Thank you for visiting Nature.com. You are using a browser version with limited CSS support. For the best experience, we recommend that you use an updated browser (or disable Compatibility Mode in Internet Explorer). In addition, to ensure ongoing support, we show the site without styles and JavaScript.
Displays a carousel of three slides at once. Use the Previous and Next buttons to move through three slides at a time, or use the slider buttons at the end to move through three slides at a time.
The blood-brain barrier and the blood-brain barrier prevent biotherapeutic agents from reaching their targets in the central nervous system, thereby hindering effective treatment of neurological diseases. To discover novel brain transporters in vivo, we introduced a T7 phage peptide library and serially collected blood and cerebrospinal fluid (CSF) using a cannulated conscious large pool model of rats. Specific phage clones were highly enriched in CSF after four rounds of selection. Testing of individual candidate peptides revealed more than 1000-fold enrichment in CSF. The bioactivity of the peptide-mediated delivery to the brain was confirmed by a 40% reduction in the level of amyloid-β in the cerebrospinal fluid using a BACE1 peptide inhibitor linked to the identified novel transit peptide. These results suggest that the peptides identified by in vivo phage selection methods may be useful vehicles for systemic delivery of macromolecules to the brain with a therapeutic effect.
Central nervous system (CNS) targeted therapy research has largely focused on identifying optimized drugs and agents that exhibit CNS-targeting properties, with less effort on discovering the mechanisms that drive active drug delivery to the brain. This is starting to change now as drug delivery, especially large molecules, is an integral part of modern neuroscience drug development. The environment of the central nervous system is well protected by the cerebrovascular barrier system, consisting of the blood-brain barrier (BBB) and the blood-brain barrier (BCBB)1, making it challenging to deliver drugs to the brain1,2. It is estimated that nearly all large molecule drugs and more than 98% of small molecule drugs are eliminated from the brain3. That is why it is very important to identify new brain transport systems that provide efficient and specific delivery of therapeutic drugs to the CNS 4,5. However, the BBB and BCSFB also present an excellent opportunity for drug delivery as they penetrate and enter all structures of the brain through its extensive vasculature. Thus, the current efforts to use non-invasive methods of delivery to the brain are largely based on the mechanism of receptor-mediated transport (PMT) using the endogenous BBB6 receptor. Despite recent key advances using the transferrin receptor pathway7,8, further development of new delivery systems with improved properties is required. To this end, our goal was to identify peptides capable of mediating CSF transport, as they could in principle be used to deliver macromolecules to the CNS or to open new receptor pathways. In particular, specific receptors and transporters of the cerebrovascular system (BBB and BSCFB) can serve as potential targets for the active and specific delivery of biotherapeutic drugs. Cerebrospinal fluid (CSF) is a secretory product of the choroid plexus (CS) and is in direct contact with the interstitial fluid of the brain through the subarachnoid space and ventricular space4. Recently it has been shown that subarachnoid cerebrospinal fluid diffuses excessively into the interstitium of the brain9. We hope to access the parenchymal space using this subarachnoid inflow tract or directly through the BBB. To achieve this, we implemented a robust in vivo phage selection strategy that ideally identifies peptides transported by either of these two distinct pathways.
We now describe a sequential in vivo phage display screening method with CSF sampling coupled with high throughput sequencing (HTS) to monitor initial selection rounds with the highest library diversity. Screening was performed on conscious rats with a permanently implanted large cisterna (CM) cannula to avoid blood contamination. Importantly, this approach selects both brain-targeting and peptides with transport activity across the cerebrovascular barrier. We used T7 phages due to their small size (~60 nm)10 and suggested that they are suitable for the transport of vesicles that allow transcellular crossing of the endothelial and/or epithelial-medulla barrier. After four rounds of panning, phage populations were isolated showing strong in vivo CSF enrichment and cerebral microvessel association. Importantly, we were able to confirm our findings by demonstrating that the preferred and chemically synthesized best candidate peptides are able to transport protein cargo into the cerebrospinal fluid. First, the pharmacodynamic effects of the CNS were established by combining a leading transit peptide with an inhibitor of the BACE1 peptide. In addition to demonstrating that in vivo functional screening strategies can identify novel brain transport peptides as effective protein cargo carriers, we expect similar functional selection approaches to also become important in identifying novel brain transport pathways.
Based on plaque-forming units (PFU), after the phage packaging step, a library of random 12-mer linear T7 phage peptides with a diversity of approximately 109 was designed and created (see Materials and Methods). It is important to note that we carefully analyzed this library before in vivo panning. PCR amplification of phage library samples using modified primers generated amplicons that were directly applicable to HTS (Supplementary Fig. 1a). Due to a) HTS11 sequencing errors, b) impact on the quality of primers (NNK)1-12, and c) the presence of wild-type (wt) phage (skeleton inserts) in the standby library, a sequence filtering procedure was implemented to extract only verified sequence information (Supplementary Fig. 1b). These filter steps apply to all HTS sequencing libraries. For the standard library, a total of 233,868 reads were obtained, of which 39% passed the filter criteria and were used for library analysis and selection for subsequent rounds (Supplementary Figure 1c–e). The reads were predominantly multiples of 3 base pairs in length with a peak at 36 nucleotides (Supplementary Fig. 1c), confirming the library design (NNK) 1-12. Notably, approximately 11% of the library members contained a 12-dimensional wild-type (wt) backbone PAGISRELVDKL insert, and almost half of the sequences (49%) contained insertions or deletions. The HTS of the library library confirmed the high diversity of peptides in the library: more than 81% of peptide sequences were found only once and only 1.5% occurred in ≥4 copies (Supplementary Fig. 2a). The frequencies of amino acids (aa) at all 12 positions in the repertoire correlated well with the frequencies expected for the number of codons generated by the degenerate NKK repertoire (Supplementary Fig. 2b). The observed frequency of aa residues encoded by these inserts correlated well with the calculated frequency (r = 0.893) (Supplementary Fig. 2c). The preparation of phage libraries for injection includes the steps of amplification and removal of endotoxin. This has previously been shown to potentially reduce the diversity of phage libraries12,13. Therefore, we sequenced a plate-amplified phage library that had undergone endotoxin removal and compared it with the original library to estimate the frequency of AA. A strong correlation (r = 0.995) was observed between the original pool and the amplified and purified pool (Supplementary Fig. 2d), indicating that competition between clones amplified on plates using T7 phage did not cause major bias. This comparison is based on the frequency of tripeptide motifs in each library, since the diversity of libraries (~109) cannot be fully captured even with HTS. Frequency analysis of aa at each position revealed a small position-dependent bias in the last three positions of the entered repertoire (Supplementary Fig. 2e). In conclusion, we concluded that the quality and diversity of the library were acceptable and only minor changes in diversity were observed due to amplification and preparation of phage libraries between several rounds of selection.
