3.1 tRF-27-ZDXPHO53KSN predicted trastuzumab resistance in breast cancer
In previous studies, we have focused on the role of tRNA-derived fragments in trastuzumab resistance in breast cancer21. We selected three cell lines, including trastuzumab-sensitive cell line SKBR3, the trastuzumab-resistant cell line JIMT1, and HBL-100 from breast epithelial cells as controls, for RNA sequencing (Fig. 1A). In our study,, 36 upregulated and 21 downregulated tRNA-derived fragments were screened between HBL-100 and SKBR3/JIMT1 cell lines, as well as 11 upregulated and one downregulated between SKBR3 and JIMT1 cell lines. Based on these results, we selected five upregulated genes (tRF-30-JZOYJE22RR33, tRF-27-ZDXPHO53KSN, tRF-26-XIP2801MK8E, tRF-29-IYEVFMD0SR1Z, and tRF-22-8B8871K92) and one downregulated gene (tRF-30-SERXPIN2NYDR) for subsequent studies.
qRT-PCR confirmed that the expression levels of the tRNA-derived fragments were consistent with the sequencing results (Fig. 1B). We then collected serum from 28 trastuzumab-sensitive and 29 drug-resistant patients, extracted total RNA, and measured the expression levels of these tRNA-derived fragments using qRT-PCR. Combined with their clinical data, we identified that tRF-30-JZOYJE22RR33 and tRF-27-ZDXPHO53KSN were sensitive to predict trastuzumab resistance (Fig. 1C and 1D). Progression-free survival (PFS) analysis also showed that patients with high expression of these two tRNA-derived fragments had a worse PFS (Fig. 1E).
We then analyzed the sources of these tRNA-derived fragments. According to the data analysis of MINTbase (https://cm.jefferson.edu/MINTbase), we found that both tRF-30-JZOYJE22RR33 and tRF-27-ZDXPHO53KSN originated from the same digestion site of tRNA-cysgca, with only difference in three bases (Fig. 1F)26. At the same time, tRF-27-ZDXPHO53KSN had a stronger ability to predict drug resistance than tRF-30-JZOYJE22RR33; therefore, we chose tRF-27-ZDXPHO53KSN (hereinafter referred to as tRF-27) as our research object.
3.2 tRF-27 overexpression enhanced the resistance of HER2 positive breast cancer cells against trastuzumab
To further investigate the significance of tRF-27 in breast cancer drug resistance, we first compared its expression levels in trastuzumab-sensitive SKBR3 and trastuzumab-resistant JIMT1 cells by qRT-PCR (Fig. 2A). As suggested in previous studies, the concentration of trastuzumab added to the medium in our study was 2 µg/ml, which is similar to the clinical concentration4. We found that tRF-27 expression was significantly upregulated in JIMT1 cells stimulated with trastuzumab; in contrast, this level was almost unchanged in SKBR3 cells. We transfected JIMT1 and SKBR3 cells with the designed lentiviruses to construct tRF-27 overexpression and knockdown cell lines for subsequent studies. Trastuzumab mainly inhibits the growth of tumor cells by reshaping the immune environment33. We used the level of cell growth and proliferation in a trastuzumab-stimulating environment to reflect the level of tolerance of cells to trastuzumab34. The CCK-8 assay was used to detect cell growth and survival. Upon stimulation with trastuzumab, the growth rate of SKBR3 cells decreased significantly. The growth of JIMT1 cells was not changed evidently. This is consistent with our previous findings. Stimulated by the trastuzumab, we compared the growth rates of cells with overexpressed or knocked down tRF-27. tRF-27 overexpression increased, but tRF-27 knockdown decreased the growth rate of cells (Fig. 2B). Cell colony formation assay showed that the colony formation rate of cells was positively correlated with the expression level of tRF-27 (Fig. 2C). EdU assay showed that tRF-27 overexpression significantly increased, but tRF-27 knockdown inhibited the proliferation in both JIMT1 and SKBR3 cells (Fig. 2D). These results suggested that overexpression of tRF-27 could promote the growth of cells stimulated by trastuzumab, thus enhancing the tolerance of cells to trastuzumab.
