Phosphatidylcholine-derived carriers facilitate brain tumor delivery and enhance glioblastoma therapy

Abstract

Glioblastoma (GBM) is the most common primary brain cancer without effective treatment. The ineffective treatment of GBM can be partially attributed to the existence of the blood-brain barrier (BBB). Lipids, which constitute over half the weight of the brain and play a vital role in brain tumor biology, can be transported to the brain in the form of lysophosphatidylcholines (LPCs) via specific LPC transporters at the BBB. We hypothesize that LPC analogs could be used as carriers for drug delivery to tumors in the brain. To test this hypothesis, we synthesized and screened a collection of LPC analogs, among which LPC analog 3 (A3), featuring a glycerophosphorylcholine (GPC) headgroup and a 15-carbon tail, exhibited a marked ability to penetrate brain tumors. We characterized A3 as a carrier for drug delivery to brain tumors by using Doxorubicin (Dox) as the therapeutic payload and found that the A3-Dox conjugate with a cathepsin B-cleavable linker has a great ability to accumulate in brain tumors, leading to effective treatment of GBM without inducing significant cytotoxicity. Our study suggests a novel approach to improving the treatment of GBM by enhancing the delivery of therapeutic agents to the brain using A3 as a carrier.

Graphical Abstract

A library of lysophosphatidylcholine (LPC) analogs was synthesized and screened. Among them, A3 demonstrated superior brain tumor penetration following intravenous administration. Delivery of doxorubicin via A3, using a cathepsin B-cleavable linker, significantly prolonged the survival of mice bearing gliomas, highlighting A3 as a promising drug carrier for brain cancer treatment.

Introduction

Glioblastoma (GBM) is the most common and deadly brain tumor in adults with a dismal prognosis. The median survival following the standard treatment regimen, including surgical resection, radiotherapy, and temozolomide chemotherapy, is only 12–15 months, and the five-year survival rate is just 5.5% [1]. A major obstacle to improving GBM treatment is the lack of efficient drug delivery methods to the brain, primarily due to the blood-brain barrier (BBB), which severely limits the penetration of most therapeutics [2, 3]. While convection-enhanced delivery (CED) can bypass the BBB through direct locoregional drug administration, its clinical application is restricted by its highly invasive nature and limited drug distribution [4, 5]. Systemic delivery presents a potential alternative but remains a significant challenge. Although the BBB is “leaky” in the tumor core, the capillaries at the tumor margins and surrounding brain tissue are nearly as impermeable as an intact BBB, preventing adequate drug penetration for a meaningful therapeutic response [6, 7]. Therefore, advancing GBM treatment requires innovative systemic drug delivery strategies capable of effectively overcoming the BBB.

In this study, we explored lysophosphatidylcholines (LPCs) as carriers for drug delivery to brain for treatment of GBM. Lipids constitute 50% to 60% of the brain’s total dry weight and play a vital role in brain tumor biology [8]. However, the brain has a limited capacity for de novo lipid synthesis and thus relies on lipid uptake from the peripheral circulation [9]. LPC has been shown to be the most efficient form for transport across the BBB via specific LPC transporters, such as the sodium-dependent LPC symporter 1 (Mfsd2a) [9–11]. We hypothesize that LPC analogs with a phosphocholine headgroup and a hydrophobic tail that mimic natural phospholipids could be used as carriers for drug delivery to brain tumors. To test this hypothesis, we synthesized and screened a collection of LPC analogs and identified analog 3 (A3), which has a glycerophosphorylcholine (GPC) headgroup and a 15-carbon fatty acid tail, as having the greatest efficiency in facilitating drug delivery to GBM in the brain. Using Doxorubicin (Dox), a highly potent anti-cancer drug that has limited ability to penetrate the brain for GBM treatment [12–14], as the therapeutic payload, we demonstrated that A3 is a promising carrier for drug delivery to the brain for the effective treatment of GBM. Our study suggests a promising approach for improving the treatment of GBM by enhancing drug delivery to the brain using A3 as a carrier.

Results and Discussion

Synthesis of LPC analogs

Using a tin-mediated mono-acylation of GPC [15], LPC analogs were synthesized with an azido moiety inserted between the carbon chain and the GPC headgroup (Figure 1A), or an azido-fatty acid attached to the GPC headgroup (Figure 1B). The resulting LPC and its analogs allow for the conjugation of DBCO-terminated payloads through a click reaction. Since the length of the fatty acid chain and headgroup composition are known key parameters for interaction with LPC transporters at the BBB [11], a small library of 18 LPC analogs with various carbon chain lengths and headgroups was synthesized, including 8 with different azido-fatty acid tails (A1–A8), 3 with inserted azido-lysine (B1–B3), 5 with altered headgroups (C1–C5), and 2 with double carbon chains (D1–D2). The chemical structures of all the compounds were verified by nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry. Triethylamine hydrochloride, a byproduct formed during the neutralization of triethylamine in the tin-mediated reaction, was observed in the ¹H NMR spectrum. This impurity is highly polar, water-soluble, and not UV-active, making it difficult to remove using standard column chromatography. Given that triethylamine hydrochloride is unlikely to affect the in vivo BBB permeability screening, we opted to use these lipids without further purification via C18 chromatography.

Figure 1.

