Conditional internalization of PEGylated nanomedicines by PEG engagers for triple negative breast cancer therapy

Triple-negative breast cancer (TNBC) lacks effective treatment options due to the absence of traditional therapeutic targets. The epidermal growth factor receptor (EGFR) has emerged as a promising target for TNBC therapy because it is overexpressed in about 50% of TNBC patients. Here we describe a PEG engager that simultaneously binds polyethylene glycol and EGFR to deliver PEGylated nanomedicines to EGFR+ TNBC. The PEG engager displays conditional internalization by remaining on the surface of TNBC cells until contact with PEGylated nanocarriers triggers rapid engulfment of nanocargos. PEG engager enhances the anti-proliferative activity of PEG-liposomal doxorubicin to EGFR+ TNBC cells by up to 100-fold with potency dependent on EGFR expression levels. The PEG engager significantly increases retention of fluorescent PEG probes and enhances the antitumour activity of PEGylated liposomal doxorubicin in human TNBC xenografts. PEG engagers with specificity for EGFR are promising for improved treatment of EGFR+ TNBC patients.

B reast cancer is the second most common cancer in the world. Triple-negative breast cancer (TNBC), which comprises 11.2-16.3% of breast cancers, lacks expression of oestrogen receptors, progesterone receptors and human epidermal growth factor receptor 2. TNBC is typically heterogeneous, aggressive and is associated with poor prognosis with limited treatment options due to the absence of well-defined therapeutic targets 1,2 . Systemic chemotherapy has been the primary treatment option for TNBC until it was discovered that epidermal growth factor receptor (EGFR) is overexpressed in B50% of TNBC tumours 3 . EGFR-targeted agents are therefore under development for the treatment of TNBC 4 . However, EGFR-targeted tyrosine kinase inhibitors such as gefitinib 5 and erlotinib 6 display minimal therapeutic efficacy in TNBC patients.
Nanomedicines, including PEGylated liposomal doxorubicin, are currently under investigation for the treatment of TNBC [7][8][9] . Nanocarriers are attractive because they can alter the pharmacokinetic profile of drugs, reduce off-target toxicity and improve the therapeutic index 10 . Tumour accumulation of nanomedicines relies on the enhanced permeability and retention effect in which the leaky blood vasculature combined with impaired lymphatic drainage can facilitate passive accumulation of nanosized particles in tumours 11 . Lung, breast and ovarian tumours display high enhanced permeability and retention effect-mediated accumulation of nanocarriers, making nanomedicines an attractive treatment alternative for TNBC 12 .
Active targeting by functionalizing the surface of nanocarriers with ligands that bind to endocytic receptors on cancer cells can promote receptor-mediated endocytosis for increased cellular uptake of nanomedicines with concomitant improved antitumour activity [13][14][15][16] . However, many technical and regulatory hurdles must be overcome to produce new nanocarriers with reproducible and homogenous ligand densities and activities 17 . Attachment of targeting ligands can also compromise the stealth features of nanocarriers and hinder tumour uptake 10,18 .
Pre-targeting strategies can help curtail these problems by separating the production of the targeting moiety and nanocarrier as well as allowing the administration of unmodified highly stealth nanocarriers 19 . Because nanomedicines are commonly grafted with poly (ethylene glycol) (PEG) to decrease uptake and clearance by the reticuloendothelial system [20][21][22] , here we develop a general pre-targeting strategy for conditional internalization of PEGylated nanomedicines for improved treatment of EGFR þ TNBC. This is accomplished by generating bispecific PEG-binding antibodies (PEG engagers) for targeted delivery of PEGylated nanomedicines to tumours. Pre-targeting of PEG engagers induce endocytosis of PEGylated nanocarriers into EGFR þ TNBC cancers leading to enhanced antitumour efficacy of PEG-modified therapeutic agents in vitro and in vivo.