Serial cerebrospinal fluid sampling can be performed by surgically implanting a cannula into the CM of conscious rats to facilitate identification of T7 phage injected intravenously (iv) via the BBB and/or BCSFB (Fig. 1a-b). We used two independent selection arms (arms A and B) in the first three rounds of in vivo selection (Fig. 1c). We gradually increased the stringency of the selection by decreasing the total amount of phage introduced in the first three rounds of selection. For the fourth round of panning, we combined samples from branches A and B and performed three additional independent selections. To study the in vivo properties of T7 phage particles in this model, wild-type phage (PAGISRELVDKL master insert) was injected into rats via the tail vein. Recovery of phages from cerebrospinal fluid and blood at different time points showed that relatively small T7 icosahedral phages had a rapid initial clearance phase from the blood compartment (Supplementary Fig. 3). Based on the titers administered and the blood volume of the rats, we calculated that only approximately 1% wt. phage from the administered dose was detected in the blood 10 minutes after intravenous injection. After this initial rapid decline, a slower primary clearance was measured with a half-life of 27.7 minutes. Importantly, only very few phages were retrieved from the CSF compartment, indicating a low background for wild-type phage migration into the CSF compartment (Supplementary Fig. 3). On average, only about 1 x 10-3% titers of T7 phage in the blood and 4 x 10-8% of initially infused phages were detected in the cerebrospinal fluid over the entire sampling period (0-250 min). Notably, the half-life (25.7 min) of wild-type phage in cerebrospinal fluid was similar to that observed in blood. These data demonstrate that the barrier separating the CSF compartment from the blood is intact in CM-cannulated rats, allowing in vivo selection of phage libraries to identify clones that are readily transported from the blood into the CSF compartment.
(a) Setting up a method for re-sampling cerebrospinal fluid (CSF) from a large pool. (b) Diagram showing the cellular location of the central nervous system (CNS) barrier and the selection strategy used to identify peptides that cross the blood-brain barrier (BBB) and the blood-brain barrier. (c) In vivo phage display screening flowchart. In each round of selection, phages (animal identifiers inside the arrows) were injected intravenously. Two independent alternative branches (A, B) are kept separately until the 4th round of selection. For selection rounds 3 and 4, each phage clone extracted from CSF was manually sequenced. (d) Kinetics of phage isolated from blood (red circles) and cerebrospinal fluid (green triangles) during the first round of selection in two cannulated rats after intravenous injection of the T7 peptide library (2 x 1012 phages/animal). Blue squares indicate the average initial concentration of phage in the blood, calculated from the amount of injected phage, taking into account the total blood volume. The black squares indicate the point of intersection of the y line extrapolated from blood phage concentrations. (e,f) Present the relative frequency and distribution of all possible overlapping tripeptide motifs found in the peptide. The number of motifs found in 1000 readings is shown. Significantly (p < 0.001) enriched motifs are marked with red dots. (e) Correlation scatterplot comparing the relative frequency of the tripeptide motif of the injected library with blood-derived phage from animals #1.1 and #1.2. (f) Correlation scatterplot comparing the relative frequencies of animal phage tripeptide motifs #1.1 and #1.2 isolated in blood and cerebrospinal fluid. (g, h) Sequence ID representation of phage enriched in blood (g) versus injected libraries and phage enriched in CSF (h) versus blood after a round of in vivo selection in both animals. The size of the one-letter code indicates how often that amino acid occurs at that position. Green = polar, purple = neutral, blue = basic, red = acidic and black = hydrophobic amino acids. Figure 1a, b was designed and produced by Eduard Urich.
We injected a phage peptide library into two CM instrument rats (clades A and B) and isolated phage from cerebrospinal fluid and blood (Figure 1d). The initial rapid clearance of the library was less pronounced compared to the wild-type phage. The mean half-life of the injected library in both animals was 24.8 minutes in blood, similar to wild-type phage, and 38.5 minutes in CSF. Blood and cerebrospinal fluid phage samples from each animal were subjected to HTS and all identified peptides were analyzed for the presence of a short tripeptide motif. Tripeptide motifs were chosen because they provide a minimal basis for structure formation and peptide-protein interactions14,15. We found a good correlation in the distribution of motifs between the injected phage library and clones extracted from the blood of both animals (Fig. 1e). The data indicate that the composition of the library is only marginally enriched in the blood compartment. Amino acid frequencies and consensus sequences were further analyzed at each position using an adaptation of the Weblogo16 software. Interestingly, we found a strong enrichment in blood glycine residues (Fig. 1g). When blood was compared with clones selected from CSF, strong selection and some deselection of motifs were observed (Fig. 1f), and certain amino acids were preferentially present at predetermined positions in the 12-member (Fig. 1h). Notably, individual animals differed significantly in cerebrospinal fluid, whereas blood glycine enrichment was observed in both animals (Supplementary Fig. 4a–j). After stringent filtering of sequence data in the cerebrospinal fluid of animals #1.1 and #1.2, a total of 964 and 420 unique 12-mer peptides were obtained (Supplementary Fig. 1d–e). The isolated phage clones were amplified and subjected to a second round of in vivo selection. Phage extracted from the second round of selection were subjected to HTS in each animal and all identified peptides were used as input to a motif recognition program to analyze the occurrence of tripeptide motifs (Fig. 2a, b, ef). Compared to the first cycle of the phage recovered from CSF, we observed further selection and deselection of many motifs in CSF in branches A and B (Fig. 2). A network identification algorithm was applied to determine if they represented different patterns of consistent sequence. A clear similarity was observed between the 12-dimensional sequences recovered by CSF in alternative clade A (Fig. 2c, d) and clade B (Fig. 2g, h). The pooled analysis in each branch revealed different selection profiles for 12-mer peptides (Supplementary Fig. 5c,d) and an increase in the CSF/blood titer ratio over time for pooled clones after the second round of selection compared to the first round of selection (Supplementary Fig. 5e). ).
