed a distinct PD profile (Fig. 2 and fig. S1), reducing pHH3 progressively over the 96-hour time course. Accurins B and E, representing faster and slower release profiles, respectively, were selected for further evaluation.

Tolerability and efficacy of AZD2811 nanoparticles in human tumor xenograft model

The efficacy and bone marrow tolerability of AZD2811 nanoparticles were evaluated in nude rats bearing SW620 xenograft tumors. Rats were used in these studies because they more faithfully model human bone marrow than mice (17). AZD1152 and Accurins B and E were compared by administering four daily doses of AZD1152 (25 mg/kg) (days 1 to 4) or two doses of Accurin B or E (25 mg/kg) (days 1 and 3). Daily bolus administration of AZD1152 was selected as a comparator on the basis of previous studies in rats showing less bone marrow toxicity and similar efficacy in comparison with AZD1152 administered by continuous infusion (18). Accurins B and E administered at half the dose intensity of AZD1152 inhibited tumor growth by 92 and 101%, respectively, and were more effective than prodrug AZD1152, which inhibited tumor growth by 58% (Fig. 3A). The nanoparticle-treated tumors initially progressed for several days before exhibiting stasis and ultimately regression relative to previous tumor measurements, consistent with extended profile of AZD2811 exposure. Neither nanoparticle formulation had an impact on body weight (Fig. 3B), suggesting a lack of overt toxicity.

Fig. 3. The effects of AZD1152 and Accurins B and E on SW620 tumor growth and bone marrow.

Nude rats bearing SW620 tumors were treated with placebo nanoparticles, AZD1152 (25 mg/kg daily for days 1 to 4; total dose 100 mg/kg) or Accurins B and E (25 mg/kg dosed on days 1 and 3; total dose 50 mg/kg). (A) Growth curves of SW620 tumors. Data are mean tumor volumes ± SEM (n = 10). **P < 0.01, ***P < 0.001, one-sided Student’s t test versus placebo. Relative times of dosing points are shown. (B) Changes in body weight relative to day 0. Data are means ± SEM (n = 10). (C) Tumors taken 5 days after first dose were analyzed for pHH3 and polyploidy, as well as tumor cell integrity by H&E staining. Bone marrow samples taken 5 and 9 days after the first dose were analyzed by staining with H&E. Scale bar, 50 μm. (D) Pathologist scoring of H&E-stained sections from bone marrow samples taken at 5, 9, and 15 days after the first dose; 0 represents intact bone marrow, 3 represents moderate impact on bone marrow. Data are individual samples, one per animal (n = 2 to 7). (E) Bone marrow cellularity as a percentage of total nuclear cell count (by FACS analysis) of bone marrow aspirates taken at 5, 9, and 15 days after the first dose. Historical data with AZD1152 showed that bone marrow returned to normal by day 9 (13), so data were not generated for day 15. Data are individual samples, one per animal (n = 2 to 5). (F) White blood cell numbers from peripheral blood samples taken at 5, 9, and 15 days after the first dose. Data are individual samples, one per animal (n = 2 to 5).

Accurins B and E showed similar PD effects in the tumor, inducing polyploid nuclei (Fig. 3C), a phenotype of Aurora B kinase inhibition by AZD1152 (19). At day 5, polyploid nuclei were evident in tumors treated with AZD1152 and Accurins B and E, confirming that each tumor had been exposed to sufficient concentrations of AZD2811 to inhibit Aurora B kinase function. Hematoxylin and eosin (H&E) staining of bone marrow showed that AZD1152 and Accurin B (but not Accurin E) reversibly reduced bone marrow cellularity at day 5, which recovered at day 9 (Fig. 3, C to E). Accurin E had minimal effect on bone marrow through day 15, confirming that the extended exposure profile does not merely delay the bone marrow effects. Accurin E had a modest effect on lymphoid cell numbers but little impact on myeloid cells; in contrast, AZD1152 had a greater impact on the myeloid cell numbers at day 5 (fig. S2, A and B). AZD1152 also reduced white blood cell counts in peripheral blood by day 5, whereas Accurin E had minimal effect over a 15-day time course (Fig. 3F).