Results
Characterization of bispecific PEG engagers. A bispecific PEG engager was generated by genetically fusing a humanized anti-PEG Fab fragment with an anti-EGFR single-chain disulfidestabilized Fv fragment. The PEG engager was designed to bind to EGFR on TNBC cells but remain dormant until contact and binding to PEG-coated nanocarriers induces rapid internalization into cancer cells (Fig. 1). Briefly, a humanized anti-PEG (6.3) Fab was constructed as a single open reading frame by fusing variable light chain with constant kappa light chain (V L -C k ) and variable heavy chain with first constant heavy chain (V H -CH 1 ) domains with an internal ribosome entry site bicistronic expression linker, allowing the coordinated expression of light and heavy chains. Single chain disulfide-stabilized variable fragments (dsFv) specific for CD19 (negative control) or EGFR were linked to the C terminus of the 6.3 Fab via a GGGGS peptide linker to generate bispecific PEG engager CD19 and PEG engager EGFR (Fig. 2a). The PEG engager genes were inserted into a lentiviral expression vector and stable 293FT producer cells were generated 41 . PEG engagers purified from the culture medium display the expected molecular sizes as visualized on a reducing and a non-reducing 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 2b). The PEG engager CD19 and PEG engager EGFR have molecular weights of 78 kDa and 79 kDa, respectively, as determined by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Fig. 2c). Size-exclusion high-performance liquid chromatography analysis showed a major peak corresponding to a monomer with minimal aggregation (Supplementary Fig. 1a). The melting temperatures of PEG engager CD19 and PEG engager EGFR (75°C and 75.8°C, respectively), as determined by differential scanning calorimetry ( Supplementary Fig. 1b), were higher than the typical melting temperatures of Fab fragments (61.9-69.4°C) 23,24 , indicating that the engagers display good thermal stabilities. Analysis of PEG engager binding to PEG by microscale thermophoresis showed that PEG engager EGFR (equilibrium dissociation constant, K D ¼ 7.6±1.13 nM) and PEG engager CD19 (K D ¼ 7.6±1.05 nM) displayed similar binding affinities (Fig. 2d). The binding affinity (K D ) of the PEG engagers to their corresponding tumour antigen targets (EGFR and CD19) was 0.96 ± 0.23 nM and 3.6 ± 0.35 nM for PEG engager EGFR and PEG engager CD19 , respectively (Fig. 2e).  PEG engager CD19 PEG engager EGFR P E G e n g a g e r E G F R P E G e n g a g e r C D 1 9 E r b i t u x I g G P E G e n g a g e r E G F R P E G e n g a g e r C D 1 9 E r b i t u x I g G  We measured anti-PEG antibody concentrations in healthy human plasma samples using an anti-PEG enzyme-linked immunosorbent assay (ELISA) developed in our lab ( Supplementary Fig. 7a). The pre-existing anti-PEG IgG concentrations ranged from 0.3 to 237.5 mg ml À 1 with a mean concentration of 5.75±16.0 mg ml À 1 in 386 anti-PEG positive samples 27 . We selected a human serum sample containing a relatively high concentration of anti-PEG IgG (51.4 mg ml À 1 ). However, PEG engager EGFR plus Doxisome in 20% human serum that was positive for anti-PEG IgG antibodies or in control human serum displayed similar EC 50 values against MDA-MB-468 cells ( Supplementary Fig. 7b). These results suggest that pre-existing anti-PEG antibodies in patients do not effectively compete with PEG engager EGFR , presumably due to the high anti-PEG affinity of the PEG engager and abundant PEG chains on Doxisome.
Pharmacokinetics and tumour targeting of the PEG engager. Pre-targeting of PEG engagers to tumours may allow for subsequent accumulation and endocytosis of PEGylated nanocarriers in cancer cells. The in vivo pharmacokinetics of PEG engagers was examined to determine a reasonable time point for administration of PEGylated nanocarriers after administration of PEG engager. The half-lives of the PEG engagers were B2.1 h (PEG engager EGFR ) and 2.2 h (PEG engager CD19 ) (Fig. 6a) after intravenous administration of 150 mg of PEG engager. Almost 90% of the PEG engagers were cleared from the circulation by 5 h after injection (Fig. 6a).
