Natasha Bray

Biologics: Transferrin' bispecific antibodies across the blood-brain barrier.

a notoriously difficult obstacle for therapeutic agents — particularly large molecules such as antibodies — that have targets within the brain. Antibodies that inhibit β‐secretase (BACE), which cleaves amyloid precursor protein into amyloid‐β (Aβ), could be useful as a treatment for Alzheimer’s disease, which is characterized by deposits of Aβ in the brain. Now, Watts and colleagues demonstrate that bispecific antibodies that target both BACE and the transferrin receptor (TFR), which is expressed on endothelial cells of the BBB, can cross into the brain in mice and in non‐human primates. First, the authors generated a humanized TFR‐specific antibody (anti‐TFR1) that binds to the receptor at a site distant from its ligand‐ binding sites. A modified version of anti‐TFR1 that had 20–25‐fold lower affinity for TFR — anti‐TFR2 — was also created to assess the impact of TFR affinity on brain uptake. Next, using ‘knobs into holes’ technol‐ ogy (in which a large amino acid is substituted for a small one in the heavy chain of one half‐antibody to create the ‘knob’ and vice versa in another half‐antibody to create the ‘hole’), the authors created two types of heterodimeric antibodies, anti‐ TFR1–BACE and anti‐TFR2–BACE, each of which had one anti‐TFR arm and one BACE‐specific arm. To test whether these antibodies could cross the BBB, mice that expressed human TFR were given an intravenous injection of anti‐ TFR1–BACE or anti‐TFR2–BACE. Both types of antibody led to a 55–65% reduction of Aβ levels in the plasma and in the brain of these animals, demonstrating that BACE function was indeed inhibited. Interestingly, the lower‐affinity anti‐ TFR2–BACE showed higher brain concentrations than anti‐TFR1–BACE; moreover, only the high‐affinity anti‐ body significantly reduced TFR pro‐ tein levels 1 day after dosing (owing to antibody‐triggered TFR degradation), indicating that anti‐TFR2–BACE has more desirable therapeutic properties. In cynomolgus monkeys, each anti‐TFR–BACE induced a >50% reduction in plasma Aβ levels after 24 hours. Moreover, injection of anti‐TFR1–BACE and anti‐TFR2– BACE reduced Aβ concentrations in the cerebrospinal fluid by ~50% and ~20%, respectively, after 48 hours. By contrast, BACE‐specific anti‐ bodies showed very little uptake into the brain and had no impact on brain Aβ levels. Importantly, there was no effect of either anti‐TFR–BACE on the number of reticulocytes (which also express TFR and thus presented a safety concern) — and, unlike in the TFR knock‐in mice, TFR levels in the brain were unaffected by these antibodies. This study demonstrates the use of bispecific antibody technology to enable therapeutic antibodies to cross the BBB via receptor‐mediated endocytosis. In principle, the BACE‐ specific arm of anti‐TFR–BACE could be replaced with antigen‐binding regions that could neutralize other target molecules on the other side of the BBB. Natasha Bray B I O LO G I C S