Synthetic routes and structures of LPC analogs. (A) The Tin-mediated mono-acylation, followed by deprotection and substitution reactions, yielded material B1. Corresponding N-hydroxysuccinimide (NHS) esters were used to synthesize materials B2 and B3. (B) The azido-fatty acid was attached to the GPC headgroup through the Tin-mediated reaction. Compounds A1–A7 were synthesized in a similar manner using different fatty acids. (C) Structures of LPC analogs.

Characterization of LPC analogs for brain tumor penetration

We evaluated the potential of LPC analogs as carriers for drug delivery to brain tumors. To enable rapid screening, Cy7, a near-infrared dye, was selected as the model payload. Cy7-conjugated LPC analogs were synthesized and intravenously (IV) injected into C57BL/6J mice at 14 days post-intracranial inoculation of luciferase-expressing GL261 cells. Mice were grouped based on luciferase expression to ensure comparable tumor volumes across treatment groups. After 5 hours, the mice were euthanized after cardiac perfusion, major organs were harvested, and fluorescence imaging was performed ex vivo. The delivery efficiency of each material was assessed by normalizing Cy7 fluorescence intensity to luciferase expression (Figure 2A, B). We found that fatty acid chain length plays a critical role in brain penetration. Among the eight compounds in group A, A3, which has a 15-carbon tail, exhibited the highest tumor penetration, whereas compounds with either shorter or longer chains (A1–2, A4–7) showed significantly reduced penetration. Replacing the fatty acid tail with a hydrophilic polyethylene glycol chain (A8) dramatically reduced delivery efficiency, highlighting the importance of lipid hydrophobicity. Compared to A3, LPC analogs B1–3, including B2 with the same fatty acid chain length, showed significantly lower penetration, suggesting that conjugating the payload at the end of the carbon chain may be preferable. Analog C1, identical in structure to A3 but lacking the glycerol backbone, exhibited reduced tumor penetration, which was further diminished by replacing the phosphatidyl moiety with a carbonate ester (C2) or removing the choline group (C4, C5). Lipid C3, with an altered hydrophilic headgroup, also showed minimal penetration, indicating that the phosphocholine headgroup is essential for brain transport. Additionally, lipid D1, which features two fatty acid tails, exhibited penetration comparable to A3, while lipid D2, lacking the GPC moiety, showed negligible fluorescence in brain tumors. Taken together, these findings demonstrate that both the hydrophobic carbon chain and the phosphocholine headgroup are critical for efficient tumor penetration. Among all tested compounds, A3 exhibited the highest brain penetration (Figure 2A, B), a result further validated in a xenograft U87 human GBM mouse model (Figure S1). To confirm that triethylamine hydrochloride does not significantly impact in vivo screening for BBB penetration, we further purified A3 using reverse-phase C18 chromatography (Figure S12) and evaluated brain tumor penetration of A3-Cy7 in mice using the purified compound. As shown in Figure S13, the results were comparable, suggesting that the presence of triethylamine hydrochloride has a minimal effect on the BBB permeability of these analogs. Based on these findings, A3 was selected as the lead carrier for further studies.

Figure 2.

Brain tumor penetration of LPC analogues. (A) Ex vivo images showing lipid-Cy7 accumulation in orthotopic GL261-Luc glioma after intravenous administration. Tumor volume was measured by luciferase-based luminescence (LUC). (B) Semi-quantification of lipid-Cy7 in brain tumors after intravenous administration by normalizing Cy7 fluorescence intensity to luciferase expression. Data are expressed as mean ± SD (n = 3). The differences between lipid-Cy7 groups and the free Cy7 group were calculated using a two-tailed, unpaired t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

We next assessed the biodistribution of the A3-Cy7 following intravenous administration. Mice bearing luciferase-expressing GL261 gliomas were established and treated with A3-Cy7 14 days after inoculation. The mice were euthanized at 1 hour, 5 hours, and 16 hours after cardiac perfusion. Major organs, including the brain, heart, liver, spleen, lung, and kidneys, were isolated and imaged ex vivo. Our findings showed that the accumulation of A3-Cy7 in tumors peaked at 5 hours post-injection. By 16 hours, A3-Cy7 primarily accumulated in the liver and brain tumors. The amount of A3-Cy7 in tumors was higher than in all other organs except the liver (Figure 3A). Further imaging showed that A3 conjugates preferentially accumulated in tumor regions but not normal brain hemisphere (Figure 3B), with Cy7 fluorescence intensity in tumor hemisphere being 11.4 times higher than in the normal brain (Figure 3C). Compared to normal brain vasculatures, tumor-associated vasculature is more abundant and exhibits higher permeability. Therefore, it is not surprising to observe significantly stronger accumulation of A3 conjugates in brain tumors, consistent with the results seen for other BBB-permeable peptides, as shown in Figure 3E. However, this does not imply the absence of A3 conjugate accumulation in non-tumor brain regions. As shown in Figure S11, compared to free Cy7, A3 conjugation significantly enhanced the accumulation of A3-Cy7 in the normal brain, indicating that the A3 conjugate can cross the BBB even in the absence of a tumor. The lack of detectable signal in non-tumor regions in Figure 2A and Figure 3B can likely be attributed to imaging conditions. These images were acquired using an IVIS system or fluorescent microscopy with an autoexposure setting. Because the signal in tumor regions was much stronger, the system automatically selected a shorter exposure time to avoid saturation, which resulted in signals in non-tumor regions being too weak to visualize. Accumulation of A3 conjugates in non-tumor regions could likely be detected if imaging were performed on brains after tumor resection. The concentration of label-free A3 in brain tumors measured by LC-MS was 4.4 times higher than in the normal brain (Figure 3D), aligning with the trend observed for A3-Cy7.