Enrichment of motifs and peptides in cerebrospinal fluid by two successive rounds of in vivo functional phage display selection.
All cerebrospinal fluid phages recovered from the first round of each animal (animals #1.1 and #1.2) were pooled, amplified, HT-sequenced and reinjected together (2 x 1010 phages/animal) 2 SM cannulated rats (#1.1 → #). 2.1 and 2.2, 1.2 → 2.3 and 2.4). (a,b,e,f) Correlation scatterplots comparing the relative frequency of tripeptide motifs of all CSF-derived phages in the first and second selection rounds. Relative frequency and distribution of motifs representing all possible overlapping tripeptides found in peptides in both orientations. The number of motifs found in 1000 readings is shown. Motifs that were significantly (p < 0.001) selected or excluded in one of the compared libraries are highlighted with red dots. (c, d, g, h) Sequence logo representation of all CSF-rich 12 amino acid long sequences based on rounds 2 and 1 of in vivo selection. The size of the one-letter code indicates how often that amino acid occurs at that position. To represent the logo, the frequency of CSF sequences extracted from individual animals between two selection rounds is compared and the enriched sequences in the second round are shown: (c) #1.1–#2.1 (d) #1.1–#2.2 (g) #1.2–#2.3 and (h) #1.2–#2.4. The most enriched amino acids at a given position in (c, d) animals no. 2.1 and no. 2.2 or (g, h) in animals no. 2.3 and no. 2.4 are shown in color. Green = polar, purple = neutral, blue = basic, red = acidic and black = hydrophobic amino acids.
After the third round of selection, we identified 124 unique peptide sequences (#3.1 and #3.2) from 332 CSF-reconstituted phage clones isolated from two animals (Supplementary Fig. 6a). The sequence LGSVS (18.7%) had the highest relative proportion, followed by the wild-type inserts PAGISRELVDKL (8.2%), MRWFFSHASQGR (3%), DVAKVS (3%), TWLFSLG (2.2%), and SARGSWREIVSLS ( 2.2%). In the final fourth round, we pooled two independently selected branches from three separate animals (Fig. 1c). Of the 925 sequenced phage clones recovered from CSF, in the fourth round we found 64 unique peptide sequences (Supplementary Fig. 6b), among which the relative proportion of wild-type phage dropped to 0.8%. The most common CSF clones in the fourth round were LYVLHSRGLWGFKLAAALE (18%), LGSVS (17%), GFVRFRLSNTR (14%), KVAWRVFSLFWK (7%), SVHGV (5%), GRPQKINGARVC (3.6%) and RLSSVDSDLSGC (3, 2%). %)). The length range of the selected peptides is due to nucleotide insertions/deletions or premature stop codons in the library primers when using degenerate codons for NNK library design. Premature stop codons generate shorter peptides and are selected because they contain the favorable aa motif. Longer peptides may result from insertions/deletions in the primers of the synthetic libraries. This positions the designed stop codon outside the frame and reads it until a new stop codon appears downstream. In general, we calculated enrichment factors for all four selection rounds by comparing the input data with the sample output data. For the first round of screening, we used wild-type phage titers as a non-specific background reference. Interestingly, negative phage selection was very strong in the first CSF cycle, but not in blood (Fig. 3a), which may be due to the low probability of passive diffusion of most members of the peptide library into the CSF compartment or relative phages tend to be more efficiently retained or removed from the bloodstream than bacteriophages. However, in the second round of panning, strong selection of phages in CSF was observed in both clades, suggesting that the previous round was enriched in phages displaying peptides that promote CSF uptake (Fig. 3a). Again, without significant blood enrichment. Also in the third and fourth rounds, the phage clones were significantly enriched in CSF. Comparing the relative frequency of each unique peptide sequence between the last two rounds of selection, we found that the sequences were even more enriched in the fourth round of selection (Fig. 3b). A total of 931 tripeptide motifs were extracted from all 64 unique peptide sequences using both peptide orientations. The most enriched motifs in the fourth round were more closely examined for their enrichment profiles across all rounds compared to the injected library (cut-off: 10% enrichment) (Supplementary Fig. 6c). General patterns of selection showed that most of the studied motives were enriched in all previous rounds of both selection branches. However, some motifs (eg SGL, VSG, LGS GSV) were predominantly from alternative clade A, while others (eg FGW, RTN, WGF, NTR) were enriched in alternative clade B.
Validation of CSF transport of CSF-enriched phage-displayed peptides and biotinylated leader peptides conjugated to streptavidin payloads.