Lung disease: Resetting the redox balance in lung fibrosis

pulmonary fibrosis (IPF) are strongly associated with ageing, but the underlying mechanisms are unclear. Hecker and colleagues now report that a loss of redox homeostasis in aged mice contributes to the persistence of senescent — but apoptosis-resistant — myofibroblasts, which may explain the reduced ability of these animals to resolve fibrosis. Moreover, the authors show that inhibition of NADPH oxidase 4 (NOX4) — an enzyme that generates reactive oxygen species — reverses this myofibroblast phenotype and prolongs survival in an aged mouse model of persistent lung fibrosis. Bleomycin-induced lung injury in young (2-month-old) mice triggers a normal fibrotic response, which peaks at 3 weeks post-injury and then self-resolves, in part through apoptosis of myofibroblasts. However, in aged (18-month-old) mice the fibrotic response persists for months and lung myofibroblasts continue to accumulate and produce extracellular matrix, leading to IPF-like symptoms. Using this ageing model of persistent fibrosis, the authors showed that after bleomycin injury, lung fibroblasts from aged mice exhibit increased levels of senescence markers and B cell lymphoma 2 (BCL-2), an anti-apoptotic marker, compared with fibroblasts from young mice. Moreover, the authors demonstrated that the senescent, apoptosis-resistant myofibroblast phenotype observed in aged mice was mediated by a redox imbalance. This imbalance was associated with deficient activation of the transcription factor NFE2-related factor 2 (NRF2) — a master regulator of antioxidant genes — and sustained activation of NOX4. This previously unknown profibrotic mechanism may help to explain why IPF develops more frequently in older individuals. Hecker et al. then investigated NOX4 as a possible therapeutic target by treating aged mice with intranasal NOX4-targeting small interfering RNA (siRNA) on alternate days between weeks 3 and 6 after bleomycin injury. Compared with lung fibroblasts from controls, fibroblasts from anti-NOX4 siRNAtreated animals exhibited markedly reduced levels of BCL-2 and senescence markers, and produced less collagen α1 (a major component of fibrotic scar tissue). Finally, animals were treated daily in weeks 3 to 6 after bleomycin exposure with oral doses (40 mg per kg) of the NOX1 and NOX4 dual inhibitor GKT137831. Compared with controls, lungs harvested from animals treated with GKT137831 6 weeks after bleomycin injury exhibited fewer senescent myofibroblasts. Furthermore, unlike the vehicle-treated controls, GKT137831-treated mice recovered their baseline body weight within the 6-week observation period. In addition, collagen levels in the lungs of GKT137831-treated animals were similar to uninjured mice and, importantly, more animals treated with GKT137831 survived the 6 weeks following bleomycin injury than did vehicle-treated animals (~95% compared with ~75%). Taken together, these data indicate that NOX4 inhibition using compounds such as GKT137831 may reset the redox balance and thus promote the susceptibility of senescent myofibroblasts to apoptosis. Therefore, NOX4 inhibitors such as GKT137831 (which is currently in Phase II trials for diabetic nephropathy) might be useful against age-associated fibrotic diseases such as IPF.

Cardiovascular disease: PDE9A inhibition mends broken hearts

preserved ejection fraction (HFPEF) could be treated by promoting intracellular signalling by the secondary messenger cyclic GMP, which has cardioprotective effects. However, blocking cGMP breakdown with the phosphodiesterase 5A (PDE5A) inhibitor sildenafil was not successful in clinical trials for HFPEF, despite showing preclinical promise. Kass and colleagues now provide an explanation for this observation, and show that inhibition of PDE9A could be more effective. Increases in cGMP levels are triggered by two independent factors: nitric oxide (NO) and atrial natriuretic peptide (ANP). PDE5A hydrolyses NO-coupled cGMP, but NO signalling is downregulated in heart disease, which might help to explain the lack of efficacy of PDE5A inhibition. Which PDEs are responsible for degrading cGMP resulting from ANP stimulation was not clear, and so the authors investigated the activity of the other cGMP-selective PDE, PDE9A, in the heart. The authors found that PDE9A was expressed in myocardial tissue from mice and from humans, and was increased in the myocardium of patients with various forms of heart failure, especially HFPEF. Interestingly, PDE9A and PDE5A were localized in different compartments of the sarcomere in myocytes, suggesting that they may have distinct functions. To determine how the function of PDE9A might differ to that of PDE5A, the authors measured levels of cGMP in rat neonatal cardiomyocytes (RNCMs) that expressed a fluorescent cGMP indicator. cGMP levels increased in RNCMs stimulated with ANP or an NO donor. Interestingly, after stimulation with ANP, but not after stimulation with the NO donor, cGMP levels were further increased by the selective PDE9A antagonist PF-9613, indicating that PDE9A degrades ANP-dependent cGMP but not NO-dependent cGMP. In the transverse aortic constriction (TAC) model of heart failure, Pde9a–/– mice showed less cardiac hypertrophy, lower expression of genes associated with myocardial pathology and better ventricular function at 3 weeks after TAC surgery than did wild-type TAC controls. Wild-type TAC mice exhibited heart hypertrophy and ventricular dysfunction 8 days after the surgery, and these symptoms were improved by 4 weeks of treatment with PF-9613 or the selective PDE5A antagonist sildenafil. However, TAC mice given the NO synthase inhibitor L-NAME — to mimic the clinical scenario in HFPEF in which NO production is suppressed — showed improvements in heart function and increases in myocardial cGMP levels and downstream PKG activity only in response to PF-9613, and not sildenafil. Together, these findings confirm that PDE5A and PDE9A target cGMP in the NOand ANPsignalling pathways, respectively. Overall, this study shows that PDE9A, unlike PDE5A, specifically hydrolyses NO-independent, ANP-coupled cGMP. So, selective PDE9A inhibitors such as PF-9613 could have greater effectiveness than PDE5A inhibitors for treating cases of HFPEF in which NO production is lower. Natasha Bray C A R D I OVA S C U L A R D I S E A S E