The impact of nanoparticle release kinetics on in vivo performance was further characterized by treating rats bearing SW620 tumor xenografts with Accurin F, which had a higher load and different ion pairing agent, but matched the release profile of Accurin E. Short-term dosing indicated that Accurin F was efficacious (Fig. 4A) and had similar PD characteristics to Accurin E (Fig. 4B), with negligible effects on bone marrow (Fig. 4C). Collectively, these data support the conclusion that AZD2811 nanoparticles with a slower release rate can regress tumors while minimizing bone marrow toxicity.

Fig. 4. The effects of Accurins E and F in vivo on SW620 tumor PD, tumor growth, and bone marrow.

For tumor growth inhibition and bone marrow studies, tumor-bearing nude rats were treated with placebo nanoparticles, AZD1152 (25 mg/kg daily for days 1 to 4; total dose 100 mg/kg), or Accurins E or F (25 mg/kg dosed on days 1 and 3; total dose 50 mg/kg). (A) Growth curves of SW620 tumors. Relative times of dosing points are shown. Data are mean tumor volumes ± SEM (n = 5). *P < 0.05, **P < 0.01, one-sided Student’s t test versus placebo. (B) Percent change in pHH3 was determined by FACS analysis of disaggregated tumors from rats treated with 25 mg/kg of AZD1152 or AZD2811 nanoparticles E and F (25 mg/kg). The percentage of pHH3+ cells is relative to placebo nanoparticle controls and is represented as a mean ± SEM (n = 3 to 5). (C) Bone marrow cellularity as a percentage of total nuclear cell count (by FACS analysis) of bone marrow aspirates taken at 5, 9, and 15 days after the first dose. Data are individual samples, one per animal (n = 3 to 5).

Retention of AZD2811 in human tumor xenografts

SW620 tumors excised from rats treated with AZD1152 or Accurin F were analyzed by imaging MS to characterize the intratumoral distribution of AZD2811 (parent drug still encapsulated in the nanoparticle or that had been released but not metabolized), an acid metabolite of AZD2811 [mass/charge ratio (m/z) 478.2 [M-H]-] (drug released from the particle locally or accumulated drug that was released in the circulation and metabolized), and the nanoparticle itself (detected as PLA, a polymer component of the nanoparticle) at various time points after dosing (Fig. 5A). In animals treated with AZD1152 prodrug, AZD2811 was detected in the tumor at 2 and 6 hours after dosing, but was undetectable at 24 hours (Fig. 5B), consistent with the PK profile of AZD1152. However, in animals treated with Accurin F on days 1 and 3, AZD2811 was detected across the entire tumor cross-section for up to 6 days after the last nanoparticle administration. Three platforms—liquid extraction surface analysis (LESA), matrix-assisted laser desorption/ionization (MALDI), and high resolution MALDI—each confirmed the presence of drug in the tumor (fig. S3), indicating that AZD2811 nanoparticles distribute active drug throughout the tumor for an extended period of time.

Fig. 5. MS imaging data collected from tumor tissue sections.

Tissues were analyzed by desorption electrospray ionization (DESI) and MALDI, and images are presented as heat maps of the relative abundance distribution of selected ions. (A) DESI-MS image for [M-H]- of AZD2811 at spatial resolution of 85 μm, displayed is m/z 506.2 [M-H]-; a metabolite at m/z 478.2 [M-H]-; and PLA polymer m/z 2034.6 (confirmed by polymer spacing). Yellow dotted line shows zoomed view of region. Analyzed after a single dose of 25 mg/kg. (B) MALDI-MS images for [M+H]+ of AZD2811 at a spatial resolution of 100 μm in tumor sections after a single AZD1152 dose (25 mg/kg) or after administration of Accurin F (25 mg/kg) on days 1 and 3. All MALDI images are displayed on same relative abundance scale. White dotted lines indicate tumor tissue section edges. Scale bars, 2 mm.