(a) Enrichment ratios calculated in all four rounds (R1-R4) based on injected (input = I) phage (PFU) titers and determined CSF phage titers (output = O). Enrichment factors for the last three rounds (R2-R4) were calculated by comparison with the previous round and the first round (R1) with weight data. Open bars are cerebrospinal fluid, shaded bars are plasma. (***p<0.001, based on Student’s t-test). (b) List of the most abundant phage peptides, ranked according to their relative proportion to all phages collected in CSF after round 4 of selection. The six most common phage clones are highlighted in color, numbered and their enrichment factors between rounds 3 and 4 of selection (insets). (c,d) The six most enriched phage clones, empty phage and parental phage peptide libraries from round 4 were analyzed individually in a CSF sampling model. CSF and blood samples were collected at the indicated time points. (c) Equal amounts of 6 candidate phage clones (2 x 1010 phages/animals), empty phages (#1779) (2 x 1010 phages/animals) and stock phage peptide libraries (2 x 1012 phages/animals) Inject at least 3 CM is administered to the cannulated animal separately via the tail vein. The CSF pharmacokinetics of each injected phage clone and phage peptide library over time is shown. (d) shows the average CSF/blood ratio for all recovered phages/mL over the sampling time. (e) Four synthetic leader peptides and one scrambled control were linked with biotin to streptavidin through their N-terminus (tetramer display) followed by injection (tail vein iv, 10 mg streptavidin/kg). At least three intubated rats (N = 3). ). CSF samples were collected at the indicated time points and streptavidin concentrations were measured by CSF anti-streptavidin ELISA (nd = not detected). (*p<0.05, **p<0.01, ***p<0.001, based on ANOVA test). (f) Comparison of the amino acid sequence of the most enriched phage peptide clone #2002 (purple) with other selected phage peptide clones from the 4th round of selection. Identical and similar amino acid fragments are color-coded.
Of all enriched phages in the fourth round (Fig. 3b), six candidate clones were selected for further individual analysis in the CSF sampling model. Equal amounts of six candidate phage, empty phage (no insert) and prophage peptide libraries were injected into three cannulated CM animals, and pharmacokinetics were determined in CSF (Fig. 3c) and blood (Supplementary Fig. 7) assays. All phage clones tested targeted the CSF compartment at a level 10-1000 times higher than that of the empty control phage (#1779). For example, clones #2020 and #2077 had about 1000 times higher CSF titers than control phage. The pharmacokinetic profile of each selected peptide is different, but all of them have a high CSF homing ability. We observed a constant decrease over time for clones #1903 and #2011, while for clones #2077, #2002 and #2009 an increase during the first 10 minutes might indicate active transport but needs to be verified. Clones #2020, #2002, and #2077 stabilized at high levels, while the CSF concentration of clone #2009 slowly decreased after the initial increase. We then compared the relative frequency of each CSF candidate with its blood concentration (Fig. 3d). The correlation of the mean titer of each CSF candidate with its blood titer at all sampling times showed that three of the six candidates were significantly enriched in blood CSF. Interestingly, clone #2077 showed higher blood stability (Supplementary Figure 7). To confirm that the peptides themselves are capable of actively transporting cargo other than phage particles into the CSF compartment, we synthesized four leader peptides derivatized with biotin at the N-terminus where the peptides attach to the phage particle. Biotinylated peptides (nos. 2002, 2009, 2020 and 2077) were conjugated with streptavidin (SA) to obtain multimeric forms somewhat mimicking phage geometry. This format also allowed us to measure SA exposure in blood and cerebrospinal fluid as cargo-transporting protein peptides. Importantly, phage data could often be reproduced when synthetic peptides were administered in this SA-conjugated format (Fig. 3e). The scrambled peptides had less initial exposure and faster CSF clearance with undetectable levels within 48 hours. To gain insight into the delivery pathways of these peptide phage clones into the CSF space, we analyzed the localization of individual phage peptide hits using immunohistochemistry (IHC) to directly detect phage particles 1 hour after intravenous injection in vivo. Notably, clones #2002, #2077, and #2009 could be detected by strong staining in brain capillaries, while control phage (#1779) and clone #2020 were not detected (Supplementary Figure 8). This suggests that these peptides contribute to the effect on the brain precisely by crossing the BBB. Further detailed analysis is required to test this hypothesis, as the BSCFB route may also be involved. When comparing the amino acid sequence of the most enriched clone (#2002) with other selected peptides, it was noted that some of them have similar amino acid extensions, which may indicate a similar transport mechanism (Fig. 3f).
Due to its unique plasma profile and significant increase in CSF over time, phage display clone #2077 was further explored over a longer 48-hour period and was able to reproduce the rapid increase in CSF observed in association with sustained SA levels (Fig. 4a). Regarding other identified phage clones, #2077 stained strongly for brain capillaries and showed significant colocalization with capillary marker lectin when viewed at higher resolution and possibly some staining in the parenchymal space (Figure 4b). To investigate whether peptide-mediated pharmacological effects could be obtained in the CNS, we performed an experiment in which biotinylated versions of i) the #2077 transit peptide and ii) the BACE1 inhibitor peptide were mixed with SA at two different ratios. For one combination we used only the BACE1 peptide inhibitor and for the other we used a 1:3 ratio of BACE1 peptide inhibitor to #2077 peptide. Both samples were administered intravenously and blood and cerebrospinal fluid levels of beta-amyloid peptide 40 (Abeta40) were measured over time. Abeta40 was measured in CSF as it reflects BACE1 inhibition in the brain parenchyma. As expected, both complexes significantly reduced blood levels of Abeta40 (Fig. 4c, d). However, only samples containing a mixture of peptide no. 2077 and an inhibitor of the BACE1 peptide conjugated to SA caused a significant decrease in Abeta40 in the cerebrospinal fluid (Fig. 4c). The data show that peptide no. 2077 is able to transport the 60 kDa SA protein into the CNS and also induces pharmacological effects with SA-conjugated inhibitors of the BACE1 peptide.