To visualize whether pre-targeting can facilitate the uptake and retention of PEGylated compounds in tumours, mice bearing established high EGFR expression levels (MDA-MB-468 and A431) or low EGFR expression levels (HepG2) tumours were intravenously injected with 6 mg kg À 1 PEG engager and then subsequently intravenously injected with 4armPEG 10k -NIR-797 probe 5 h later. In vivo imaging system (IVIS) optical imaging of these mice at 24, 48 and 72 h after probe injection showed that the fluorescence signal in PEG engager EGFR targeted tumours was significantly enhanced as compared with the PEG engager CD19 control group ( Fig. 6b and Supplementary Fig. 8  engager CD19 -treated tumours, respectively (Fig. 6c). Neither PEG engager EGFR nor PEG engager CD19 enhanced the fluorescence signal in HepG2 (low EGFR expression levels) tumour-bearing mice ( Supplementary Fig. 9).
Antitumour activity of pre-targeted PEG engager. To investigate whether PEG engager EGFR can inhibit EGFR signalling, EGFR-positive A431 cells were stimulated with or without epidermal growth factor and then co-incubated with PEG engagers or control antibodies. Both Erbitux (monoclonal anti-EGFR IgG) and PEG engager EGFR inhibited the phosphorylation of EGFR and Erk as compared with negative control Herceptin (anti-HER2 IgG) and PEG engager CD19 ( Supplementary Fig. 10  The maximum-tolerated dose of doxorubicin in severe combined immunodeficiency (SCID) mice is around 2.5-3 mg kg À 1 due to defective DNA repair in these mice 28 . Increasing the dose of Doxisome to 3 mg kg À 1 did not provide better therapeutic activity because the mice experienced significant body weight loss and early deaths (Fig. 7b). Our results demonstrate that pre-targeting PEG engager EGFR to EGFR overexpressing TNBC tumours can markedly enhance the therapeutic efficacy of PEGylated liposomal doxorubicin (Doxisome) with limited side effects as shown by body weight loss analysis (Fig. 7b). We further investigated whether pre-docking of PEGylated nanoparticles with PEG engagers could enhance their therapeutic efficacy. We used a molar ratio of PEG engager and PEG-lipid on Doxisome of 1:55 ( Supplementary Fig. 11a). Mice were administrated with a mixture of PEG engager and Doxisome (pre-incubated at 4°C for 1 h) and blood samples were then periodically collected from the tail vein of the mice. The half-lives of the PEG engagers as determined by quantitative ELISA were B3.5 h (PEG engager EGFR ) and 3.8 h (PEG engager CD19 ) ( Supplementary Fig. 11b) after intravenous administration of PEG engager-docked Doxisome (containing 30 mg PEG engager). To examine the therapeutic activity of PEG engager-docked Doxisome, non-obese diabetic-severe combined immunodeficiency (NOD SCID) mice bearing human MDA-MB-468 TNBC xenografts were intravenously injected with PBS, 3 mg kg À 1 free doxorubicin, 6 mg kg À 1 PEG engager EGFR alone, PEG engager CD19 -decorated Doxisome or PEG engager EGFRdecorated Doxisome (1 mg kg À 1 of doxorubicin) on days 1, 8, 15 and 22. Doxorubicin slightly inhibited tumour growth while significantly better antitumour activity was observed in mice treated with Doxisome (1 mg kg À 1 ) or PEG engager CD19decorated Doxisome as compared to mice treated with PBS vehicle (Supplementary Fig. 11c, Po0.005). A higher dose (3 mg kg À 1 ) of Doxisome was toxic to the mice ( Fig. 7b and Supplementary Fig. 11d). PEG engager EGFR -decorated Doxisome significantly suppressed TNBC tumour growth as compared to mice treated with Doxisome alone (Supplementary Fig. 11c, Po0.05). Thus, pre-docking PEG engager EGFR on Doxisome markedly enhanced therapeutic efficacy with minimal side effects as measured by body weight loss ( Supplementary Fig. 11d).