Efficacy of Accurin E in mouse xenograft models

To further explore the efficacy of AZD2811 nanoparticles, AZD1152 and Accurin E were compared in several mouse models of human tumors. In all studies, tumor-bearing mice were treated intravenously with Accurin E (25 mg/kg) on days 1 and 3, or daily doses of AZD1152 (25 mg/kg) on days 1 to 4. Treatment of nude mice bearing SW620 tumors yielded effects consistent with those seen in nude rats. Accurin E–treated animals displayed 96% tumor growth inhibition at half the dose of AZD1152 (65% tumor growth inhibition), with minimal impact on body weight (Fig. 6, A and B). Single doses of AZD1152 or Accurin E resulted in reductions in pHH3 in the tumor (Fig. 6C) and coincident induction in polyploid nuclei (Fig. 6D). The slower reduction in pHH3 with Accurin E (over the course of 96 hours) compared to AZD1152 (≈24 hours) is indicative of controlled release of AZD2811.

Fig. 6. The effects of AZD1152 and Accurin E in vivo in murine tumor xenograft models.

Tumor-bearing mice were treated with placebo nanoparticle, AZD1152 (25 mg/kg daily for days 1 to 4; total dose 100 mg/kg), or Accurin E (25 mg/kg dosed on days 1 and 3; total dose 50 mg/kg). (A) Growth curves of SW620 tumors in nude mice. Relative times of dosing points are shown. Data are mean tumor volumes ± SEM (n = 10 to 12). ***P < 0.001, one-sided Student’s t test versus placebo. (B) Changes in body weight of SW620-bearing mice relative to day 0. Dosing time course is same as in (A). Data are means ± SEM (n = 10 to 12). (C) Percent change in pHH3 was determined by FACS analysis of disaggregated SW620 tumors from nude mice treated once with AZD1152 or Accurin E (25 mg/kg). Data are mean percentages relative to placebo nanoparticle ± SEM. (D) SW620 tumors were analyzed for pHH3 and polyploidy, as well as tumor cell integrity by H&E staining from nude mice treated once with placebo nanoparticle or Accurin E (25 mg/kg). (E) Growth curves of OCI-LY19 tumors grown in CB17-scid mice or TMD8 tumors grown in NOD.CB17-scid mice. Relative times of dosing points are shown. Data are means ± SEM (n = 10 to 12). *P < 0.05, ***P < 0.001, one-sided Student’s t test versus placebo.

Accurin E was next tested in tumor xenografts derived from two diffuse large B cell lymphoma (DLBCL) cell lines, TMD8 and OCI-LY19. AZD1152 previously displayed signs of clinical activity in DLBCL, but was not advanced due to dosing and tolerability considerations (20). In animals bearing OCI-LY19 tumors, AZD1152 and Accurin E both regressed the tumor up to day 7 (Fig. 6E). At day 17, the tumor growth inhibition for AZD1152-dosed mice was 90% compared with 99% for mice receiving Accurin E. Tumor regrowth in this aggressive model occurred in both treatment groups beginning at day 17, with slower regrowth seen in tumors treated with Accurin E than AZD1152. In the TMD8 model, AZD1152 was only partially active (54% tumor growth inhibition at day 18), delivering brief tumor stasis before rapid regrowth (Fig. 6E). In contrast, Accurin E led to sustained inhibition of tumor growth (95% inhibition at day 18) and extended the time to tumor regrowth by more than 15 days.


Therapeutic agents targeting the cell cycle have shown encouraging antitumor activity in late-stage clinical trials but are limited by mechanism-related toxicity because they inhibit processes that are fundamental to the replication of all cells (7, 19). Compelling efficacy, such as the activity of AZD1152 in AML (13, 21), warrants the development of novel nanomedicines to improve the therapeutic index of these molecules. We therefore sought to maximize the therapeutic effect of Aurora B kinase inhibitors in tumors while sparing healthy tissues by using the Accurin nanomedicine platform to systematically vary drug release kinetics. Our study is the first to investigate the nanoencapsulation of an MTA targeting the cell cycle and makes use of in situ hydrophobic ion pairing to generate a diverse panel of nanoparticle formulations from which a candidate was selected for further interrogation in humans.

