Anti-HDGF Antibody Targets EGFR Tyrosine Kinase Inhibitor–Tolerant Cells in NSCLC Patient-Derived Xenografts

Lung cancer remains a major disease and is the leading cause of cancer-related death worldwide (1). In non–small cell lung cancer (NSCLC), the major form of lung cancer, the activating mutation of EGFR is an important oncogenic driver. The incidence rate of NSCLC with EGFR mutation ranges from 10% to 15% in patients of European origin and 30% to 60% in East Asian patient populations (2). Patients with lung cancer with EGFR mutation are often nonsmokers, younger than the median age of average patients with lung cancer, and a higher percentage of them are women.

EGFR is a member of the ERBB family receptor tyrosine kinase vital for the survival, growth, migration, and differentiation of epithelial cells (3). Structurally, EGFR consists of an amino-terminal extracellular ligand-binding domain, linked to a transmembrane region, followed by the kinase domain and a carboxyl-terminal-regulatory domain. Engagement of EGFR ligands, such as EGF or amphiregulin, induces receptor dimerization, followed by autophosphorylation of several tyrosine residues in the regulatory domain, leading to the release of inhibition on the kinase and activation of EGFR. The activated kinase in turn recruits and activates several downstream pathways, including Ras-MAPK, PI3K/AKT/mTOR, that are major drivers of survival and proliferation (4). Dysregulated EGFR activation in the absence of proper physiologic signal plays a crucial role in oncogenesis and in maintaining the malignancy of lung cancer and therefore an ideal therapeutic target (5).

Gain-of-function mutation is a common mechanism that can result in abnormal EGFR activation. Among the more than 30 activating mutations in EGFR kinase domain identified in treatment-naïve tumors, exon 19 in-frame deletion and exon 21 L858R mutation account for about 90% of the incidences. Exon 20 in-frame deletion and exon 19 mutations account for about 10% (6). EGFR with exon 19 deletion, L858R, and G719X mutations can be inhibited by several classes of EGFR tyrosine kinase inhibitors (TKI). They are often referred to as the classical sensitizing EGFR mutation. Inhibition of their activity in NSCLC cells causes growth arrest and cell death, typically at nanomolar concentrations under cell culture conditions. Clinically approved EGFR inhibitors for NSCLC treatment include the first-generation competitive inhibitors, such as erlotinib and gefitinib, the second-generation irreversible pan-ERBB family inhibitors, such as afatinib and dacomitinib, and the third-generation mutation-specific inhibitors, such as osimertinib and lazertinib. EGFR-targeted therapy has achieved great clinical success. In patients with NSCLC with metastatic or recurring diseases harboring sensitizing EGFR mutation, targeted therapy using these small-molecule TKIs resulted in significantly increased progression-free survival (PFS), compared with platinum doublet chemotherapy (10–19 vs. 5.4 months), and extension of overall survival (OS; 31–38 vs. 26 months; refs. 7–12). It has become one of the preferred first-line choices for these patients.

However, despite the high response rate (∼80%) and significantly extended PFS, few patients have complete responses. In most patients, the disease will eventually progress as clones with acquired drug resistance emerge. Extensive studies have identified a complex mechanism of resistance that causes treatment failure (13, 14). About 50% of the resistance to the first or second-generation TKIs occurs because of secondary T790M mutation in EGFR. Other mechanisms that confer resistance include amplification of EGFR or activation of other receptor tyrosine kinases, such as ERBB2 and MET, mutation of EGFR downstream signaling effectors, and alteration in cell differentiation status, such as small cell lung cancer transformation, epithelial-to-mesenchymal transition, or acquisition of stem cell–like state (15). Mutation in C797S or loss of T790M can result in resistance to osimertinib in addition to other target-independent alterations (16, 17). Many of these secondary alterations bestow acquired resistance by reactivating the major prosurvival/proliferative signaling pathways.