(a) Clonal injection (2 × 10 phages/animal) of T7 phage showing long-term pharmacokinetic profiles of CSF peptide #2077 (RLSSVDSDLSGC) and uninjected control phage (#1779) in at least three CM-intubated rats. (b) Confocal microscopic image of representative cortical microvessels in phage-injected rats (2 × 10 10 phages/animal) showing counterstaining of peptide #2077 and vessels (lectin). These phage clones were administered to 3 rats and allowed to circulate for 1 hour before perfusion. Brains were sectioned and stained with polyclonal FITC-labeled antibodies against the T7 phage capsid. Ten minutes prior to perfusion and subsequent fixation, DyLight594-labeled lectin was administered intravenously. Fluorescent images showing lectin staining (red) of the luminal side of microvessels and phages (green) in the lumen of capillaries and perivascular brain tissue. The scale bar corresponds to 10 µm. (c, d) Biotinylated BACE1 inhibitory peptide alone or in combination with biotinylated transit peptide #2077 was coupled to streptavidin followed by intravenous injection of at least three cannulated CM rats (10 mg streptavidin/kg). BACE1 peptide inhibitor-mediated reduction in Aβ40 was measured by Aβ1-40 ELISA in blood (red) and cerebrospinal fluid (orange) at the indicated time points. For better clarity, a dotted line is drawn on the graph at a scale of 100%. (c) Percentage reduction in Aβ40 in blood (red triangles) and cerebrospinal fluid (orange triangles) in rats treated with streptavidin conjugated to transit peptide #2077 and BACE1 inhibitory peptide in a 3:1 ratio. (d) Percentage reduction in blood Aβ40 (red circles) and cerebrospinal fluid (orange circles) of rats treated with streptavidin coupled to a BACE1 inhibitory peptide only. The Aβ concentration in the control was 420 pg/ml (standard deviation = 101 pg/ml).
Phage display has been successfully applied in several areas of biomedical research17. This method has been used for in vivo vascular diversity studies18,19 as well as studies targeting cerebral vessels20,21,22,23,24,25,26. In this study, we extended the application of this selection method not only to the direct identification of peptides targeting cerebral vessels, but also to the discovery of candidates with active transport properties to cross the blood-brain barrier. We now describe the development of an in vivo selection procedure in CM intubated rats and demonstrate its potential to identify peptides with CSF homing properties. Using the T7 phage displaying a library of 12-mer random peptides, we were able to demonstrate that the T7 phage is small enough (approximately 60 nm in diameter)10 to be adapted to the blood-brain barrier, thereby directly crossing the blood-brain barrier or choroid plexus. We observed that CSF harvesting from cannulated CM rats was a well-controlled in vivo functional screening method, and that the extracted phage not only bound to the vasculature but also functioned as a transporter across the blood-brain barrier. Furthermore, by simultaneously collecting blood and applying HTS to CSF and blood-derived phages, we confirmed that our choice of CSF was not influenced by blood enrichment or fitness for expansion between rounds of selection. However, the blood compartment is part of the selection procedure, since phages capable of reaching the CSF compartment must survive and circulate in the bloodstream long enough to enrich themselves in the brain. In order to extract reliable sequence information from raw HTS data, we implemented filters adapted to platform-specific sequencing errors in the analysis workflow. By incorporating kinetic parameters into the screening method, we confirmed the rapid pharmacokinetics of wild-type T7 phages (t½ ~ 28 min) in blood24, 27, 28 and also determined their half-life in cerebrospinal fluid (t½ ~ 26 min) per minute). Despite similar pharmacokinetic profiles in blood and CSF, only 0.001% of the blood concentration of phage could be detected in CSF, indicating low background mobility of wild-type T7 phage across the blood-brain barrier. This work highlights the importance of the first round of selection when using in vivo panning strategies, especially for phage systems that are rapidly cleared from the circulation, as few clones are able to reach the CNS compartment. Thus, in the first round, the reduction in library diversity was very large, as only a limited number of clones were eventually collected in this very strict CSF model. This in vivo panning strategy included several selection steps such as active accumulation in the CSF compartment, clone survival in the blood compartment, and rapid removal of T7 phage clones from the blood within the first 10 minutes (Fig. 1d and Supplementary Fig. 4M). ). Thus, after the first round, different phage clones were identified in CSF, although the same initial pool was used for individual animals. This suggests that multiple strict selection steps for source libraries with large numbers of library members result in a significant reduction in diversity. Therefore, random events will become an integral part of the initial selection process, greatly influencing the result. It is likely that many of the clones in the original library had a very similar CSF enrichment propensity. However, even under the same experimental conditions, selection results may differ due to the small number of each particular clone in the initial pool.
The motifs enriched in CSF differ from those in the blood. Interestingly, we noted the first shift towards glycine-rich peptides in the blood of individual animals. (Fig. 1g, Supplementary Figs. 4e, 4f). Phage containing glycine peptides may be more stable and less likely to be taken out of circulation. However, these glycine-rich peptides were not detected in the cerebrospinal fluid samples, suggesting that the curated libraries went through two different selection steps: one in the blood and another allowed to accumulate in the cerebrospinal fluid. CSF-enriched clones resulting from the fourth round of selection have been extensively tested. Nearly all of the individually tested clones were confirmed to be enriched in CSF compared to blank control phage. One peptide hit (#2077) was examined in more detail. It showed a longer plasma half-life compared to other hits (Figure 3d and Supplementary Figure 7), and interestingly, this peptide contained a cysteine residue at the C-terminus. It has recently been shown that the addition of cysteine to peptides can improve their pharmacokinetic properties by binding to albumin 29 . This is currently unknown for peptide #2077 and requires further study. Some peptides showed a valence-dependency in CSF enrichment (data not shown), which may be related to the displayed surface geometry of the T7 capsid. The T7 system we used showed 5-15 copies of each peptide per phage particle. IHC was performed on candidate lead phage clones injected intravenously into the cerebral cortex of rats (Supplementary Fig. 8). The data showed that at least three clones (No. 2002, No. 2009 and No. 2077) interacted with the BBB. It remains to be determined whether this BBB interaction results in the accumulation of CSF or the movement of these clones directly to the BCSFB. Importantly, we show that the selected peptides retain their CSF transport capacity when synthesized and bound to the protein cargo. Binding of N-terminal biotinylated peptides to SA essentially repeats the results obtained with their respective phage clones in blood and cerebrospinal fluid (Fig. 3e). Finally, we show that lead peptide #2077 is able to promote the brain action of a biotinylated peptide inhibitor of BACE1 conjugated to SA, causing pronounced pharmacodynamic effects in the CNS by significantly reducing Abeta40 levels in CSF (Fig. 4). We were unable to identify any homologues in the database by performing a peptide sequence homology search of all hits. It is important to note that the size of the T7 library is approximately 109, while the theoretical library size for 12-mers is 4 x 1015. Therefore, we only selected a small fraction of the diversity space of the 12-mer peptide library, which may mean that more optimized peptides can be identified by evaluating the adjacent sequence space of these identified hits. Hypothetically, one of the reasons why we have not found any natural homologues of these peptides may be deselection during evolution to prevent the uncontrolled entry of certain peptide motifs into the brain.