Neurodegenerative diseases: A tale of two taus in traumatic brain injury.

major risk factor for developing chronic traumatic encephalopathy (CTE) or Alzheimer disease (AD) — both of which are characterized by the formation of neurofibrillary tangles containing hyperphosphorylated tau (P-tau). How TBI triggers this long-term tauopathy is not known. Now, Kondo et al. demonstrate a potential molecular basis for this link: in TBI mice, an antibody targeting the cis isoform of P-tau can protect against the development of tauopathy and synaptic and behavioural deficits. Previously, the group showed that, in AD, proline isomerase 1 can inhibit the progression of tauopathy by changing the configuration of the phosphorylated Thr231-Pro motif in P-tau from cis to trans. They developed polyclonal antibodies specific for the cis or trans form of P-tau to show that whereas trans P-tau is physiological, the cis form is an early pathogenic conformation that leads to tauopathy. In this study, the authors further developed monoclonal antibodies (mAbs) that were able to specifically detect and eliminate the cis or trans form of P-tau — cis mAb and trans mAb, respectively. To investigate the expression of cis P-tau and trans P-tau after TBI, the authors assessed antibody labelling of neurons in the brains of mice that underwent TBI. Whereas TBI had no effect on the level of trans P-tau in the brain, it increased the amount of cis P-tau in injured axons after ~12 hours, with increases in levels lasting longer (more than 6 months after injury) in the brains of mice with repeated mild TBIs or a single severe TBI than in the brains of mice with a single mild TBI. Intriguingly, cis P-tau expression was more widespread throughout the brain 6 months after TBI than at 2 months post-injury. To investigate the possibility that cis P-tau released by injured cells could enter neighbouring cells, the authors collected media from hypoxia-stressed cells that expressed green fluorescent protein (GFP)-tagged tau. Naive neurons incubated with this stressedcell media — but not media from stressed cells incubated with cis mAb — showed uptake of the fluorescent tau and neuronal damage, indicating that cis P-tau expressed by stressed or injured neurons may ‘spread’ to nearby neurons and have neurotoxic effects. Next, the authors tested the effects of peripheral injections of cis mAb starting before or just after a single severe TBI in mice. Two weeks’ treatment with cis mAb eradicated cis P-tau in the brain, prevented axonal pathology and apoptosis, and restored cortical long-term potentiation. In the elevated plus maze test of risk-taking behaviour, TBI mice treated with cis mAb for 2 months showed control levels of risk-taking behaviour; by contrast, TBI mice treated with immunoglobulin G spent more time in the exposed arms of the maze. Moreover, 6 months’ treatment of TBI mice with cis mAb protected against the development of tauopathy and brain atrophy. These findings support a role for cis P-tau in the development of tauopathy and neurodegeneration after TBI. Interestingly, the authors also found that cis P-tau co-localized with tau pathology in neurons of brains of individuals with CTE, but not in control brains, indicating that antibody-targetable cis P-tau may also be involved in CTE in humans.