Initial attempts to encapsulate and control the release of AZD2811 were hampered by its poor solubility in benzyl alcohol, ethyl acetate, and other solvent systems suitable for use in nanoemulsion processes. The resulting nanoparticles displayed low drug loads and rapid release of payload with minimal sensitivity to formulation parameters. We therefore explored the use of pharmaceutically acceptable organic acid additives to increase the drug load and enable a broader range of nanoparticle release profiles. We found that incorporation of stoichiometric quantities of organic acids resulted in up to a fourfold increase in drug load along with an extension of the time to 50% release in vitro from less than 2 hours to as long as 72 hours. The rate and duration of AZD2811 release was dependent on the selection of the organic acid as well as other formulation and process parameters. We hypothesized that the observed effects are mediated by proton exchange and formation of ion pairs involving the AZD2811 and organic acid.

AZD2811 nanoparticle formulations spanning a fivefold range of release rates displayed slower onset and more prolonged inhibition of Aurora B kinase in tumors compared to the prodrug AZD1152, together with more effective tumor growth inhibition in multiple preclinical models when administered at one half the dose of AZD1152. Accurins E and F, which displayed slower release profiles, similarly inhibited tumor growth with minimal bone marrow toxicity. In contrast, AZD1152 and Accurin B were significantly toxic to bone marrow. These findings highlight the influence of nanoparticle release kinetics on therapeutic index and illustrate the use of hydrophobic ion pairing to modulate drug release across a wide range of release rates. Hydrophobic ion pairing agents have been used previously to facilitate encapsulation of hydrophilic drugs in nanoparticles (22, 23).

The preclinical data presented in this study establish the potential for nanoparticles encapsulating molecular targeted cancer agents with narrow therapeutic indices. Limitations of this work include the clinical relevance of xenografted animal models used to assess efficacy of nanoparticle formulations, and the degree to which improvements in therapeutic index seen in this study can be extrapolated to other MTAs. Although we selected a lead formulation using a tumor model (SW620) that supported the AZD1152 program—and, as such, we had extensive comparator data from which to benchmark the tolerability, PD, and efficacy of candidate nanoparticles—the model is subject to the known limitations of xenografted human tumor cell lines in assessing therapeutic candidates in oncology. Moreover, although rat bone marrow is commonly used to model myelotoxicity in humans, interrogation of the nanoparticle dose and schedule in patients may be required to achieve optimal clinical results. We have focused on bone marrow effects based on clinical experience with AZD1152; however, it is possible that PK and biodistribution of the nanoparticle formulation may result in other potential toxicities associated with inhibiting the cell cycle, such as gastrointestinal effects. Clinical studies will be required to understand these risks in more detail.

Our study showed that slow release is critical for the Aurora B kinase inhibitor, but other MTAs may require different release kinetics to overcome toxicity and maximize antitumor activity. For agents with ionizable functional groups, ion pairing can enable the evaluation of various release profiles to determine empirically which pharmacokinetics afford optimal results. For example, preliminary unpublished data suggest that this approach can substantially improve the preclinical tolerability and antitumor activity of the Akt inhibitor MK-2206, which targets the phosphatidylinositol 3-kinase pathway (24). Attempts to drug this pathway, which is up-regulated in many human cancers, have been largely unsuccessful due to therapeutic index limitations of conventional agents, including MK-2206.

The nanoparticle PK and biodistribution properties described here are well suited to improve the therapeutic index of cell cycle inhibitor payloads, including Aurora B kinase inhibitors, which have disappointed in the clinic largely due to dose-limiting bone marrow toxicity (14). The particles are retained in the vascular compartment and cleared very slowly while releasing their payload at a controlled rate. This results in an increased concentration of the payload at the tumor site and a reduction in the maximum systemic released drug concentration. Furthermore, whereas short-term exposure of Aurora kinase inhibitors leads to cell cytostasis, extended inhibition forces cells into mitotic catastrophe, leading to cell death (19, 25). Thus, the ability to exert prolonged PD effects is potentially desirable because rapidly proliferating tumor cells, such as cells with dysregulated cell cycle check points, are sensitive to Aurora kinase inhibition, as are cells with mutations or amplifications leading to up-regulation of myc expression (26). In addition, cell cycle agents are likely to give maximum benefit when used in combination. The improved bone marrow profile observed with slow-releasing nanoparticles may enable efficacious combination treatments with acceptable tolerability. In particular, combining Aurora B kinase inhibitors with DNA damaging agents [chemotherapy, radiotherapy, or poly(adenosine diphosphate–ribose) polymerase (PARP) inhibitors] that exploit the weakened polyploid DNA structures caused by Aurora kinase inhibition warrants investigation.