Figure 3.

Assessment of A3 conjugates in brain tumors and other organs. (A) Quantitative analysis of A3-Cy7 and free Cy7 distribution in various organs at 1, 5, and 16 hours after intravenous injection (n = 3). Mice were inoculated with orthotopic GL261-luc glioma. (B) Fluorescence images of brain sections from mice intravenously treated with free Cy3 and A3-Cy3 (red). The tumor regions were delineated with white dotted circles. (C) Quantification of A3-Cy7 in the tumor hemisphere and normal cerebral hemisphere (n = 3). (D) Quantitative analysis of LPC A3 by LC-MS. (E) Ex vivo images and (F) semi-quantification of Cy7-conjugated A3 and other BBB-permeable peptides in brain tumors. Cy7 fluorescence was normalized to luminescence. Data are expressed as mean ± SD (n = 3). Differences between A3 and other peptides were evaluated by the unpaired t test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Chemical conjugation strategies to improve drug delivery to brain tumors have been previously explored, often through the use of peptides targeting the BBB or brain tumors. We compared A3 with peptides that are well-documented for this purpose, including angiopep-2 and L57 peptides which bind to LRP1 [16–18], cRGD peptide targeting integrins [19, 20], and peptide HAI with a high affinity for the transferrin receptor [21, 22]. All peptides, similar to A3, were conjugated with the Cy7 dye. The Cy7-conjugated peptides were then intravenously administered to mice bearing GL261 gliomas and evaluated using the same procedures described above. The amount of conjugates was normalized, ensuring each mouse received an equivalent dose of Cy7 fluorescence. Our results indicated that A3 outperformed these peptide carriers, exhibiting enhanced fluorescent intensity in brain tumors (Figure 3E, F). This underscores the significant potential of A3 as a carrier for drug delivery to brain tumors.

Design and characterization of A3 conjugates for Dox delivery to GBM

We evaluated A3 as a drug carrier for brain tumor treatment by using Dox, a potent anti-cancer drug that cannot penetrate the brain to treat brain cancer [12–14], as the therapeutic payload. Three A3-Dox conjugates with different linkers, including Dox-NH-A3, Dox-CO-A3, and Dox-OH-A3, were synthesized (Figure 4A). Dox-NH-A3 consists of a linker that is cleavable by cathepsin B (Val-Cit-PAB) [23]. Since the Val-Cit-PAB linkage differs from the carbon chain linker used in A3-Cy7 and A3-Cy3, we evaluated A3-Val-Cit-PAB-Cy7 in mice bearing GL261 gliomas and confirmed that Val-Cit-PAB does not impact A3’s ability to enhance brain tumor penetrability (Figure S2). Dox-CO-A3 contains an acid-sensitive linker attached to the ketone group at position C13 of Dox [24]. Dox-OH-A3 was synthesized by conjugating Dox to the hydroxyl group at position C14, and it is not designed for stimuli-responsive drug release. The cytotoxicity of the three A3-Dox conjugates was evaluated in GL261 cells. Compared to free Dox, Dox-NH-A3 and Dox-CO-A3, with conjugations at the amine and ketone groups respectively, showed reduced cytotoxicity. In contrast, Dox-OH-A3, with A3 attached to the hydroxyl group at position C14 which does not significantly affect the bioactivity of Dox [24, 25], exhibited minimal loss in cytotoxicity (Figure 4B).

Figure 4.

Structures and evaluation of Doxorubicin-A3 conjugates. (A) Structures of Dox-NH-A3, Dox-CO-A3, and Dox-OH-A3. (B) Cytotoxicity of free Dox and Dox-A3 conjugates against GL261-luc after 72 hours’ incubation. (C) Drug release of Dox-NH-A3 in the presence of cathepsin B. (D) Mass spectrum of released Dox and fragment from Dox-NH-A3. (E) Acid-triggered cleavage of Dox-CO-A3 after 18 hours of incubation at different pH values. (F) Quantitative analysis of Dox-A3 conjugates in brain tumors by LC-MS. Data are presented as mean  ±  SD (n = 3). NA, not applicable.

Next, we characterized the drug release properties of Dox-NH-A3 and Dox-CO-A3 using LC-MS. Val-Cit-PAB cleavage was detected within 30 minutes in the presence of cathepsin B, with complete Dox release observed within 6 hours (Figure 4C). The drug and released fragments were confirmed via mass spectrometry (Figure 4D). The expression of cathepsin B in GL261 and U87 cells was verified by Western Blot (Figure S3). Since cathepsin B is mainly located in endosome/lysosome in cells, we evaluated the subcellular distribution of Dox-NH-A3 using confocal microscopy (Figure S8). Co-staining images showed that Dox-NH-A3 partially colocalized with Lysotracker Deep Red, indicating accumulation in endosomes/lysosomes following cellular uptake. This supports the mechanism of cathepsin B–responsive drug release. Acid-triggered drug release of Dox-CO-A3 was measured after 18 hours of incubation at various pH values. Dox was released in acidic conditions, with only 30.4% of Dox-CO-A3 remaining at pH 5.0 (Figure 4E). Because esterases are present within cells, we investigated the hydrolysis of Dox-OH-A3 following cellular uptake. GL261 cells were treated with Dox-OH-A3, then washed and homogenized. The resulting cell lysates were analyzed by LC-MS. As shown in the Figure S10a, significant hydrolysis and release of free Dox were observed at 24- and 48-hours post-treatment. We also characterized the blood circulation profile of Dox-OH-A3 following intravenous administration (Figure S10b). We found that Dox-OH-A3 was rapidly cleared from the bloodstream. No free Dox was detected in this blood, likely due either to the short in vivo half-life of free Dox (Figure S9) or to minimal hydrolysis of Dox-OH-A3 in the blood.