Analgesia: Anti-itch and anti-ouch antibody

Nav1.7 plays a key part in the sensation of pain, but past efforts to develop analgesics that target this channel have often not achieved high enough Nav subtype selectivity to avoid adverse off-target effects. In a recent paper in Cell, Lee et al. developed a Nav1.7-specific antibody called SVmab1, which not only reduces inflammatory and neuropathic pain in mouse models but also reveals a role of Nav1.7 in itch. The authors produced a monoclonal antibody against the voltagesensor paddle of human Nav1.7 — a region that modulates channel gating and shows high variation among Nav subtypes. Patch-clamp experiments showed that, unlike a control antibody (CTmab) against a Nav1.7 region not involved in gating, SVmab1 reduced the size of sodium currents through cell-lineexpressed Nav1.7 channels by ~87%, with a half-maximal inhibitory concentration of 30 nM. Moreover, these effects were specific to Nav1.7 over other Nav subtypes; besides Nav1.7, SVmab1 only partially inhibited Nav1.6. Voltage-clamp recordings from mouse dorsal root ganglion neurons revealed that the inhibitory effects of SVmab1 on sodium currents were specific to small-sized nociceptive neurons. In spinal cord slices, the frequency of excitatory postsynaptic currents (EPSCs) in interneurons that receive input from nociceptive neurons was markedly reduced by SVmab1. Furthermore, in slices taken from mice 4 days after sciatic nerve ligation, EPSC frequency was increased, but reduced to normal levels by SVmab1; thus, SVmab1 reduces excitatory transmission in pain-sensitive neurons. The authors tested the analgesic properties of SVmab1 in mouse models of inflammatory and neuropathic pain. Intraplantar injection of formalin induces paw inflammation and pain-associated behaviour. Unlike CTmab, both intrathecal and systemic application of SVmab1 prevented licking and flinching behaviours after formalin injection, and systemic injection of SVmab1 reduced formalin-induced paw oedema. Similarly, SVmab1 — but not CTmab — increased the mechanical pain threshold in mice with chronic nerve constriction injury, with analgesia lasting ~24 hours after SVmab1 injection. Finally, the authors investigated the role of Nav1.7 in the sensation of itch. Compared with CTmab-treated animals, mice treated with intrathecal SVmab1 exhibited less scratching behaviour after intradermal injection of itch-inducing agents and after spinal injection of gastrin-releasing peptide, which evokes itch sensation via spinal cord neurons. Likewise, scratching behaviour in chronic itch models of dry skin and allergic contact dermatitis was reduced by intrathecal or intravenous application of SVmab1. Compared with healthy controls, spinal cord neurons from mice with dry itchy skin exhibited higher-frequency EPSCs that were suppressed by SVmab1, implying that itch increases Nav1.7-mediated excitatory activity in the spinal cord. This study indicates that specific inhibition of peripheral and central Nav1.7 channels may be a viable strategy for treating inflammatory or neuropathic pain, as well as itch-related disorders. The authors’ approach for developing a channelspecific antibody might also be applicable to other voltage-gated cation channels of therapeutic interest.

Antibacterial drugs: Antibody-antibiotic conjugate tracks down hidden S. aureus.

resides outside host cells, but it can also infect cells such as phagocytes and thus disseminate around the body and cause chronic or recurrent infections in tissues distant from the initial site of infection. The antibiotics that are currently used to treat S. aureus infections only target extracellular, planktonic bacteria, so the effects of targeting intracellular S. aureus are not known. In a new study, Lehar et al. develop an antibody– antibiotic conjugate (AAC) that binds to the S. aureus cell wall and releases the antibiotic moiety upon entry into host cells. First, the authors compared the infectivity of intracellular methicillin-resistant S. aureus (MRSA) with that of planktonic bacteria. In medium containing vancomycin (one of the antibiotics used to treat MRSA infections), planktonic MRSA was rapidly sterilized, whereas MRSA in peritoneal cells survived and could even infect co-incubated osteoblasts. Mice injected with MRSA-infected immune cells developed higher bacterial burdens than did mice injected with an equivalent dose of free bacteria. These findings indicate that bacteria can ‘escape’ antibiotics by hiding in host cells, and that these intracellular bacteria can thus be more virulent than free bacteria. To combat these hiding bacteria, the authors designed an AAC containing an antibody that would bind to the bacterium, and an antibiotic that would be released only after entry of the opsonized bacterium into a mammalian cell. The authors screened a panel of 40 patient-derived S. aureus-specific antibodies against several strains of S. aureus; an antibody specific for sugar modifications of wall-techoic acid (WTA; a type of glycopolymer on the outer layer of Gram-positive bacteria) showed binding to all tested strains. For the antibiotic moiety, the authors used a rifamycin derivative (a rifalogue), which could kill dormant antibiotic-resistant bacterial cells (known as ‘persister’ cells), non-replicating but viable cells, and intracellular MRSA. To allow intracellular release of the rifalogue from the AAC, the authors attached it to the WTA-specific antibody using a linker that could be cleaved by the lysosomal enzyme cathepsin D. The AAC, but not the antibody alone or a non-cleavable version of the AAC, killed bacteria in macrophages, suggesting that the construct worked as intended. Moreover, unlike a mixture of rifampicin and the antibody alone, the AAC markedly reduced the transfer of bacteria from infected peritoneal cells to non-infected osteoblasts in the presence of vancomycin. Mice injected with vancomycin and intracellular MRSA developed colonies of the bacteria in the brain, but these colonies did not develop in mice treated with vancomycin and the AAC. Furthermore, a single dose of the AAC 24 hours after infection prevented colonization of the kidneys, whereas treatment with vancomycin, the unconjugated rifalogue, the unconjugated WTA-specific antibody or the non-cleavable AAC beginning at the same time point did not. Overall, this study shows that targeting intracellular S. aureus — for example with this AAC — may be a clinically relevant and effective way of preventing colonization and reducing the risk of developing chronic infections. Natasha Bray A N T I B AC T E R I A L D R U G S