The AZD2811 nanoparticles identified in this study have the potential to increase efficacy at tolerable doses using a more convenient dosing regimen, which may in turn extend the utility of Aurora B kinase inhibition to a broader range of hematological and solid tumor cancer indications. Translational next steps include process development and manufacture of clinical supplies, investigational new drug (IND)–enabling toxicology studies in rodents and nonhuman primates, and finally initiation of a phase 1 clinical trial ( identifier: NCT02579226) to assess the safety, tolerability, and PK of AZD2811 Accurins in patients with advanced solid tumors.


Study design

The objective of this study was to develop and characterize a nanoparticle formulation of AZD2811 with increased therapeutic index relative to the parent drug via reduced toxicity and increased antitumor activity, and with extended release characteristics to enable more convenient dosing of AZD2811 to patients. All animal studies were conducted in accordance with UK Home Office legislation, the Animal Scientific Procedures Act 1986, and the AstraZeneca Global Bioethics policy. All experimental work is outlined in project license 40/3483, which has gone through the AstraZeneca Ethical Review Process. Studies in the United States were conducted in accordance with the guidelines established by the internal IACUC (Institutional Animal Care and Use Committee) and reported following the ARRIVE (Animal Research: Reporting In Vivo experiments) guidelines. Randomization of animals onto study was based on initial tumor volumes to ensure equal distribution across groups. A power analysis was performed whereby group sizes were calculated to enable statistically robust detection of tumor growth inhibition (more than six per group) or PD endpoint (more than four per group). Bone marrow effects and modulation of tumor biomarkers after drug treatment were evaluated using two independent measures, with expert pathologist input. MS-based analysis of nanoparticle and drug distribution in tumors was determined using three complimentary platforms at independent laboratories.

Preparation of AZD2811 nanoparticles

Accurins were produced using a nanoemulsion process shown schematically in fig. S4. For initial formulation work, nanoparticles were produced using a benchtop scale process (batch size, 250 to 500 mg of nanoparticles). For animal studies, batch sizes of 20 to 36 g were generated for Accurins B to F. For Accurin A, where only PK studies were planned, a batch size of 5 g was used. The larger-scale process used the same conditions and similar equipment as the small-scale process used for formulation screening. AZD2811 (2-[3-[[7-[3-[ethyl(2-hydroxyethyl)amino]propoxy]quinazolin-4-yl]amino]-1H-pyrazol-5-yl]-N-(3-fluorophenyl)acetamide) anhydrous free base; AstraZeneca) and PLA-PEG (Evonik Industries; part #100 DL mPEG 5000 3.5CE) were dissolved in an organic phase comprising benzyl alcohol and ethyl acetate and then dispersed in an immiscible aqueous phase to form a two-phase nanoemulsion. Nanoparticles containing organic acids except for DOSA were produced using a benzyl alcohol/ethyl acetate 20/80 organic phase solvent mixture and a 5% organic phase solids concentration. Nanoparticles containing DOSA used a benzyl alcohol/ethyl acetate 30/70 organic phase with a solids concentration of 30%. The emulsion was quenched by addition to cold water, resulting in extraction of solvent from the organic phase and hardening of nanoparticles. Subsequent processing included tangential flow filtration, which removes unencapsulated API and processing aids, and addition of aqueous sucrose solution for stabilization and cryopreservation of the nanoparticle suspension. AZD2811 nanoparticles were characterized with respect to particle size, AZD2811 load, and in vitro release kinetics (Supplementary Methods).

Bioanalysis of AZD2811 in plasma

Blood samples (20 μl) were taken from the tail vein using a capillary tube, diluted in 80 μl of phosphate-buffered saline (PBS), and centrifuged at 13000 rpm for 3 min at 3° to 4°C. Plasma was removed and stored at −80°C before analysis. Subsequent steps were carried out on ice. Plasma samples (50 μl) were diluted using an appropriate dilution factor. Acetonitrile (150 μl) was added with the internal standard, and samples were clarified by centrifugation at 4500 rpm for 10 min. Supernatant (50 μl) was then diluted in 300 μl of water and analyzed via LC-MS/MS against an 11-point standard calibration curve (1 to 10,000 nM). The LC-MS/MS conditions are in tables S3 and S4.