We validated our hypothesis using tumor xenografts in nude mice. We injected 2*10^7 tumor cells subcutaneously into the left groin of BALB/c (nude) of each mouse. The mice were divided into four groups (four in each group), respectively treated with wild-type SKBR3 cells (control), SKBR3 cells overexpressing tRF-27 (tRF-27 OE), SKBR3 cells expressing blank control (tRF-27 NC), and SKBR3 cells with knocked down tRF-27 (tRF-27 IN). At day 7 after subcutaneous tumourisation, the latter three groups of mice were injected with trastuzumab. At day 28, the tumor disappeared in the tRF-27 NC group and the tRF-27 IN group (Fig. 2E). The growth of tumor in the remaining groups proven the role of tRF-27 in promoting drug resistance (Fig. 2F and 2G). Taken together, overexpression of tRF-27 could enhance trastuzumab resistance in breast cancer tumor cells.
3.3 tRF-27 bound to G3BPs and acted independent of stress granules
To determine the mechanism by which tRF-27 exerts its biological functions, we constructed biotin-labeled tRF-27 probes and scramble probes for pull-down assay35. SKBR3, JIMT1 and drug-sensitive BT474 cell lines were used for the experiment. Having been cultured with trastuzumab for 48 h, 2*10^8 cells of each type of cells were collected for pull-down assay (Fig. 3A and S1A). The obtained protein was silver-stained and detected using mass spectrometry36. The mass spectrometry results are appended to the Supplementary Materials (Table S2). The intersection in pull-down assays results of the tRF-27 group and the scramble RNA group was removed (Fig. 3B and S1B). We found that only one protein, G3BP2, was bound simultaneously in all three cell lines15. We found that G3BP2 and G2BP1 share 70% homologous sequences (Figure S1C)19; therefore, we tested their expression with Western blotting. In SKBR3 and JIMT1 cell lines, both G3BP1 and G3BP2 could bind to tRF-27 (Fig. 3C). In SKBR3 and JIMT1 cell lines, RNA immunoprecipitation (RIP) assay was performed with antibodies against G3BP1 and G3BP2. The RNA obtained was detected by qRT-PCR. The results confirmed the binding pattern (Fig. 3D).
G3BPs serve as key proteins in stress granules, and function in various environments17, 20. When stress granules form, G3BPs can competitively bind to regulatory associated protein of MTOR complex 1 (RPTOR), a key protein of the MTORC1 complex through sperm associated antigen 5 (SPAG5, also known as Astrin)37, thereby inhibiting the activation of MTORC1 by riveting the TSC complex on the surface of lysosomes without stress granules. Abnormal activation of the PI3K/AKT/mTOR pathway is engaged in trastuzumab resistance38. This information moved us to explore an association between G3BPs and trastuzumab resistance.
First, we clarified whether trastuzumab causes the formation of stress granules39. We cultured SKBR3 and JIMT1 cells in a trastuzumab-exposed environment and a glucose-deficient environment, considering that nutrient deficiency was a clear trigger for the formation of stress granules40. In immunofluorescence assay, stress granules formed in a nutrient-deficient environment, whereas the trastuzumab-treated group did not show nucleated stress granules (Fig. 3E and S2A). We collected SKBR3 cells stimulated by trastuzumab for co-IP assay, and found that SPAG5 did not bind to G3BPs (Fig. 3F). Western blotting was used to detect the expression levels of G3BPs in cells with tRF-27 overexpression or knockdown. Similarly, tRF-27 did not alter the expression levels of G3BPs in cells stimulated by trastuzumab (Fig. 3G)19. This indicated that stress granules did not form under the stimulation of trastuzumab. In addition, tRF-27 might function in a manner independent of stress granules.
We constructed a cell line for G3BP2 knockdown using siRNAs, and found that G3BP2 knockdown did not significantly increase G3BP1 expression (Fig. 3H). The cells with G3BP2 knockdown were subjected to another RNA pull-down assay, and we found that tRF-27 still bound to G2BP1(Fig. 3I). This suggested that tRF-27 could bind directly to the two proteins without relying on the dimeric relationship of G3BPs.
3.4 G3BPs were stably expressed in breast cancer cells, but G3BPs overexpression reduced trastuzumab tolerance
Although the regulatory mechanism of G3BPs in the MTORC1 signaling pathway has been well understood, evidence still lacks that G3BPs are involved in trastuzumab resistance. It has proven that the MTORC1 signaling pathway is repressed as the expression of G3BPs increases, thus inhibiting the growth and proliferation of cells20. The abnormal activation of the PI3K/AKT/mTOR signaling pathway is essential for trastuzumab resistance. Here, we obtained breast cancer-related data from the TCGA database and used the Genefu R package to classify all patients into 5 subtypes according to PAM50. Interestingly, we found that the expression levels of G3BP1 and G3BP2 did not vary across the five subtypes (Figure S2B). In the GSE76360 of the GEO database, by comparing the pre- and post-treatment sequencing data of 50 patients who received neoadjuvant trastuzumab-containing regimens, we found no significant changes in G3BP1 and G3BP2 expression before and after treatment (Figure S2C).