Screening: A platform for probing protein-aggregation inhibitors

by the aggregation of proteins; for example, type II diabetes is associated with the formation of human islet amyloid polypeptide (hIAPP) amyloid fibrils. One strategy to prevent such aggregation is to use small molecules that bind specifically with the protein precursor before it misfolds and aggregates; however, screens to identify such small molecules are challenging and require large amounts of precursor protein. Saunders et al. now introduce a new Escherichia coli-based platform that can predict the potential of different proteins to aggregate, and use this system to identify specific small-molecule inhibitors of hIAPP aggregation. First, the authors developed a way of testing the propensity of different test proteins to aggregate. They linked the gene sequence encoding the test protein between the sequences that encode the two domains of the bacterial β-lactamase (β-la) TEM1, which localizes in the space between the inner and outer membranes of the bacterium. If the test protein sequence folds normally here, the two β-la domains are brought together to form the functional enzyme, rendering the bacterium resistant to β-lactam antibiotics such as ampicillin. By contrast, if the test protein sequence aggregates, the β-la domains do not join, and the bacterium is susceptible to β-lactam antibiotics. The authors confirmed the validity of this system by measuring the antibiotic resistance of E. coli carrying sequences encoding β-la constructs that contained test proteins known to have high aggregation propensities (hIAPP, amyloid-β1–40 (Aβ1–40) and Aβ1–42) or rat IAPP (rIAPP), which does not aggregate into amyloid fibrils. Whereas E. coli expressing β-la–hIAPP, β-la–Aβ1–40 or β-la–Aβ1–42 were all susceptible to 80 μg per mL ampicillin, bacteria expressing β-la–rIAPP or a control sequence were relatively resistant. Measures of antibiotic resistance correlated with the propensity of each protein to aggregate in vitro (assessed using biophysical methods). Next, the authors tested a panel of 20 compounds known to have different abilities to prevent aggregation of hIAPP in this system. Indeed, the compounds known to strongly inhibit hIAPP aggregation dose-dependently restored the resistance of β-la–hIAPP-expressing E. coli to ampicillin, whereas the compounds with no ability to prevent aggregation did not. To find new inhibitors of hIAPP aggregation, 59 additional compounds were tested in a miniaturized version of the E. coli system (to reduce the amount of test compound needed to ~0.07 mg). The 59 compounds included 31 compounds known to inhibit aggregation of other proteins, and 28 compounds that were computationally selected for their structural similarity to a known inhibitor of hIAPP called JCS-1. Of the 59 test compounds, 6 compounds moderately or strongly rescued the resistance of β-la–hIAPP-expressing E. coli to ampicillin; the ability of the strongest hit, dopamine, to limit hIAPP aggregation was confirmed using biophysical methods. This platform could potentially be adapted to screen small molecules for inhibitors of the aggregation of many different proteins, using only small amounts of screening compounds, and without the need to produce large amounts of test protein. Natasha Bray S C R E E N I N G

Obesity: Reversing resistance to leptin in obesity.