Tumor growth inhibition and PD studies in vivo

All animals included on studies were greater than 6 weeks old at the time of cell implant (TMD8, OCI-LY19, or SW620; Supplementary Methods). Tumor growth was monitored twice weekly by bilateral caliper measurements. Tumor volumes [(3.142 * MAX(len:wid) * MIN(len:wid) * MIN(len:wid))/6000] and tumor growth inhibition [((RTV(Control) − RTV(Treatment)) * 100)/(RTV(Control) − 1)] were calculated, where RTV is the geometric mean for control and treatment groups, which was calculated using final tumor volume/initial tumor volume for individual animals.

For studies in SW620 athymic (Hsd:RH-Foxn1rnu) male rats and OCI-LY19 severe combined immunodeficient female mice (CB17 scid), 1 × 107 cells in 50% Matrigel were inoculated subcutaneously on the left flank of the animal. For SW620 athymic (nu/nu:Alpk) mouse studies, 1 × 106 cells in 50% Matrigel were inoculated subcutaneously on the left flank of the animal. For studies in TMD8 tumors, 1 × 107 cells in 50% Matrigel were inoculated subcutaneously on the left flank of female NOD.CB17-scid mice (Charles River Laboratories).

For acute PD studies, nude rats bearing SW620 xenografts were dosed intravenously with a single bolus dose of 30% sucrose, AZD1152 (25 mg/kg), or Accurins B to E containing a single dose equivalent of AZD2811 (25 mg/kg). Tumors were excised postmortem at specified time points and frozen for analysis of pHH3. For PD measurements accompanying tumor growth inhibition studies, tumors were excised postmortem at specified time points after the first dose, fixed in 10% buffered formalin for 24 to 48 hours, and then processed to paraffin block. Femurs were fixed in 10% buffered formalin for 24 to 48 hours, decalcified in 10% formic acid (Sigma), and processed to paraffin block, or bone marrow was flushed from the femur using 50% fetal calf serum/50% PBS and analyzed by flow cytometry.

pHH3 analysis

Frozen SW620 tumors were disaggregated in cold PBS using Medicon Technology and fixed in 80% ethanol. Resulting cell suspensions were blocked with 1% bovine serum albumin (BSA) before incubation with primary antibody to pHH3 (Millipore) and subsequent incubation with a fluorescein isothiocyanate (FITC)–conjugated anti-rabbit IgG (H+L) secondary antibody (Millipore). Finally, cells were incubated with ribonuclease (1 mg/ml) (Sigma) and propidium iodide (0.4 mg/ml) (Sigma) and analyzed by flow cytometry. Activity was measured as an inhibition of histone H3 phosphorylation on Ser10. Average numbers of pHH3 positive cells in G2/M were calculated for each treatment group and compared to the pHH3 level in G2/M phase cells from the placebo nanoparticle group.

Bone marrow analysis

Nucleated bone marrow cells were analyzed using a method adapted from (27). The bone marrow cell suspensions were filtered through a 100-mm disposable filter device (Filcons, Dako), then underlayered with 1 ml of fetal bovine serum (Sigma), and centrifuged at 300g for 5 min at 4°C. The cell pellet was resuspended in 4 ml of ice-cold PBS containing 0.5% BSA. FITC-conjugated mouse anti-rat CD45 (5 ml) and 10 μl of phycoerythrin (PE)–conjugated mouse anti-rat CD71 monoclonal antibodies (Serotec) were added to 100 μl of adjusted bone marrow cell suspension, mixed well, and incubated on ice in the dark for 20 min. Cells were washed with ice-cold PBS containing 0.5% BSA and recentrifuged. The resulting cell pellet was resuspended in 0.4 ml of ice-cold PBS containing 0.5% BSA, and then 20 μl of LDS-751 staining solution (Molecular Probes) was added and kept in the dark for 20 min before flow cytometric analyses. Sample analysis was performed using FACSDiva.

White blood cell analysis

Dipotassium EDTA anticoagulated peripheral whole blood samples were analyzed on an Advia 2120i automated hematology analyzer with veterinary software (Siemens). Total white blood cell counts were calculated for each group.