All three drug conjugates were evaluated for brain tumor penetration in mice bearing GL261 gliomas. Mice were grouped based on luciferase expression to ensure comparable tumor volumes and received intravenous administration of Dox-NH-A3, Dox-CO-A3, or Dox-OH-A3. Mice receiving free Dox served as controls. After 5 hours, cardiac perfusion was performed, and the mice were euthanized. Tumors in the brain were harvested and homogenized for quantification by LC-MS, which detected Dox and its conjugates with different sensitivities, with the lowest detectable concentrations being 0.53 ng/g for Dox, 10.29 ng/g for Dox-NH-A3, 92.59 ng/g for Dox-CO-A3, and 30.86 ng/g for Dox-OH-A3. As expected, no Dox was detected in tumors from the control group. In tumors from mice receiving Dox-CO-A3, only free Dox, not the parent drug, was detected (Figure 4F). Given that free Dox cannot penetrate brain tumors and complete drug release within 5 hours is unlikely (Figure 4E), we speculated that the detected free Dox was released from Dox-CO-A3 within the brain tumors, while the parent drug’s concentration was below LC-MS detection limits for Dox-CO-A3 (92.59 ng/g). In tumors from mice treated with Dox-OH-A3, neither free Dox nor the parent drug was detected. In contrast, significant drug conjugate and a small amount of free Dox, potentially released from the parent drug, were detected in tumors from mice receiving Dox-NH-A3 (Figure 4F). A comparable degree of drug accumulation was observed in the U87 GBM xenograft model (Figure S4), validating the brain tumor penetrability of Dox-NH-A3. The blood circulation profiles of Dox and Dox-NH-A3 after intravenous administration were also determined using LC-MS (Figure S9). In mice treated with free Dox, the drug was detectable in the blood immediately after injection (0 hour), but not at 2, 5, 16, or 24 hours, consistent with its short half-life [26]. In contrast, Dox-NH-A3 exhibited prolonged circulation without detectable release of free Dox in the bloodstream. These results suggest that A3 conjugation not only enhances brain tumor penetration but also improves the metabolic stability of Dox-NH-A3, making it a promising candidate for brain tumor therapy.

Evaluation of Dox-A3 conjugates for GBM treatment

We evaluated Dox-NH-A3 for brain cancer treatment in the GL261 orthotopic mouse model. Six days after inoculation with luciferase-expressing GL261 cells, the mice were randomly grouped and received intravenous treatment of Dox-NH-A3, PBS, or free Dox. Dox-NH-A3 was administered at a dose of 16 mg/kg, equivalent to 5 mg/kg for free Dox, the maximum dose safely administered to mice. Treatments were repeated every 2–3 days for a total of 12 days. The mice were monitored for tumor growth using luciferase imaging, body weight changes, and survival. We found that treatment with Dox-NH-A3, but not free Dox, significantly inhibited tumor growth and prolonged the survival of tumor-bearing mice by 12 days (P = 0.0004) (Figure 5A–C). H&E staining confirmed that mice receiving Dox-NH-A3 treatment had the smallest tumor lesions (Figure 5E). Terminal deoxynucleotidyl transferase (TUNEL) staining revealed massive cell apoptosis in tumors from mice treated with Dox-NH-A3, but not in those treated with PBS (Figure 5E). Treatment with Dox-NH-A3 appeared safe, without causing significant loss in body weight throughout the study (Figure 5D). Blood analysis of aspartate aminotransferase (AST), alanine aminotransferase (ALT), and blood urea nitrogen (BUN) did not indicate significant hepatotoxicity or renal toxicity (Figure 5F).

Figure 5.

In vivo therapeutic efficacy and toxicity profiles of Dox-A3 conjugates in mice. (A) Tumor volume measured once a week by luminescence from GL261-luc inoculated in the brain. (B) Quantitative analysis of tumor growth. Data were expressed as mean  ±  SD (n = 7). (C) Kaplan–Meier survival analysis of mice treated with the indicated drugs (n = 7). The dosage of Dox-NH-A3 was equivalent to 5 mg/kg of Dox. (D) Changes in body weight during the treatments. (E) H&E and TUNEL staining of brain tumors isolated from mice received the indicated treatments. (F) Analysis of AST, ALT and BUN in blood of mice receiving the indicated treatments. Data are expressed as mean ± SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, NS, no significant difference.

We also assessed the therapeutic efficacy of Dox-CO-A3 and Dox-OH-A3, which compared to Dox-NH-A3 demonstrated less brain tumor penetrability (Figure 4F). Treatments with both conjugates showed a decreased therapeutic effect in mice bearing GL261 gliomas, even when administered at a dose equivalent to 10 mg/kg of Dox (Figure S5). While 10 mg/kg is a lethal dose for free Dox, both conjugates did not significantly reduce body weight nor induce significant hepatotoxicity or renal toxicity (Figure S6).