leptin acts mainly in the hypothalamus to promote satiety and increase energy expenditure. Efforts to treat obesity by supplementing leptin levels have been unsuccessful, however, as obesity is associated with a loss of sensitivity to leptin, despite high levels of the hormone in circulation. Now, Ozcan and colleagues describe the identification and effects of celastrol, which reinstates leptin sensitivity in diet-induced obese (DIO) mice. Endoplasmic reticulum (ER) stress in the brain can lead to leptin resistance and thus obesity, so the authors sought to identify compounds that could reinstate ER homeostasis. They analysed the changes in gene expression that were induced in six models of reversal of ER stress in mouse liver and hypothalamus, and used the Connectivity Map (CMAP) database to identify compounds that induced similar changes in the expression of analogous human genes. Of these compounds, celastrol, which is extracted from the ‘thunder god vine’ Triptergium wilfordii, most strongly effected gene expression changes associated with return to ER homeostasis. Intraperitoneal injection or oral gavage of celastrol for 3 weeks led to substantial loss of body weight (mainly from fat) in DIO mice, but not in lean mice. The weight-loss effect was associated with an initial drastic reduction in food intake that gradually normalized as body weight and plasma leptin levels lowered. The effects of celastrol were not seen in ob/ob obese mice (which lack leptin) or db/db obese mice (which lack the leptin receptor), suggesting the drug promotes sensitivity to leptin. Moreover, cela strol enhanced the effects of leptin injection on reducing food intake in ob/ob mice; thus, cela strol potentiates the satiety-inducing effects of leptin. The respiratory exchange ratio of celastrol-treated DIO, lean or ob/ob mice during the dark phase was reduced, which indicates that these animals were metabolizing more fat than were vehicle-treated controls. Furthermore, the compound also ameliorated histological signs of hepatic steatosis, lowered plasma cholesterol and improved glucose homeostasis in DIO mice. The hypothalamus of DIO mice 15 hours after treatment with celastrol showed increased levels of phosphorylated signal transducer and activator of transcription 3 (STAT3), which is involved in leptin signalling, indicating that celastrol affects this central pathway. The drug also reduced levels of phosphorylated eukaryotic translation initiation factor 2α kinase 3 (PERK; a sign of ER stress) in the hypothalamus of DIO mice, confirming the ER homeostasis-stabilizing effects of the compound. Although the exact molecular mechanism of the actions of celastrol remains unknown, these results show that celastrol reduces hypo thalamic ER stress and re-establishes hypothalamic sensitivity to leptin in obesity. Moreover, the ‘reverse-engineered’ approach used to identify celastrol could be useful for pinpointing other drugs that induce characterized gene expression profiles. Natasha Bray O B E S I T Y

Fibrotic disease: Fixing a feedback loop in fibrosis

growth factor‐β (TGFβ) is crucial in tissue repair, but in fibrotic diseases, TGFβ signalling is chronically upregulated, leading to overproduc‐ tion of collagen and extracellular matrix by activated fibroblasts. Now, Palumbo‐Zerr and colleagues in the Distler group show that this may be due to a broken regulatory loop involving the nuclear receptor NR4A1, and that cytosporone B (Csn‐B; a known selective NR4A1 transcriptional agonist) can ameliorate several mouse models of fibrosis. First, the authors found that levels of NR4A1 mRNA were higher in fibrotic skin of patients with systemic sclerosis than with healthy skin, and showed that TGFβ signalling induces NR4A1 expression in dermal fibroblasts. In mice overexpressing a consti‐ tutively active TGFβ receptor type I (TBRI mice; a model of fibrosis), as well as in two other mouse models of skin fibrosis and a model of lung fibrosis, genetic deficiency of NR4A1 exacerbated fibrotic symptoms, including increased TGFβ signalling and numbers of fibroblasts. The authors demonstrated that NR4A1 recruits complexes that contain SP1 and histone deacetylase 1 to epigenetically silence TGFβ target genes, such as those encoding collagen. Together, these data indicate that NR4A1 usually downregulates TGFβ signalling, and NR4A1 deficiency exacerbates TGFβ‐associated fibrosis. Interestingly, fibroblasts chronically or repeatedly exposed to TGFβ showed an initial increase followed by a reduction in levels of NR4A1. By contrast, levels of phosphorylated NR4A1 (pNR4A1) — which cannot bind SP1 — rose steadily in fibroblasts stimulated with TGFβ, and were elevated in TBRI mice and in fibrotic tissues of animals in other models of fibrosis. These findings suggest that long‐term exposure to TGFβ — as in fibrosis — may induce NR4A1 phosphorylation and thus impair NR4A1‐mediated repression of TGFβ target genes. Finally, the authors investigated whether activating NR4A1 with Csn‐B could reduce fibrosis. In TBRI mice, but not in Nr4a1–/– TBRI mice, Csn‐B (injected intraperito‐ neally) lowered TGFβ‐target‐gene expression, collagen production, dermal thickening and myoblast differentiation. Csn‐B ameliorated fibrosis in models of skin, pulmo‐ nary, hepatic and renal fibrosis, even when the drug was administered after fibrosis had been established. Substantial efforts have been made to target TGFβ signalling in disease — particularly in ca ncer — but have often been limited by the challenges of achieving a viable therapeutic window, given the complex, ‘double‐edged’ nature of TGFβ signalling. Overall, this study indicates that NR4A1 agonists could provide a new way to therapeutically intervene in a TGFβ regulatory loop that is impaired in fibrosis, but potential side effects will need careful consideration.