Off-target effects of PEG engager-mediated therapy. EGFR is expressed at low levels on normal cells of epithelial, mesenchymal and neuronal origin. We hypothesized that the density of EGFRs on cells might be an important factor for conditional internalization of PEGylated nanocarriers by PEG engager EGFR (Fig. 8a). Indeed, a linear correlation was observed between the logarithm of EGFR expression levels on cancer cell lines and the logarithm of the anti-proliferative activity (EC 50  It has been reported that EGFR-targeted therapies can cause hepatotoxicity 29 . EGFR expression in normal human hepatocytes (mean fluorescence intensity ¼ 38), however, is relatively low (Fig. 8c), which suggests reduced off-target toxicity by PEG engager EGFR therapy.

Discussion
We report bispecific PEG engagers that simultaneously bind to PEG on nanomedicines and EGFR on cancer cells for selective delivery of PEGylated stealth nanocarriers to EGFR þ TNBC cells. PEG engager EGFR bound to EGFRs on TNBC cells but remained in a dormant state on the plasma membrane until contact with PEGylated nanocarriers triggered rapid endocytosis of both the engager and PEGylated nanocarrier. PEG engager EGFR markedly increased the anticancer activity of PEG-liposomal doxorubicin (Doxisome) against EGFR þ TNBC in vitro and in vivo. The PEG engager tolerated the presence of pre-existing anti-PEG antibodies and effectiveness correlated to the EGFR expression levels on cells. This simple and flexible strategy appears promising for enhanced delivery of PEGylated medicines for improved therapy of TNBC patients.
TNBC is highly aggressive and metastatic with high rates of recurrence 2 . About 50% of TNBC tumours overexpress EGFR, which is correlated with poor prognosis in TNBC patients 3 . Overexpressed EGFR can induce cell proliferation, promote angiogenesis, increase cell migration and enhance chemoresistance 30 . Therefore, EGFR has been considered as a therapeutic target for the treatment of TNBC. However, several receptor tyrosine kinase inhibitors (RTKIs) including gefitinib and erlotinib did not benefit TNBC patients 5,6 . Compared to wild-type EGFR, EGFRs with activating mutations in non-small cell lung cancers display higher binding to RTKIs, leading to 100-fold greater sensitivity to RTKIs 31 . Thus, the poor efficacy of RTKIs in TNBC may be due to the low occurrence of activating EGFR mutations in TNBC (3-11%) 32,33 as compared to non-small cell lung cancers (10-35%) 34,35 . In contrast to RTKIs that target activated intracellular EGFR kinase domains, antibodies that recognize the extracellular domain of the EGFR can target both wild-type and mutated EGFRs ( Fig. 5 and Supplementary Fig. 4). Therapeutic monoclonal antibodies that recognize extracellular domains of EGFR, such as cetuximab, have been applied for TNBC therapy. However, the response rate of cetuximab alone or combined with carboplatin for TNBC treatment was not promising 36 . These EGFR-targeted antibodies and RTKIs focus on the blockage of EGFR signalling pathways, which involves complicated signalling networks. For instance, temsirolimus can inhibit the EGFR-PI3K-AKT-mTOR pathway resulting in cancer cell death. However, MDA-MB-231 TNBC cells, which possess normal phosphatase and tensin homologue function (a tumour suppressor gene that inactivates AKT) are resistant to temsirolimus, whereas MDA-MB-468 TNBC cells (with loss of phosphatase and tensin homologue function) were 8,000-fold more sensitive to this drug 37 . Alternatively, delivery of anticancer drugs to TNBC cells by targeting EGFRs may bypass the drawbacks of EGFR signalling inhibition. Thus, PEG engager EGFR may be useful for selective delivery of PEGylated nanomedicines to EGFR overexpressing TNBC.