Mechanism of A3-mediated brain penetration

A3, featuring a 15-carbon tail and a GPC headgroup, is similar to LPC, a natural substrate of the Mfsd2a transporter that is highly expressed in the endothelium of the BBB and facilitates the uptake of plasma fatty acids in the form of LPC [11]. We speculated whether A3 penetrates the brain via the Mfsd2a transporter. To test this hypothesis, we overexpressed the Mfsd2a transporter in HEK293 cells. The increased expression was validated by Western Blot (Figure S7). We treated both control and Mfsd2a-overexpressing HEK293 cells with A3-Cy3 or NBD-LPC, a fluorescent lipid known to be a substrate of Mfsd2a [11]. As expected, HEK293 cells overexpressing Mfsd2a exhibited significantly greater uptake of NBD-LPC. However, the uptake of A3-Cy3 was comparable in cells regardless of Mfsd2a expression levels (Figure 6A), suggesting that A3 may not interact with the Mfsd2a transporter. To confirm these observations, we tested A3-Cy7 in mice bearing GL261 gliomas, with and without pretreatment with LPC-DHA, another natural substrate of Mfsd2a [11]. We found that pretreatment with a 200-fold molar excess of LPC-DHA did not significantly reduce brain tumor penetration of Cy7-A3 (Figure 6B, C).

Figure 6.

Mechanism of A3-mediated uptake. (A) Fluorescence images of A3-Cy3 uptake in HEK293 cells with or without Mfsd2a transfection at 0.5 h and 4 h. Scale bar: 100 μm. (B) In vivo competition assay with natural Mfsd2a ligand, LPC-DHA. (C) Semi-quantification of A3-Cy7 in brain tumors with or without LPC-DHA pretreatment. Data are expressed as mean ± SD (n = 3). (D) Flow cytometry analysis of A3-Cy3 uptake in bEnd.3 cells at 4°C and 37°C. (E) ATP depletion by NaN₃ in bEnd.3 cells after 1 h and 3 h of incubation. (F) Flow cytometry analysis of A3-Cy3 uptake in bEnd.3 cells following 2-hour incubation. Cells were treated with or without 30 mM NaN₃ one hour prior to material addition. Data are presented as mean  ±  SD (n = 3). **** indicates p < 0.0001, NS, no significant difference.

To exclude the possibility that A3 penetrates the brain via passive diffusion, like many lipophilic small molecules [27], and albumin-associated free fatty acids [9], we compared the uptake of A3-Cy3 in bEnd.3 mouse brain endothelial cells at 4°C and 37°C. Active transport, such as endocytosis, is inhibited at low temperature [28]. Quantitative flow cytometry analysis revealed a drastic decrease in fluorescent intensity in bEnd.3 cells when the temperature was lowered from 37°C to 4°C (Figure 6D), suggesting a temperature-dependent process. We further examined the effect of ATP depletion on A3-Cy3 uptake in bEnd.3 cells using sodium azide (NaN₃), an agent that depletes ATP and inhibits cellular ATP production by affecting mitochondrial oxidative phosphorylation [29, 30]. ATP depletion by NaN₃ in bEnd.3 cells was validated using a luciferase luminescence assay (Figure 6E). Flow cytometry analysis showed that A3-Cy3 uptake was significantly reduced by NaN₃, indicating that the process is energy-dependent (Figure 6F). Collectively, these findings suggest that A3 crosses the blood-brain barrier via an active transport mechanism, likely independent of Mfsd2a-mediated transport.

Discussion

Delivering therapeutic agents to brain tumors remains challenging due to the presence of the BBB [31]. Unlike nanoparticle approaches, where nanoparticles are engineered for brain penetration through surface functionalization and thus are inherently complex [32, 33], the drug conjugate approach offers a more straightforward and simplified solution, enhancing potential clinical translation [34, 35]. In this study, we synthesized and characterized a small library of LPC analogs and identified A3 as a material that efficiently penetrates brain tumors (Figure 1B). We demonstrated that A3 could be used as a carrier for the efficient delivery of drugs to brain tumors. Using Doxorubicin (Dox) as the therapeutic payload, we showed that the delivery efficiency is related to the type of linker used for drug conjugation. Among three different linkers, including a cathepsin B-responsive linker, an acid-labile linker, and an ester bond linker, the cathepsin B-responsive linker allowed the greatest efficiency in brain tumor penetration. Consistently, treatment with Dox-NH-A3, which contains the cathepsin B-responsive linker, most effectively inhibited tumor growth without significant side effects.

The current understanding of lipid transport across the brain remains limited. Passive diffusion has long been considered the primary mechanism [9, 36]. However, recent studies suggest that some lipids may traverse the BBB via receptor-mediated transcytosis involving low-density lipoprotein (LDL) receptors or through internalization mechanisms mediated by fatty acid transport proteins (FATPs) [9]. The discovery of Mfsd2a further shifted this paradigm, identifying it as a critical transporter for the uptake of LPCs [11]. Notably, although A3 shares structural features with known Mfsd2a ligands, it appears to penetrate the brain through a mechanism independent of Mfsd2a. Continued advances in understanding lipid transport may provide valuable insights into the exact mechanism underlying A3-mediated delivery.

Development of LPC analogs as carriers for drug delivery to the brain was previously explored. PC-DHA, an LPC analog, was reported to enhance neuronal uptake and brain retention of small interfering RNA (siRNA) following local injection, though it exhibited limited brain distribution after systemic administration [37, 38]. PC-DHA differs from LPC A3 in that it includes a linker between the GPC head and the fatty acid tail. Our study found that this type of modification, as seen with LPC B1–B3 in Figure 2, was a less effective approach for facilitating brain penetration. This suggests that conjugating the payload at the end of the LPC carbon chain may be crucial for brain tumor penetrability, although PC-DHA was not evaluated in brain tumors. The ability of LPC A3 to deliver other macromolecular payloads, such as siRNA and ASO, presents a promising avenue for future research.