Anticancer drugs: On-site CAR parking

against cancer cells by transducing T cells from patients with chimeric antigen receptors (CARs) that are specific for antigens expressed by tumours has recently resulted in dramatic clinical responses for some blood cancers. However, a major issue in extending the approach to solid tumours is that if the antigen is not specific enough to the tumour, there is a risk of CAR–T cell toxicity to non-tumour tissues. Johnson et al. now describe a CAR–T cell therapy that targets epidermal growth factor receptor variant III (EGFRvIII), an oncogenic mutant of EGFR that is specifically expressed in ~30% of glioblastomas. First, the authors created CARs based on EGFRvIII-specific singlechain variable fragments (scFvs). When mice were co-injected daily with temozolomide (a drug used to treat glioma), EGFRvIII-specific CAR–T cells induced rapid regression of EGFRvIII-positive glio blastoma; in particular, a CAR containing the murine 3C10 scFv substantially reduced tumour burden after just 7 days and so was investigated further. The authors created eight humanized versions of the 3C10 scFv and picked one, 2173, for its favourable selectivity for EGFRvIII over the wild-type EGFR (EGFRwt) — a quality that suggested it might elicit fewer off-tumour effects. To investigate this possibility, the authors tested the 2173-CAR–T cells versus T cells expressing CARs based on the EGFR-specific antibody, cetuximab (cetux), which binds to EGFRwt and EGFRvIII with equal affinity. Cetux-CAR–T cells proliferated, produced type I cytokines and caused lysis in equal amounts in response to EGFRwtor EGFRvIII-expressing cells. By contrast, 2173-CAR–T cells exhibited these responses (to a similar extent as did cetuxCAR–T cells) only in response to EGFRvIII-expressing cells. Next, immunodeficient mice were grafted with human skin, which expresses low levels of EGFR, and injected intravenously 4 weeks later with CAR–T cells. Cetux-CAR–T cell injection led to a high degree of lymphocyte infiltration and lymphocyte-induced apop tosis in the human skin, whereas there was minimal infiltration or cell death in the human skin on animals injected with 2173-CAR–T cells. Together, these results imply that 2173-CAR–T cells do not target the immune response to EGFRwt-expressing tissues. Finally, the authors tested the 2173-CAR–T cells in an immuno deficient mouse model of intracranial EGFRvIII-expressing glioma; 7 days after tumour cell implantation, 2173-CAR–T cells, untransduced T cells or vehicle solution were injected intravenously. Three days later (11 days after tumour implantation), tumours in the 2173-CAR–T cell-treated group were ~65% smaller than in either of the two control groups. Overall, this study indicates that CAR–T cells can effectively target EGFRvIII-expressing tumours, while avoiding potential toxic effects from interactions with EGFRwt. Given the current lack of safe and effective drugs for glioblastoma, it may be more feasible to test therapies in this indication, even if they potentially carry relatively high risks of toxicity. A Phase I study of 2173-CAR–T cells in EGFRvIII-positive glioblastoma is currently enrolling. Natasha Bray A N T I C A N C E R D R U G S