Active targeting of nanomedicines to tumours by functionalizing nanocarriers with targeting ligands is attractive to improve tumour targeting and intracellular delivery of nanocargos 13,14 . However, the complexity involved in production of targeted nanomedicines is time-consuming, technically challenging and expensive. The technical challenges include difficulty in controlling ligand conjugation stoichiometry, aggregation of  Figure 6 | Pharmacokinetics and imaging of PEG engagers in mice. (a) NSG mice were intravenously injected with 6 mg kg À 1 PEG engager EGFR (white circles) or PEG engager CD19 (red squares). Mean plasma concentrations of the PEG engagers were measured by sandwich ELISA (n ¼ 3 mice). (b) Five hours before intravenous administration of 4armPEG 10k -NIR-797 probe (5 mg kg À 1 ), NSG mice bearing subcutaneous MDA-MB-468 tumours were intravenously injected with 6 mg kg À 1 PEG engager EGFR or PEG engager CD19 and the whole-body imaging were sequentially imaged at 24, 48 and 72 h with an IVIS Spectrum imaging system. (c) The uptake of PEG-NIR797 in MDA-MB-468 tumours was determined by measuring fluorescence intensities (n ¼ 3). Data are shown as mean ± s.d. Significant differences in mean fluorescent intensity between PEG engager EGFR and PEG engager CD19 groups are indicated: *Pr0.01, **Pr0.001 (two-way analysis of variance). NS, not significant. ligands on nanocarrier surfaces, intercarrier variations in ligand density, loss of ligand bioactivity, deattachment of targeting ligands from nanocarriers in vivo and poor scalability or reproducibility of the conjugation process 17 . The resulting batch-to-batch variations can hinder clinical translation and commercialization of targeted nanocarriers. To help overcome these problems, Brinkmann and colleagues developed bispecific digoxigenin-binding antibodies for delivery of digoxigenmodified nanocarriers to disease sites 38,39 . Likewise, we previously described bispecific PEG-binding antibodies for tumour-specific delivery of PEGylated compounds 40,41 . However, direct attachment of targeting ligands to nanocarriers can reduce the stealth feature of the nanocarriers, resulting in accelerated clearance and reduced uptake into tumours 10 . In addition, attachment of targeting ligands can increase nanocarrier size, further impeding tumour uptake 18 .
Pre-targeting strategies may help translate targeted nanomedicines to the clinic by decreasing nanocarrier complexity, maintaining the stealth properties of nanocarriers and facilitating personalized combinations of targeting ligands and nanomedicines. For example, monovalent ligands can be targeted to high density receptors on cancer cells for subsequent internalization of nanocarriers modified with biorthogonal click-chemistry agents 42,43 . However, both the targeting ligand and the nanocarrier require chemical modification and the kinetics of the biorthogonal click-chemistry reactions may limit nanocarrier capture in vivo 44 . By contrast, PEG engagers can be used with any nanomedicine that is coated with PEG, which is commonly used to decrease nanocarrier uptake in macrophages and increase nanocarrier circulation time. Thus, even 'off-the-shelf' nanomedicines, such as Doxil, can be targeted to EGFR þ TNBC tumours by this approach. Antibody-antigen interactions also rapidly occur in vivo, promoting effective capture and internalization of stealth nanocarriers.
Conditional internalization of nanocarriers into target cells is desirable for successful pre-targeted delivery 43,45 . Receptor dimerization stimulates EGFR endocytosis for turnover and downregulation of EGFRs 46,47 . Anti-EGFR monoclonal antibodies, such as cetuximab and necitumumab, recognize the domain III ligand-binding region of EGFR and rapidly trigger crosslinking of EGFRs due to their bivalent format, thereby stimulating internalization and degradation of the receptor 48 . Since bivalent antibodies are typically required to crosslink receptors and trigger endocytosis, we designed a monovalent PEG engager EGFR to accumulate at EGFRs without promoting receptor internalization. PEG engagers were retained on target cancer cells until contact with PEGylated nanocarriers, which crosslinked EGFRs and induced rapid endocytosis of the nanomedicines.