In summary, we synthesized and screened a series of LPC analogs and identified A3 as having a superior ability to penetrate brain tumors following intravenous administration. A3 demonstrated greater efficiency compared to peptide ligands, including L57, cRGD, HAI, and angiopep-2, which have been documented for their ability to enable drug delivery to the brain. We established A3 as a carrier for drug delivery to brain tumors, though the exact mechanism of its brain penetration remains to be determined. Our study suggests that A3 is a promising carrier for drug delivery to brain tumors and may offer an effective solution for the treatment of GBM.

Materials and Methods

Materials

Lipids C3, D1 and NBD-LPC were purchased from Avanti Polar Lipids. Lipid D2 was purchased from BroadPharm. Lipid C5 was purchased from AA blocks. Mfsd2a plasmid was purchased from Genscript. Mfsd2a and Cathepsin B antibodies were purchased from Proteintech (67965-1-Ig) and CST (31718S), respectively. Peptides including cRGD ([Cyc(1,9)]H-CRGDKGPDC(KAzide)-NH₂), HAI (H-HAIYPRH(KAzide)-NH₂), L57 (H-TFFYGGSRGKRNNFKTEEYK(N₃)-OH), and Angiopep-2 (H-TFFYGGSRGKRNNFKT EEYK(N₃)-OH) were purchased from Biosynth. Sulfo-Cy7-DBCO, Sulfo-Cy3-DBCO and Sulfo-Cy7-amine were purchased from Lumiprobe. GL261, HEK293, U87 and bEnd.3 cells were obtained from the American Type Culture Collection (ATCC, USA). Cells were cultured in Dulbecco’s modified Eagle medium (DMEM, Gibco) supplemented with 10 % fetal bovine serum (Gibco) and 1% penicillin-streptomycin (Gibco) at 37 °C with 5 % CO₂.

Animal models and materials screening

All animal studies follow the ethical guidelines set by the Yale University Institutional Animal Care and Utilization Committee (IACUC). The procedure of GL261-luc inoculation in female C57BL/6 mice (Charles River Laboratories, USA) was described in previous study [14] Briefly, mice were anesthetized by ketamine and xylazine via intraperitoneal injection. 5 ×10⁴ GL261-luc cells in 5 μl PBS were injected into the right striatum using a stereotactic apparatus with an UltraMicroPump (UMP3) (World Precision Instruments, USA). The injection site was 3 mm below the cortical surface, 2 mm laterally, and 0.5 mm posterior to the bregma. The tumor growth was monitored by luciferase-based luminescence on the In Vivo Imaging System (IVIS, Xenogen). To establish the U87 xenograft model, 2 ×10⁵ U87-luc cells in 5 ul PBS were injected into female BALB/c nude mice. The mice were grouped according to the tumor volume two weeks after inoculation.

Sulfo-Cy7-DBCO (Lumiprobe) was stirred with 3-fold excess of materials A-D in H₂O (or 5% DMSO) for 2 hours, which was monitored by TLC (DCM/MeOH). Once the sulfo-Cy7-DBCO disappeared on TLC plate, these fluorescence-labeled lipids were intravenously injected into mice. After 5 hours, cardiac perfusion with 20 ml PBS was applied under anesthesia, and then mice were euthanized. The isolated brain tissues were stored in 4% paraformaldehyde. The fluorescence intensity in brain tumors was measured by the In Vivo Imaging System (IVIS, Xenogen).

Cryosection

Sulfo-Cy3-DBCO (Lumiprobe) was conjugated to A3 via the same way described above for the synthesis of Sulfo-Cy7-A3. Sulfo-Cy3-A3 and free Sulfo-Cy3-DBCO were injected into mice via tail vein. After 5 hours, cardiac perfusion was applied under anesthesia, and then mice were sacrificed. The isolated brain tissues were stored in 4% paraformaldehyde and then dehydrated in 5% sucrose. The brain tissue was embedded in the Tissue-Tek^® O.C.T. compound, which was frozen at −20°C and sliced into 15 μm sections. The images were obtained on a fluorescence microscope (Keyence BZ series, USA).

Biodistribution of A3-Cy7

Female C57BL/6 mice inoculated with GL261-luc glioma for 14 days were intravenously injected with free Sulfo-Cy7-DBCO (Lumiprobe) or A3-Cy7. After 1, 5 or 16 hours, cardiac perfusion with 20 ml PBS was applied under anesthesia, and then organs were harvested and fixed in 4% paraformaldehyde. The fluorescent intensity in brain tumor, heart, liver, spleen, lung and kidney was measured and quantified via the In Vivo Imaging System (IVIS, Xenogen). To measure the BBB permeability of A3-Cy7 in normal brain, female C57BL/6 mice without tumors were intravenously injected with free Sulfo-Cy7-DBCO or A3-Cy7, followed by cardiac perfusion and euthanized at 5 hours after administration. The brain was collected and imaged ex vivo.

Transfection of Mfsd2a

HEK293 cells were seeded in 6-well plates to achieve 80% confluence on the second day. Mfs2a plasmids (Genscript) were transfected via FuGENE^® HD Transfection Reagent (Promega) following the manufacturer’s instructions. After 48 hours, the cells were treated with 50 μM of NBD-LPC or A3-Cy3 and inoculated for another 0.5 h or 4 h. Cells were imaged on the fluorescence microscope (Keyence BZ series, USA).