In contrast to the mechanism of resistance, the mechanism that drives the tumor to transition from drug-sensitive to drug-resistant is less understood. In most patients, this transition probably happens during the period of PFS that lasts for months. In cultured cells harboring a sensitizing EGFR mutation, most of the cells are eliminated by TKI treatment at physiologically achievable plasma concentration within a few days. Long-term exposure to escalating concentrations of TKI is typically used to induce acquired resistant clones (18–23). Many of these in vitro established resistant clones have MET amplification, demonstrating MET signaling as an effective mechanism to bypass EGFR blockage. Indeed, cell lines with high-level MET expression, such as Hcc827 or PC-9, are the easiest to induce TKI-resistant clones in cell culture by MET gene amplification. Clinically, about 5% to 20% of the tumors in patients relapsed on TKI treatment have MET amplification (24–28). This suggests other mechanisms probably exist in vivo to promote the survival of tumor cells during the PFS in which steady-state plasma concentration of TKI can be achieved in 1 to 2 weeks.

Hepatoma-derived growth factor (HDGF) is a heparin-binding protein first isolated from the hepatoma cell line Hu-7 conditioned medium base on its ability to stimulate Swiss 3T3 cell proliferation (29). It is distinct from hepatocyte growth factor, the ligand of MET, and there is no report of direct interaction of HDGF with MET. In several tumor cell lines, HDGF has been shown to activate PI3K/Akt signaling through interaction with cell surface receptor (30, 31) or activate MAPK/ERK signaling by enhancing KRAS expression (32). Dysregulated overexpression of HDGF has been identified in multiple solid tumors and is associated with inferior clinic outcome of patients with cancer (33). We have shown in our prior studies that there is a strong association of HDGF overexpression with poor performance in patients with NSCLC. Reducing the expression of HDGF in NSCLC cells by siRNA suppresses the tumorigenicity and malignancy of NSCLC cells (34, 35). Furthermore, anti-HDGF antibody could enhance the chemotherapeutic treatment of NSCLC tumor xenografts (36, 37).

In this study, we investigated the mechanism underlying incomplete response to TKI in EGFR-targeted therapy and the role of HDGF in this process in PDX models of NSCLC harboring EGFR mutation.

Osimertinib methanesulfonate (>99%) was obtained from LC Laboratories. The purification of HDGF and generation of antibody were described previously (36). To produce humanized anti-HDGF antibody, cDNA encoding the variable domain of the murine antibody was amplified and then sequenced, and the complementary determinant regions were grafted onto a human IgG1 framework, cloned into a mammalian expression vector, and then expressed in the Expi293 transient expression system (Invitrogen). Recombinant anti-HDGF antibody H3 was purified by standard affinity chromatography, and endotoxin was removed to <0.02 EU/mg. The binding affinity of the humanized antibody to HDGF was determined by surface plasmon resonance on Biacore 3000 using GST-HDGF as the analyte.

The DNA sequence encoding GST-HDGF fusion proteins and a series of C-terminal truncations were cloned into pBiEx-1 (Novagen), expressed in BL21(DE3) cells, fractionated by SDS-PAGE, and blotted and probed with recombinant anti-HDGF antibody H3.

HDGF single-guide RNA CRISPR/Cas9-targeting constructs (Applied Biological Materials Inc., Cat#23141111) were packaged in lentivirus and used to transduce HEK293 cells. The sequences targeted were targets 1 to 3: GTC​GCG​ATC​CAA​CCG​GCA​GA; targets 2 to 159: TGG​CCT​CCA​CTC​ACG​TCT​CG; and targets 3 to 195: TCT​CCT​TGG​ATT​CCT​CGT​AA. The transduced cells were selected using puromycin to generate polyclonal (pooled) cell lines.

PDX models of NSCLC, TM00219 and TM00199 (Table 1), were obtained from The Jackson Laboratory and propagated in immunodeficient NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ mice (38). The initial engraftment by the vendor, designated as P0, was harvested and cryopreserved. Low-passage tumors (P2 to P4) were used to prepare treatment cohorts. Briefly, a fresh tumor was cut into ∼2-mm pieces and implanted into the flank of the recipient animal. Female mice were used for both PDX models to match the sex with the human donors. When the tumor volume reached 250 to 300 mm³, the mice were randomized into two arms to receive osimertinib (10 mg/kg per os for 5 days a week) or osimertinib plus H3 (13 mg/kg intraperitoneally every 3 days). Tumor sizes were measured using a caliper twice a week. The approximate volume was calculated using the formula V = L × 2W × H. The use of animals in this study was approved by the UMB Institutional Animal Care and Use Committee.