Doxil, PEGylated liposomal doxorubicin, is approved by FDA for treatment of ovarian and breast cancer patients 9,49 . Although the standard dose of Doxil recommended by the FDA (50 mg m À 2 ) significantly benefits cancer patients, a high incidence of hand-foot syndrome is also observed 9,50 . Doxil treatment is often delayed for patients with handfoot syndrome, limiting its effectiveness. PEG engager EGFR enhanced the antitumour activity of low-dose Doxisome (1 mg kg À 1 ¼ 3 mg m À 2 ), to TNBC tumours 51 . This significant tumour suppression may be due to the specific targeting and rapid endocytosis of Doxisome via pre-targeted PEG engagers. Therefore, the PEG engager-mediated liposomal doxorubicin therapy may allow effective therapy at lower doses, thus reducing side effects such as hand-foot syndrome. We also demonstrated (a) Groups of eight SCID mice bearing MDA-MB-468 tumours were intravenously injected with 6 mg kg À 1 PEG engager CD19 (red squares), 6 mg kg À 1 PEG engager EGFR (white circles) or 18 mg kg À 1 PEG engager EGFR (white squares) 5 h before intravenous injection of 1 mg kg À 1 Doxisome. Groups of eight mice were also intravenously injected with 6 mg kg À 1 free doxorubicin (white triangles), 3  that pre-decoration of Doxisome with PEG engager EGFR could enhance the half-life of the engagers and improve the therapeutic index of Doxisome treatment against EGFR þ human tumour xenografts, indicating that both pre-mixing and pre-targeting of PEG engagers are promising therapeutic approaches.
Stepwise accumulation of mutations can lead to distinct cancer cell subpopulations within the same tumour 52,53 . For example, discordant HER2 expression has been observed in the same biopsy from cancer patients 54 . This could be a major obstacle for conventional single-ligand targeted cancer therapy since subpopulations lacking expression of target receptors may repopulate the tumour after treatment. To overcome this problem, multiligand nanocarriers can be employed to improve drug delivery to heterogeneous tumours 55,56 . A cocktail of PEG engagers with specificity to different tumour-associated antigens (for example, insulin-like growth factor 1 receptor, HER2, HER3, HER4 and c-Met) might offer a flexible yet effective treatment option for heterogeneous tumours.
Antibody-drug conjugates (ADC) are one of the most promising targeted drug delivery systems. ADCs are composed of monoclonal antibodies conjugated with cytotoxic drug molecules via a biodegradable linker to allow drug release after ADC internalization into cancer cells 57 . Low drug-to-antibody ratios of two-four are required to prevent alteration of antibody properties and maintain efficient cancer therapy, thereby requiring attachment of highly cytotoxic drugs such as monomethyl auristatin E and maytansine [58][59][60] . EGFR is also a potential target for developing ADCs as two EGFR-targeting ADCs are in clinical trials (ABT-414/Abbvie 61 and IMGN289/ ImmunoGen). However, off-target uptake of ADCs in normal cells which express low levels of the target receptors can cause severe side effects 62 . We determined that EGFR expression density is critical for efficient PEG engager EGFR -mediated endocytosis of PEGylated liposomal doxorubicin, suggesting that PEG engagers may possess lower off-target toxicity.
This may expand the range of targetable tumour antigens as well as allow delivery of a broader range of therapeutic cancer drugs since each nanocarrier can encapsulate up to 10,000 drug molecules 63 . On the other hand, PEG engagers might be less effective for treatment of cancers that express low levels of target antigens.