ATP depletion by NaN₃

bEnd.3 cells in 100 μL medium were seeded in 96-well plates to achieve 90% confluence. NaN₃ (Sigma) were added to cells at different concentrations. After 1 hour or 3 hours of incubation at 37°C, the luminescence was measured by CellTiter-Glo^® 2.0 (Promega). As the CellTiter-Glo^® 2.0 assay primarily reflects cell viability rather than directly measuring intracellular ATP levels, the observed decrease in luminescence serves as an indirect indicator of ATP depletion.

A3-Cy3 uptake in bEnd.3

bEnd.3 cells in 2 ml medium were seeded in 6-well plates to achieve 90% confluence. To evaluate the effect of ATP on cell uptake of A3 conjugate, 30 mM of NaN₃ was applied one hour before the addition of fluorescent materials. To explore the effect of low temperature, cells were moved to 4°C one hour ahead of material addition. And then, A3-Cy3 was added to cells at a concentration of 5 μM, followed by 2–3 hours of incubation at 37°C or 4°C. Cells were then collected and washed for flow cytometry analysis.

Cell viability

1 ×10⁴ GL261 cells in 100 μL medium per well were seeded in 96-well plates. After 24 h, 20 μL medium containing drugs was added. Cells were incubated at 37% for 3 days. CellTiter-Glo^® 2.0 Cell Viability reagent (Promega) was added to cells following the manufacturer’s instructions. The luminescence was measured by a Spectramax M5 Microplate Reader (Molecular Devices, USA). The IC₅₀ was calculated by GraphPad Prism 10 software.

In vitro drug release

Dilute Cathepsin B (Acrobiosystem, CTB-M52H9) to 100 μg/mL in Activation Buffer (25 mM MES, 5 mM DTT, pH 5.0). Incubate at room temperature for 15 minutes according to the manufacturer’s instructions. 50 μL of Dox-NH-A3 in assay buffer (25 mM MES, pH 5.0) was mixed with 50 μL of cathepsin B (100 μg/mL) solution. 10 μL of the reaction mixture was taken out at 0 h, 0.5 h, 2 h, and 6 h and added to 1 mL of MeOH for LC-MS testing. For acid-triggered drug release, Dox-CO-A3 was incubated in PBS buffer at different pH values for 18 h. 20 μL of PBS buffer was added to LC-MS vials and diluted with 1 mL MeOH.

Subcellular distribution of Dox-NH-A3

U87 cells were seeded in 20 mm culture dishes (Nest scientific). On the second day, cells were treated with 3 μM Dox-NH-A3. 18 h later, the cells were stained with Hoechst 33342 and 70 nmol lysotracker deep red (Invitrogen) for 1h. The subcellular distribution was measured by a laser scanning confocal microscope Leica SP8 Gated (Leica, Germany). Images were generated on 60x oil objective using DAPI light cube for Hoechst 33342, Cy3 for Dox-NH-A3 and Cy5 for Lysotracker deep red. The scale bar is 10 μm.

LC-MS measurement

Mass spectrometric measurements were performed with a Shimadzu Scientific Instruments quadrupole time-of-flight (QToF) 9030 LC-MS system, equipped with a Nexera LC-40D xs UHPLC, consisting of a CBM-40 Lite system controller, a DGU-405 Degasser Unit, two LC-40D XS UHPLC pumps, a SIL-40C XS autosampler and a Column Oven CTO-40S. UV data was collected with a Shimadzu Nexera HPLC/UHPLC Photodiode Array Detector SPD M-40 in the range of 190 – 800nm. Samples were held in the autosampler compartment. LC was performed with a Shim-pack Scepter C18-120, 1.9 um, 2.1×100 mm Column, equilibrated at 4°C in a column oven. A gradient eluent with Acetonitrile and Water (0.1% FA) in a mode of 12 min was applied. Flow was held constant at 0.4000 mL/min and the composition of the eluent was changed according to the following gradient: 0 to 2 min, held at 95% A, 5% B; 2 to 8 min, change to 5% A, 95% B; 8 to 10 min, held at 5% A, 95% B; 10 to 10.01 min, change to 95% A, 5% B; 10.01 to 12 min, held at 95% A, 5% B. The ionization source was run in “ESI” mode, with the electrospray needle held at +4.5kV. Nebulizer Gas was at 2 L/min, Heating Gas Flow at 10 L/min. Mass spectra were recorded in the range of 100 to 2000 m/z in positive ion mode. Measurements and data post-processing were performed with LabSolutions 5.97 Realtime Analysis and PostRun. The LC-MS tests were completed by the Chemical and Biophysical Instrumentation Center at Yale University.

Drug elimination in blood

The mice were administrated with a single dose of Dox (5 mg/kg) or Dox-NH-A3 (16 mg/kg, equivalent to 5 mg/kg free Dox) intravenously. The blood samples were harvested at 0, 2, 5, 16 and 24h, which were centrifuged immediately at 1200g for 10 min to obtain plasma. Due to the low blood volume of mice, samples at 0, 2 and 5 hours were collected from the same mice and the other three mice were used for collection at 0, 16 and 24 hours. Mice were sacrificed after the last sample harvest. 1 ml MeOH was added to 20ul plasma to precipitate protein, followed by centrifugation at 5000g. The supernatant was filtered before LC-MS analysis. Mass peaks at 544.1741 m/z (M+H⁺) and 879.9001 m/z (M+2H⁺) were extracted for quantification of Dox and Dox-NH-A3, respectively. The stability of Dox-OH-A3 in blood was evaluated using the same method with mass peak at 677.3049 m/z (M+2H⁺) extracted.