Taken together, our results provide a basis for future work to identify and characterize the transport systems of the cerebrovascular barrier in vivo in more detail. The basic setup of this method is based on a functional selection strategy that not only identifies clones with cerebral vascular binding properties, but also includes a critical step in which successful clones have intrinsic activity to cross biological barriers in vivo into the CNS compartment. is to elucidate the mechanism of transport of these peptides and their preference for binding to the microvasculature specific to the brain region. This may lead to the discovery of new pathways for the transport of the BBB and receptors. We expect that the identified peptides can directly bind to cerebrovascular receptors or to circulating ligands transported through the BBB or BCSFB. The peptide vectors with CSF transport activity discovered in this work will be further investigated. We are currently investigating the brain specificity of these peptides for their ability to cross the BBB and/or BCSFB. These new peptides will be extremely valuable tools for the potential discovery of new receptors or pathways and for the development of new highly efficient platforms for the delivery of macromolecules, such as biologics, to the brain.
Cannulate the large cisterna (CM) using a modification of the previously described method. Anesthetized Wistar rats (200-350 g) were mounted on a stereotaxic apparatus and a median incision was made over the shaved and aseptically prepared scalp to expose the skull. Drill two holes in the area of the upper sash and fasten the fixing screws in the holes. An additional hole was drilled in the lateral occipital crest for stereotactic guidance of a stainless steel cannula into the CM. Apply dental cement around the cannula and secure with screws. After photo-curing and cement hardening, the skin wound was closed with 4/0 supramid suture. Proper placement of the cannula is confirmed by spontaneous leakage of cerebrospinal fluid (CSF). Remove the rat from the stereotaxic apparatus, receive appropriate postoperative care and pain management, and allow it to recover for at least one week until signs of blood are observed in the cerebrospinal fluid. Wistar rats (Crl:WI/Han) were obtained from Charles River (France). All rats were kept under specific pathogen-free conditions. All animal experiments were approved by the Veterinary Office of the City of Basel, Switzerland, and were performed in accordance with Animal License No. 2474 (Assessment of Active Brain Transport by Measuring Levels of Therapeutic Candidates in the Cerebrospinal Fluid and Brain of Rat).
Gently keep the rat conscious with the CM cannula in hand. Remove Datura from the cannula and collect 10 µl of spontaneously flowing cerebrospinal fluid. Since the patency of the cannula was ultimately compromised, only clear cerebrospinal fluid samples with no evidence of blood contamination or discoloration were included in this study. In parallel, approximately 10–20 μl of blood was taken from a small incision at the tip of the tail into tubes with heparin (Sigma-Aldrich). CSF and blood were collected at various time points after intravenous injection of T7 phage. Approximately 5–10 μl of fluid was discarded before each CSF sample was collected, which corresponds to the dead volume of the catheter.
Libraries were generated using the T7Select 10-3b vector as described in the T7Select system manual (Novagen, Rosenberg et al., InNovations 6, 1-6, 1996). Briefly, a random 12-mer DNA insert was synthesized in the following format:
The NNK codon was used to avoid double stop codons and amino acid overexpression in the insert. N is a manually mixed equimolar ratio of each nucleotide, and K is a manually mixed equimolar ratio of adenine and cytosine nucleotides. Single stranded regions were converted to double stranded DNA by further incubation with dNTP (Novagen) and Klenow enzyme (New England Biolabs) in Klenow buffer (New England Biolabs) for 3 hours at 37°C. After the reaction, double-stranded DNA was recovered by EtOH precipitation. The resulting DNA was digested with restriction enzymes EcoRI and HindIII (both from Roche). The cleaved and purified (QIAquick, Qiagen) insert (T4 ligase, New England Biolabs) was then ligated in-frame into a pre-cleaved T7 vector after amino acid 348 of the 10B capsid gene. Ligation reactions were incubated at 16° C. for 18 hours prior to in vitro packaging. Phage packaging in vitro was performed according to the instructions supplied with the T7Select 10-3b cloning kit (Novagen) and the packaging solution was amplified once to lysis using Escherichia coli (BLT5615, Novagen). The lysates were centrifuged, titrated and frozen at -80° C. as a stock solution of glycerol.
Direct PCR amplification of phage variable regions amplified in broth or plate using proprietary 454/Roche-amplicon fusion primers. The forward fusion primer contains sequences flanking the variable region (NNK) 12 (template-specific), GS FLX Titanium Adapter A, and a four-base library key sequence (TCAG) (Supplementary Figure 1a):
The reverse fusion primer also contains biotin attached to capture beads and the GS FLX Titanium Adapter B required for clonal amplification during emulsion PCR:
The amplicons were then subjected to 454/Roche pyrosequencing according to the 454 GS-FLX Titanium protocol. For manual Sanger sequencing (Applied Biosystems Hitachi 3730 xl DNA Analyzer), T7 phage DNA was amplified by PCR and sequenced with the following primer pairs:
Inserts from individual plaques were subjected to PCR amplification using the Roche Fast Start DNA Polymerase Kit (according to the manufacturer’s instructions). Perform a hot start (10 min at 95 °C) and 35 boost cycles (50 s at 95 °C, 1 min at 50 °C, and 1 min at 72 °C).
Phage from libraries, wild-type phage, phage rescued from CSF and blood, or individual clones were amplified in Escherichia coli BL5615 in TB broth (Sigma Aldrich) or in 500 cm2 dishes (Thermo Scientific) for 4 h at 37°C. Phage were extracted from the plates by rinsing the plates with Tris-EDTA buffer (Fluka Analytical) or by collecting the plaques with sterile pipette tips. Phage were isolated from culture supernatant or extraction buffer with one round of polyethylene glycol (PEG 8000) precipitation (Promega) and resuspended in Tris-EDTA buffer.