Covalent attachment of PEG to peptides, proteins, nucleic acids and nanoparticles can improve their pharmacokinetic properties and stability 64 . Although PEG is thought to be non-immunogenic, several studies have detected pre-existing anti-PEG antibodies in normal donors 26,65,66 . Pre-existing anti-PEG antibodies could theoretically compromise the effectiveness of PEG engagers by blocking engager binding to PEGylated medicines. However, we found that human serum that contained high levels of anti-PEG antibodies did not interfere with PEG engager-mediated delivery of Doxisome into TNBC cells. This may be because PEG engagers display high affinity to PEG (K D ¼ 7.55 nM) and are not easily competed by naturally occurring anti-PEG antibodies.
In conclusion, the humanized PEG engager generated in this study is anticipated to help overcome major bottlenecks in targeted nanomedicines by reducing manufacturing complexity and providing flexibility in choosing the appropriate disease targets by simply changing the targeting portion of these molecules. Flexible targeting coupled with the capability to directly deliver nanocarriers into target cells may expand the range of therapeutic agents available for therapy.

Methods
Cell lines and animals. Human 293FT cells (Thermo Fisher Scientific, San Jose, CA) were cultured in Dulbecco's modified Eagle's medium (Sigma-Aldrich, St Louis, MO) supplemented with 10% heat-inactivated foetal calf serum (HyClone, Logan, Utah), 100 U ml À 1 penicillin and 100 mg ml À 1 streptomycin at 37°C in an atmosphere of 5% CO 2 in air. Human hepatocytes were purchased from BD Biosciences (San Jose, CA) EGFR-mutated PC9 (EGFR exon19del E746-A750) cell line was kindly provided by Dr Pan-Chyr Yang (President of the National Taiwan Collection (Manassas, VA) and were maintained in RPMI-1640 containing the same supplements. Knockdown of EGFR in BT20 cells is described in Supplementary Methods. Except for BT-20 cells, the cell lines used in this study are not listed in the database of commonly misidentified cell lines maintained by the International Cell Line Authentication Committee. The cell lines were not authenticated by our laboratory. All cell lines were tested for mycoplasma and propagated for less than 6 months after resuscitation. Female NSG mice (non-obese diabetic. Cg-Prkdc scid Il2rg tm1Wjl /SzJ, 6-8 week old), NOD SCID mice (NOD.CB17-Prkdc scid /NcrCrl, 6-8 week old) and BALB/c nude mice (BALB/cAnN.Cg-Foxn1 nu /CrlNarl) were obtained from the National Laboratory Animal Center, Taipei, Taiwan and were maintained under specific pathogen-free conditions. All animal experiments were performed in accordance with institutional guidelines and ethically approved by the Laboratory Animal Facility and Pathology Core Committee of IBMS, Academia Sinica. Mice were randomly assigned to treatment groups but the investigators were not blinded to the treatments.
DNA plasmid construction. pAS3w.Ppuro, pMD.G (VSV-G envelope plasmid) and pCMVDR8.91 (packaging plasmid) vectors were obtained from the National RNAi Core Facility (Institute of Molecular Biology/Genomic Research Center, Academia Sinica, Taiwan) 67 . To generate the anti-PEG Fab-based bispecific PEG engager antibodies, the mouse V L and V H domains of the 6.3 antibody were cloned from cDNA prepared from the 6.3 hybridoma 40 . The anti-PEG antibody was humanized by first aligning the V H and V L sequences of the mouse 6.3 antibody to human immunoglobulin germline sequences using the IgBLAST program (http://www.ncbi.nlm.nih.gov/igblast/). The human germline V H IGHV7-4-1*02 and V L IGKV4-1*01 exons were selected based on the framework homology. The complementarity-determining regions of mouse 6.3 V H and V L domains were then grafted onto human V H IGHV7-4-1*02 and V L IGKV4-1*01 genes using assembly PCR. Human immunoglobulin G 1 (IgG 1 ) C k and CH 1 constant domains were cloned from extracted human