We constructed siRNA of G3BP2 and transfected it into SKBR3 cells. A plasmid overexpressing G3BP2 was constructed and transfected into SKBR3 cells. All experimental and control cells were cultured under trastuzumab exposure. Cell growth was detected by colony formation and EdU assays (Figure S3A and S3B). We found that G3BP2 knockdown promoted, but G3BP2 overexpression inhibited the growth and proliferation of cells.
Western blotting was used to detect protein expression in cells exposed to trastuzumab (Figure S3C). We found that G3BP2 knockdown promoted phosphorylation of Tuberin, but inhibited the phosphorylation of substrate eukaryotic translation initiation factor 4E binding protein 1 (4EBP1) of MTORC1. G3BP2 overexpression inhibited phosphorylation of Tuberin and promoted phosphorylation of 4EBP119. Taken together, G3BP2 could inhibit cell growth under trastuzumab exposure, suggesting that tRF-27 might affect the function but not the expression of G3BPs.
3.5 tRF-27 bound to G3BPs through a specific structure
Because stress granules were not formed, we suspected that G3BPs function through another pathway. G3BPs, as connexins, rivet lysosomal membrane protein LAMPs and TSC complex key protein Tuberin (also known as TSC2), thereby activating the TSC complex and inhibiting MTORC1. We transfected tRF-27 with 5'-FAM into SKBR3 and JIMT1 cells and immunofluorescently stained G3BP2 protein, finding that both tRF-27 and G3BP2 were predominantly located in the cytoplasm (Fig. 4A and S4A). First, we verified the bindings between these proteins. Proteins were extracted from cells cultured in a trastuzumab-stimulated environment and co-immunoprecipitated. G3BPs combined with LAMP1, RPTOR and Tuberin (Fig. 4B). This finding is consistent with the results of previous studies, indicating that G3BPs function through this pathway when stimulated by trastuzumab.
We further explored the binding regions between tRF-27 and G3BPs. According to the sequence of tRF-27, we designed three fragments through missing the front, back, and middle nine bases. Biotin-labeled probes were constructed (Fig. 4C). Drug-stimulated cells were collected for Pull-down assay (Figure S4B). The results showed that the ability of the probe to bind to G3BP2 was greatly reduced when the middle nine bases were missed, suggesting that tRF-27 relies on the middle nine bases to bind to G3BP2 (Fig. 4D). Online docking tools (http://hdock.phys.hust.edu.cn/) were used to predict the spatial structure of these four RNAs (Figure S4C)41. It was found that the complete tRF-27 had a loop, and this configuration was destroyed when tRF-27 lost its middle nine bases.
The protein structure of G3BP2 was obtained from UniPort (https://www.uniprot.org). We focused on 3 domains in G3BPs. Previous studies confirmed that G3BPs bind to LAMPs in the NTF2 domain. The RGG and RRM domains are considered as the main domains binding to RNAs (Fig. 4E). We found that only the NTF2 domain, as the spatial structure of G3BPs, was resolved, and Alphafold did not provide predictions (Figure S4D)42, 43. Based on this, we constructed plasmids with HA-TAG tags expressing G3BP2 and lacking these three domains44. Full-length and truncated G3BP2 plasmids were transfected into HEK293T cells, and Pull-down and co-IP assays were performed (Figure S5A and S5B). We found that tRF-27 did not bind to the G3BP2 which lacked the NTF2 region (Fig. 4F). Consistent with previous reports, LAMP1 was integrated into this domain (Fig. 4G). Binding between G3BP1 and G3BP2 was more likely to depend on the RGG domain. To confirm our results, we constructed truncated G3BP1 bodies (Fig. 4H and S5C). We found that tRF-27 could not bind to G3BP1, regardless of the absence of NTF2 or RRM (Fig. 4I).