Hydrolysis of Dox-OH-A3 within cells

GL261 cells were planted to a 12-well plate on day 1 and Dox-OH-A3 was added at a concentration of 10 μM on Day 2. After 24 or 48 hours, cells were washed with PBS and homogenized in 0.2 mL H₂O, which was mixed with 1 mL MeOH. The mixture was subjected to an ultrasound for 2 minutes. After centrifugation at 4°C for 10 min, the supernatant was filtered before LC-MS analysis. Mass peaks at 544.1741 m/z (M+H⁺), and 677.3049 m/z (M+2H⁺) were extracted for Dox and Dox-OH-A3, respectively.

Quantification of drugs in the brain tumors

Female C57BL/6 mice were intracranially injected with GL261-luc cells as described above. Tumor volume was measured by IVIS. After 20 days, the mice were treated with Dox-NH-A3, Dox-CO-A3, or Dox-OH-A3 at a dosage equivalent to 20 mg/kg Dox via tail vein. For the U87 model, nude mice were injected with U87-luc and received intravenous administration of Dox-A3 12 days after tumor inoculation. At 5 hours post-injection, cardiac perfusion with 20 ml PBS was applied under anesthesia, and then mice were sacrificed. The brain tumors (100–200 mg) were harvested and homogenized in 1 mL H₂O, which was mixed with 5 mL MeOH. The mixture was subjected to an ultrasound for 2 minutes. After centrifugation at 4°C for 10 min, the supernatant was collected and concentrated to 1 mL, which was filtered before LC-MS analysis. Standard curves for each drug were established for mass spectrum-based quantitative analysis. Mass peaks at 544.1741 m/z (M+H⁺), 879.9001 m/z (M+2H⁺), 637.2966 m/z (M+2H⁺), and 677.2934 m/z (M+2H⁺) were extracted for quantification of Dox, Dox-NH-A3, Dox-CO-A3 and Dox-OH-A3, respectively. To compare the penetration of LPC A3 in brain tumors and normal brain, mice bearing GL261-luc cells were intravenously injected with LPC A3 at a dose of 20 mg/kg. At 5 hours post-injection, cardiac perfusion was applied under anesthesia. The tumor hemisphere and the normal cerebral hemisphere (100–200 mg) were harvested and homogenized in 1 mL H₂O. The 200 ul tissue suspension was mixed with 800 ul MeOH, which was subject to an ultrasound for 2 minutes. After centrifugation at 4°C for 10 min, the supernatant was collected and filtered before LC-MS analysis. The standard curve was established for mass spectrum-based quantitative analysis. The mass peak at 523.3255 m/z (M+H⁺) was extracted for quantification of LPC A3.

Therapeutic study

Female C57BL/6 mice (6–8 weeks) bearing orthotopic GL261-luc glioma were randomly grouped on day 6 post-implantation. Mice received biweekly treatments for two weeks with PBS, Dox (5 mg/kg), Dox-NH-A3 (equivalent to 5 mg/kg Dox), or Dox-OH-A3 (equivalent to 10 mg/kg Dox). Dox-CO-A3 (equivalent to 10 mg/kg Dox) were injected three times a week for two weeks. Tumor volume was measured weekly via the In Vivo Imaging System (IVIS, Xenogen). The brains were collected and fixed for immunohistochemistry analysis.

H&E and TUNEL staining

Tissues were fixed in 4 % PFA, embedded in paraffin, and sectioned to a thickness of 5 μm, followed by standard H&E and TUNEL staining procedures. This experiment was completed by the Pathology-based Central Tissue Resource Lab (YPTS) at Yale University. Slides were imaged on the fluorescence microscope (Keyence BZ series, USA).

AST, ALT and BUN assays

Normal C57BL/6 mice were intravenously injected with Dox-NH-A3 (equivalent to 5 mg/kg of Dox), Dox-CO-A3, and Dox-OH-A3 (equivalent to 10 mg/kg of Dox) twice a week for two weeks, which was identical with the therapeutic study. On the third day after the last dose, blood was collected from the orbital venous plexus of mice under anesthesia. The supernatant was collected in the tubes with heparin after centrifugation at 2,000 × g for 15 minutes. Samples were stored at −80°C. The activities of aspartate aminotransferase (Abcam, 105135), alanine transaminase (Abcam, 105134), and the urea nitrogen level (Invitrogen) were tested following manufacturers’ instructions.

Statistical Analysis

In vitro tests were performed in triplicate. All data were presented as the means ± standard deviations (SD). Survival study was analyzed based on Kaplan–Meier analysis. Differences between experimental groups were evaluated by the unpaired t test using Prism 10 (GraphPad) *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

1. A library of phosphatidylcholine-derived carriers was synthesized and screened in vivo.

2. Among them, LPC A3 demonstrated superior ability to penetrate brain tumors following intravenous administration.

3. LPC A3 demonstrated greater delivery efficiency compared to previously reported peptide ligands, including L57, cRGD, HAI, and angiopep-2.

4. Intravenous administration of doxorubicin-A3 conjugate with a cathepsin B-cleavable linker significantly prolonged the survival of mice bearing gliomas.

Data Availability

Data will be made available on request.