Sections (4 μm) were deparaffinized with xylene and rehydrated through graded alcohols into water. Antigen retrieval was carried out in a Milestone RHS microwave rapid histoprocessor for 5 min at 110°C in pH 6 citrate buffer (Dako). Tissues were placed on a Lab Vision Autostainer, and endogenous peroxidase was blocked with 3% H2O2 for 10 min, followed by washing twice in tris-buffered saline/0.05% Tween (TBS-T). For pHH3, serum-free protein block (Dako; X0909) was applied for 15 min before incubation with primary antibody (Upstate Biotechnology, 1:1000 dilution) for 1 hour. For detection of pHH3, sections were incubated for 30 min with Rabbit EnVision polymer detection system (Dako). All samples were developed in liquid 3,3′-diaminobenzidine (DAB; Dako) for 10 min. Sections were then counterstained with Carazzi’s hematoxylin, dehydrated, cleared, and mounted with coverslips. All washes were performed in TBS-T, and all steps were conducted at room temperature. Tumor and femur were scored visually by a pathologist for levels of pHH3 or a reduction in bone marrow cellularity using a scoring system, where 0 = no change, 1 = minimal, 2 = mild, 3 = moderate, 4 = severe change compared to placebo nanoparticle–treated animals.

Statistical analysis

Statistical analysis of tumor growth inhibition studies uses a comparison of Log(RTV(Control)) to Log(RTV(Treatment)), with a one-sided Student’s t test and pooled variability across all groups. The statistical test uses a 5% significance level, and the test is one-sided because only a reduction in tumor volume compared to control is relevant.



Fig. S1. PD of AZD2811 nanoparticles in vivo in SW620 tumor xenograft models.

Fig. S2. Total nucleated cells from bone marrow in nude rats.

Fig. S3. MS reveals relative abundance of [M+H]+ metabolite.

Fig. S4. Process flow diagram for the generation of Accurin nanoparticles.

Table S1. AZD2811 solubility characteristics.

Table S2. Characteristics of AZD2811 nanoparticles produced during formulation screening.

Table S3. LC-MS/MS parameters.

Table S4. Optimized parameters for MS analyses.


  1. Acknowledgments: We thank R. Langer for his critical review of the manuscript and helpful suggestions, N. Floch and V. Campbell for help with figure formatting, M. Dymond for statistical support, and staff in Laboratory Animal Sciences and Drug Safety and Metabolism Alderley Park and Nanhua Deng Gatehouse Park for technical support. Funding: This work was funded by AstraZeneca. Author contributions: P.J.J. and S.T.B. performed conception and design of research and writing the manuscript; M.A., J.H., and J.M. conceived and designed the research; S.A. performed animal study design, analysis and interpretation, and writing the manuscript; E.C., R.O., P.T., C.L.R., and D.T. performed animal study conduct, design, analysis, and interpretation; J.F. and P.A.H. performed pathology analysis and interpretation; R.E. and U.M.P. performed flow cytometry analysis and interpretation; J.W., C.H., A.S., M.W., and S.L. performed bioanalysis and PK analysis; R.J.A.G., J.G.S., N.S., Z.T., A.N., and P.A. performed imaging MS design, analysis, and interpretation; Y.H.S., J.N., G.T., and D.D.W. performed nanoparticle formulation and process development; D.P. led analytical method development and nanoparticle characterization; S.Z. performed data interpretation and writing the manuscript. Competing interests: S.A., E.C., J.M., R.O., J.F., P.A.H., P.T., R.E., U.M.P., J.W., C.H., A.S., R.J.A.G., J.G.S., D.T., M.W., C.L.R., M.A., P.J.J., and S.T.B. are current or former AstraZeneca employees and shareholders. Y.H.S., S.L., G.T., J.N., D.D.W., D.P., J.H., and S.Z. are current or former BIND employees and shareholders. Results reported herein are disclosed in PCT Patent Application WO2015/036792. Data and materials availability: All reasonable requests for collaboration involving materials used in the research will be fulfilled provided that a written agreement is executed in advance between AstraZeneca or BIND Therapeutics Inc. and the requester (and his or her affiliated institution). Such inquiries or requests for additional data should be directed to the corresponding authors.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article