This led us to speculate that tRF-27 might occupy the binding domain of G2BPs and LAMPs, thus affecting the binding of the two. To test this hypothesis, we synthesized tRF-27 and fragments of scrambled RNA and incubated two excess RNAs (10 nmol) with equal amounts of HEK-293 cell lysate (200 µl of cell lysate, 2*10^7 cells) for 2 h. The G3BP2 antibody-attached protein A/G magnetic beads were used to perform co-immunoprecipitation, and the protein that could bind to G3BP2 was detected by Western blotting (Fig. 5A). We found that G3BP2 binding to LAMP1 was significantly reduced in lysates co-incubated with tRF-27, compared to that in the scramble RNA group (Fig. 5B and 5C). At the same time, as a control, the binding of G3BP2 to G3BP1 did not change significantly. We repeated the experiment three times, measured and compared the band gray values (Fig. 5D), measured and compared the band quantitation values. The results showed that the tRF-27 group had a statistically significant decrease in quantitation, compared with the scramble group. We believed that tRF-27 might inhibit the ability of G3BPs to connect tubulin and lysosomes by occupying the binding site of G3BPs to LAMPs, thereby relieving the inhibition of TSC complexes on MTORC145. We used HDOCK (http://hdock.phys.hust.edu.cn/) to predict the binding regions of tRF-27 with NTF2 in G3BPs and demonstrated them with PyMOL (Fig. 5E).
Functional rescue experiments were also conducted. G3BP2-overexpressing plasmids were transfected into SKBR3 cells in a trastuzumab-exposed environment. Cell growth was detected by colony formation essay (Figure S5D). EdU assay showed that the proliferation of cells in the overexpression tRF-27 group was significantly inhibited, but not in the overexpression of G3BP2 which lacked NTF2 (Figure S5E).
3.6 tRF-27 promoted the activation of MTORC1
Consistent with previous reports, we found that trastuzumab inhibited the entire PI3K/AKT/mTOR signaling pathway in SKBR3 cells. Abnormal activation of this signaling pathway was observed in drug-resistant cells (Figure S6A and S6B).
We measured the expression of proteins involved the PI3K/AKT/mTOR signaling pathway in cells with tRF-27 overexpression and knockdown in a trastuzumab-stimulated environment (Fig. 6A and 6B)46. We found that tRF-27 did not affect the expression of proteins upstream MTORC1, but significantly promoted MTORC1 activation. This result was consistent with our previous assumptions.
We also treated cells with rapamycin, an inhibitor of MTORC1, in combination with trastuzumab, and found that the effect of tRF-27 in promoting cell proliferation was largely eliminated (Figure S6C). This result was consistent with our hypothesis.
We further used immunofluorescence to detect the colocalization of proteins. In SKBR3 cells, after tRF-27 was overexpressed, the colocalization of G3BP1 or Tubulin with LAMP1 was reduced (Fig. 6C and 6D). Knockdown of tRF-27 increased the colocalization of G3BP1 with LAMP1 and increased the colocalization of Tuberin with LAMP1. Overexpression of tRF-27 spatially separated G3BP1 and RPTOR (Fig. 6E), and inhibited Tuberin phosphorylation (Fig. 6F). This indicated that the TSC complex was activated by tRF-27 overexpression.
To test our hypothesis, we collected tumor samples from HER2-positive breast cancer patients who had received neoadjuvant trastuzumab-containing regimens, and divided the patients into trastuzumab-sensitive and trastuzumab-resistant groups based on whether their Miller-Payne scores were greater than 2. Tissue samples from two additional patients who did not receive neoadjuvant therapy were used as controls. Immunohistochemical analysis showed that trastuzumab treatment did not affect G3BP1 expression, but trastuzumab-sensitive patients showed stronger expression of p-tuberin and weaker expression of p-4EBP1 (Fig. 7A and 7B).
In addition, we detected the expression of these proteins in mouse tumor samples (Fig. 8A). Immunohistochemical analysis showed that overexpression and knockdown of tRF-27 achieved consistent results observed in the cells (Fig. 8B).
In summary, stimulated by trastuzumab, G3BPs attached to TSC complexes to LAMPs on the surface of lysosomes, thereby inhibiting MTORC1 activation. tRF-27 overexpression blocked the lysosomal localization of TSC complexes by competitively binding with G3BPs to LAMPs. As the inhibition on MTORC1 was relieved, the growth of trastuzumab-stimulated cells was promoted, thus enhancing their resistance against trastuzumab (Fig. 8C).