The amplified phage was subjected to 2-3 rounds of endotoxin removal using endotoxin removal beads (Miltenyi Biotec) prior to intravenous (IV) injection (500 μl/animal). In the first round, 2×1012 phages were introduced; in the second, 2×1010 phages; in the third and fourth selection rounds, 2×109 phages per animal. Phage content in CSF and blood samples collected at the indicated time points was determined by plaque counting according to the manufacturer’s instructions (T7Select system manual). Phage selection was performed by intravenous injection of purified libraries into the tail vein or by re-injection of phage extracted from CSF from the previous selection round, and subsequent harvests were performed at 10 min, 30 min, 60 min, 90 min, 120 min, 180 min, and 240 min respectively CSF and blood samples. A total of four rounds of in vivo panning were conducted in which the two selected branches were separately stored and analyzed during the first three rounds of selection. All phage inserts extracted from CSF from the first two rounds of selection were subjected to 454/Roche pyrosequencing, while all clones extracted from CSF from the last two rounds of selection were manually sequenced. All blood phages from the first round of selection were also subjected to 454/Roche pyrosequencing. For injection of phage clones, selected phages were amplified in E. coli (BL5615) on 500 cm2 plates at 37°C for 4 hours. Individually selected and manually sequenced clones were propagated in TB medium. After phage extraction, purification and removal of endotoxin (as described above), 2×1010 phages/animal in 300 μl were injected intravenously into one tail vein.
Preprocessing and qualitative filtering of sequence data. Raw 454/Roche data was converted from a binary standard stream map format (sff) to a Pearson human readable format (fasta) using vendor software. Further processing of the nucleotide sequence was performed using proprietary C programs and scripts (unreleased software package) as described below. The analysis of primary data includes strict multi-stage filtering procedures. To filter out reads that did not contain a valid 12mer insert DNA sequence, the reads were sequentially aligned to start label (GTGATGTCGGGGATCCGAATTCT), stop label (TAAGCTTGCGGCCGCACTCGAGTA) and background insert (CCCTGCAGGGATATCCCGGGAGCTCGTCGAC) using the global Needleman-Wunsch test. alignment allowing up to 2 inconsistencies per alignment31. Therefore, reads without start and stop tags and reads containing background inserts, i.e., alignments that exceed the allowed number of mismatches, were removed from the library. As for the remaining reads, the N-mer DNA sequence extending from the start mark and ending before the stop mark was excised from the original read sequence and further processed (hereinafter referred to as “insert”). After translation of the insert, the portion after the first stop codon at the 5′ end of the primer is removed from the insert. In addition, nucleotides leading to incomplete codons at the 3′ end of the primer were also removed. To exclude inserts containing only background sequences, translated inserts beginning with the amino acid pattern “PAG” were also removed. Peptides with a post-translational length of less than 3 amino acids were removed from the library. Finally, remove redundancy in the insert pool and determine the frequency of each unique insert. The results of this analysis included a list of nucleotide sequences (inserts) and their (read) frequencies (Supplementary Figures 1c and 2).
Group N-mer DNA inserts by sequence similarity: To eliminate 454/Roche-specific sequencing errors (such as problems with sequencing homopolymer extensions) and remove less important redundancies, previously filtered N-mer DNA sequence inserts (inserts) are sorted by similarity. insertions (up to 2 non-matching bases allowed) using an iterative algorithm defined as follows: insertions are sorted first by their frequency (highest to lowest), and if they are the same, by their secondary sort by length (longest to shortest) ). Thus, the most frequent and longest insertions define the first “group”. The group frequency is set to the key frequency. Then, each insertion remaining in the sorted list was tried to be added to the group by pairwise Needleman-Wunsch alignment. If the number of mismatches, insertions, or deletions in an alignment does not exceed a threshold of 2, an insertion is added to the group, and the overall group frequency is increased by how often the insertion was added. Inserts added to a group are marked as used and excluded from further processing. If the insert sequence cannot be added to an already existing group, the insert sequence is used to create a new group with the appropriate insert frequency and marked as used. The iteration ends when each insertion sequence has either been used to form a new group or can be included in an already existing group. After all, grouped inserts consisting of nucleotides are eventually translated into peptide sequences (peptide libraries). The result of this analysis is a set of insertions and their corresponding frequencies that make up the number of consecutive reads (Supplementary Fig. 2).
Motif Generation: Based on a list of unique peptides, a library was created containing all possible amino acid patterns (aa) as shown below. Each possible pattern of length 3 was extracted from the peptide and its inverse pattern was added along with a common motif library containing all patterns (tripeptides). Libraries of highly repetitive motifs were sequenced and redundancy removed. Then, for each tripeptide in the motif library, we checked for its presence in the library using computational tools. In this case, the frequency of the peptide containing the found motif tripeptide is added and assigned to the motif in the motif library (“number of motifs”). The result of motif generation is a two-dimensional array containing all occurrences of tripeptides (motifs) and their respective values, which are the number of sequencing reads that result in the corresponding motif when the reads are filtered, grouped, and translated. Metrics as described in detail above.
Normalization of the number of motifs and corresponding scatterplots: The number of motifs for each sample was normalized using
where ni is the number of reads containing topic i. Thus, vi represents the percentage frequency of reads (or peptides) containing motif i in the sample. P-values for the non-normalized number of motifs were calculated using Fisher’s exact test. Regarding correlograms of the number of motives, Spearman’s correlations were calculated using the normalized number of motives with R.
To visualize the content of amino acids at each position in the peptide library, web logograms 32, 33 (http://weblogo.threeplusone.com) were created. First, the content of amino acids at each position of the 12-mer peptide is stored in a 20×12 matrix. Then, a set of 1000 peptides containing the same relative amino acid content at each position is generated in fasta-sequence format and provided as input to web-logo 3, which generates a graphical representation of the relative amino acid content at each position. for a given peptide library. To visualize multidimensional datasets, heat maps were created using an internally developed tool in R (biosHeatmap, a yet-to-be-released R package). The dendrograms presented in the heat maps were calculated using Ward’s hierarchical clustering method with the Euclidean distance metric. For statistical analysis of motif scoring data, P values for unnormalized scoring were calculated using Fisher’s exact test. P-values for other datasets were calculated in R using Student’s t-test or ANOVA.