peripheral blood mononuclear cells cDNA. Humanized 6.3 V L -C k and 6.3 V H -CH 1 domains were assembled by overlap polymerase chain reaction from humanized 6.3 V L and 6.3 V H and human C k and CH 1 fragments 41 . The humanized 6.3 V L -C k and 6.3 V H -CH 1 were joined by a composite internal ribosome entry site bicistronic expression peptide linker and inserted into the pAS3w.Ppuro plasmid. In vivo pharmacokinetics. NSG mice were intravenously injected with 150 mg PEG engager CD19 or PEG engager EGFR and blood samples were periodically collected from the tail vein of the mice. Plasma was prepared by centrifugation (5 min, 12,000g). The PEG engager levels in plasma were determined by quantitative sandwich ELISA. Maxisorp 96-well microplates were coated with 50 ml per well of anti-6 Â His tag antibody (GeneTex) (2 mg ml À 1 ) in bicarbonate buffer, pH 8.0 for 4 h at 37°C and then at 4°C overnight. The plates were blocked with 200 ml per well 5% skim milk in PBS for 2 h at room temperature and then washed with PBS three times. Graded concentrations of PEG engager CD19 , PEG engager EGFR or plasma samples in dilution buffer (2% skim milk in PBS) were added to the wells for 2 h at room temperature. After washing with PBS four times, the plates were stained with 50 ml per well horseradish peroxidase-conjugated anti-human IgG Fab antibody (Jackson ImmunoResearch Laboratories) (5 mg ml À 1 ). The plates were washed with PBS six times and 100 ml per well ABTS solution (0.4 mg ml À 1 2,2 0 -azino-di(3-ethylbenzthiazoline-6-sulfonic acid), 0.003% H 2 O 2 , 100 mM phosphate citrate, pH 4.0) was added for 30 min at room temperature. The absorbance of the wells at 405 nm was measured on a microplate reader. The initial and terminal half-lives of the PEG engagers were estimated by fitting the data to a two-phase exponential decay model with Prism 5 software (Graphpad Software).
Synthesis of PEGylated near-infrared probes. 4arm-PEG 10K -NH 2 (Laysan Bio) dissolved in dimethyl sulfoxide at 2 mg ml À 1 was mixed with a sixfold molar excess of NIR-797 isothiocyanate (Santa Cruz Biotechnology) (in dimethyl sulfoxide) for 2 h at room temperature to produce 4arm-PEG 10K -NIR-797 probes. These compounds were diluted in a fivefold volume of ddH 2 O and dialysed (molecular weight cutoff B12,000-14,000 daltons) against ddH 2 O to remove free NIR-797 isothiocyanate. The probes were sterile filtered and stored at À 80°C.
Pre-docked PEG engagers. Characterization, in vivo blood half-life and in vivo therapy of pre-docked PEG engagers is described in Supplementary Methods.
In vivo antitumour therapy. Groups of NSG or NOD SCID mice bearing 44.7 ± 10.7 mm 3 MDA-MB-231 (n ¼ 6) or 84.3 ± 4.3 mm 3 subcutaneous MDA-MB-468 (n ¼ 8) tumours on their right flank were intravenously injected with PBS or 6 or 18 mg kg À 1 PEG engagers. After 5 h, the mice were intravenously administrated with free doxorubicin (3 mg kg À 1 ) or Doxisome (1 or 3 mg kg À 1 ). Treatment was repeated once a week for a total of 4 weeks. Tumour sizes were measured every 7 days. Tumour volumes were calculated according the formula: length Â width Â height Â 0.5.
Statistical analysis. Results are presented as the mean±s.d. All experiments were repeated at least two times with representative data shown. Animal sample size was chosen based on similar well-characterized literature. Statistical analyses were examined using the two-way analysis of variance. Differences in tumour volumes between groups were examined for statistical significance using one-way analysis of variance followed by Dunnett's multiple comparisons; a probability value o 0.05 was considered statistically significant. No statistical method was used to predetermine sample sizes.
Data availability. Data supporting the findings of this study are available in the article and its Supplementary Information files, or from the corresponding author on reasonable request.