Selected phage clones and phages without inserts were injected intravenously via the tail vein (2×1010 phages/animal in 300 μl PBS). Ten minutes prior to perfusion and subsequent fixation, the same animals were intravenously injected with 100 μl of DyLight594-labeled lectin (Vector Laboratories Inc., DL-1177). 60 minutes after phage injection, rats were perfused through the heart with 50 ml PBS followed by 50 ml 4% PFA/PBS. Brain samples were additionally fixed overnight in 4% PFA/PBS and soaked in 30% sucrose overnight at 4°C. Samples are flash frozen in the OCT mixture. Immunohistochemical analysis of frozen samples was performed at room temperature on 30 µm cryosections blocked with 1% BSA and incubated with polyclonal FITC-labeled antibodies against T7 phage (Novus NB 600-376A) at 4 °C. Incubate overnight. Finally, the sections were washed 3 times with PBS and examined with a confocal laser microscope (Leica TCS SP5).
All peptides with a minimum purity of 98% were synthesized by GenScript USA, biotinylated and lyophilized. Biotin is bound via an additional triple glycine spacer at the N-terminus. Check all peptides using mass spectrometry.
Streptavidin (Sigma S0677) was mixed with a 5-fold equimolar excess of biotinylated peptide, biotinylated BACE1 inhibitory peptide, or a combination (3:1 ratio) of biotinylated BACE1 inhibitory peptide and BACE1 inhibitory peptide in 5–10% DMSO/incubated in PBS. 1 hour at room temperature before injection. Streptavidin-conjugated peptides were injected intravenously at a dose of 10 mg/kg into one of the tail veins of rats with a cerebral cavity.
The concentration of streptavidin-peptide complexes was assessed by ELISA. Nunc Maxisorp microtiter plates (Sigma) were coated overnight at 4°C with 1.5 μg/ml mouse anti-streptavidin antibody (Thermo, MA1-20011). After blocking (blocking buffer: 140 nM NaCL, 5 mM EDTA, 0.05% NP40, 0.25% gelatin, 1% BSA) at room temperature for 2 hours, wash the plate with 0.05% Tween-20/PBS (wash buffer) for 3 Second, CSF and plasma samples were added to wells diluted with blocking buffer (plasma 1:10,000, CSF 1:115). The plate was then incubated overnight at 4°C with detection antibody (1 μg/ml, anti-streptavidin-HRP, Novus NB120-7239). After three washing steps, streptavidin was detected by incubation in TMB substrate solution (Roche) for up to 20 min. After stopping color development with 1M H2SO4, measure the absorbance at 450 nm.
The function of the streptavidin-peptide-BACE1 inhibitor complex was assessed by Aβ(1-40) ELISA according to the manufacturer’s protocol (Wako, 294-64701). Briefly, CSF samples were diluted in standard diluent (1:23) and incubated overnight at 4°C in 96-well plates coated with BNT77 capture antibody. After five washing steps, HRP-conjugated BA27 antibody was added and incubated for 2 hours at 4° C., followed by five washing steps. Aβ(1–40) was detected by incubation in TMB solution for 30 minutes at room temperature. After color development has been stopped with stop solution, measure the absorbance at 450 nm. Plasma samples were subjected to solid phase extraction prior to Aβ(1–40) ELISA. Plasma was added to 0.2% DEA (Sigma) in 96-well plates and incubated at room temperature for 30 minutes. After successively washing the SPE plates (Oasis, 186000679) with water and 100% methanol, plasma samples were added to the SPE plates and all liquid was removed. Samples were washed (first with 5% methanol then 30% methanol) and eluted with 2% NH4OH/90% methanol. After drying the eluate at 55°C for 99 min at constant N2 current, the samples were reduced in standard diluents and Aβ(1–40) was measured as described above.
How to cite this article: Urich, E. et al. Cargo delivery to the brain using transit peptides identified in vivo. the science. 5, 14104; doi:10.1038/srep14104 (2015).
Likhota J., Skjorringe T., Thomsen L.B. and Moos T. Delivery of macromolecular drugs to the brain using targeted therapy. Journal of Neurochemistry 113, 1–13, 10.1111/j.1471-4159.2009.06544.x (2010).
Brasnjevic, I., Steinbusch, HW, Schmitz, C., and Martinez-Martinez, P. Delivery of peptide and protein drugs across the blood-brain barrier. Prog Neurobiol 87, 212–251, 10.1016/j.pneurobio.2008.12.002 (2009).
Pardridge, W. M. The blood-brain barrier: a bottleneck in brain drug development. NeuroRx 2, 3–14, 10.1602/neurorx.2.1.3 (2005).
Johanson, K. E., Duncan, J. A., Stopa, E. G., and Byrd, A. Prospects for improved drug delivery and targeting to the brain via the choroid plexus-CSF pathway. Pharmaceutical Research 22, 1011–1037, 10.1007/s11095-005-6039-0 (2005).
Pardridge, W. M. Modernization of biopharmaceuticals with molecular Trojan horses for brain delivery. Bioconjug Chem 19, 1327–1338, 10.1021/bc800148t (2008).
Pardridge, WM receptor-mediated peptide transport across the blood-brain barrier. Endocr Rev. 7, 314–330 (1986).
Niewoehner, J. et al. Increase brain penetration and efficacy of therapeutic antibodies using monovalent molecular shuttles. Neuron 81, 49–60, 10.1016/j.neuron.2013.10.061 (2014).
Bien-Lee, N. et al. Transferrin receptor (TfR) transport determines brain uptake of affinity variants of TfR antibodies. J Exp Med 211, 233–244, 10.1084/jem.20131660 (2014).