Formalin-fixed, paraffin-embedded tumor sections were prepared and stained as described (34). Briefly, duplicate sections were stained with diluted antibody at 4°C overnight and then visualized using Elite ABC kit (VectorLabs) and hematoxylin counterstain. The mounted slides were scanned using Olympus VS120 or Aperio CS2 Pathology Slides Scanner at 40× resolution without adjusting or editing. The staining intensity was semiquantitatively evaluated by two team members and a trained pathologist (AS), in which each tumor section was scored negative or weak (i), moderate (ii), and strong (iii).

Soluble proteins were extracted from frozen tumor using PBS buffer containing 1% Triton X-100 supplemented with protease and phosphatase inhibitors, then separated on 4% to 12% SDS-PAGE gradient gel, blotted to the nitrocellulose membrane, and probed sequentially with 5 to 7 antibodies. The membrane was stripped between staining. Antibody reactivity was detected electrochemiluminescently on film, scanned using CanoScan 9000F, and quantified using ImageJ. Images were not adjusted or edited manually.

Genomic DNA was purified from pellet after tumor protein extraction with proteinase K digestion and silica membrane binding. A 1,320-bp fragment in EGFR gene flanking exon 20 was amplified using high-fidelity PCR and primers in introns 19 to 20 (5′-CAC​AGC​ACA​GAG​AGA​CCA​CT-3′) and introns 20 to 21 (5′-CAA​GGT​AAG​CAA​GCC​AGG​CC-3′) and purified and sequenced using a primer in exon 20 (5′-GAA​GCC​TAC​GTG​ATG​GCC​A-3′).

Antibodies to the following were obtained from Cell Signaling Technology: phospho-EGFR (Tyr1068; #3777, RRID: AB_2096270 and #2234, RRID: AB_331701), phospho-MEK1/2 (Ser217/221; #9121, RRID: AB_331648), MEK1/2 (#9122, RRID: AB_823567), phospho-p44/42 MAPK (Erk1/2; Thr202/Tyr204; #9101, RRID: AB_331646 and #4307, RRID: AB_2315112), p44/42 MAPK (Erk1/2; #9102, RRID: AB_330744), phospho-Akt (Ser473; #4060, RRID: AB_2315049 and #9271, RRID: AB_329825), Akt (#9272, RRID: AB_329827), phospho-PRAS40 (Thr246; #2997, RRID: AB_2258110), PRAS40 (#2610, RRID: AB_916206), phospho-p70 S6 kinase (Thr389; #9234, RRID: AB_2269803), p70 S6 kinase (#2708, RRID: AB_390722), phospho-4EBP1 (Thr37/46; #2855, RRID: AB_560835), 4EBP1 (#9644, RRID: AB_2097841), Met (#8198, RRID: AB_10858224), and phospho-Met (Tyr1234/1235; #3077, RRID: AB_2143884). Anti-EGFR was obtained from Santa Cruz Biotechnology (Cat#sc-03, RRID: AB_631420). Anti–β-actin was obtained from Sigma (Cat#A1978, RRID: AB_476692). Anti–HDGF T221 was used in previous studies (34, 39).

HEK293 (CRL-1573) and Hcc827 (CRL-2868) were obtained from ATCC. Expi293F was part of Invitrogen Cat#A14635. The cell lines were expanded upon arrival, and multiple aliquots of low-passage cells (P2–P3) were cryopreserved. Upon recovery, the cells were cultured in media containing tiamulin fumarate, followed by minocycline (Roche, BM-Cyclin). All cells were tested negative for Mycoplasma contamination monthly using the MycoFluor Kit (Invitrogen). Cells within passage 6 were used for all experiments.

Differences in PFS were analyzed by Kaplan–Meier survival curves using SRplot (http://www.bioinformatics.com.cn/srplot). Significance was tested using the log-rank test. Quantitative data of IHC and Western blotting analysis were analyzed using the t test (one-tailed homoscedastic).

The data generated in this study are available upon request from the corresponding author.

The recombinant humanized anti-HDGF antibody H3 binds HDGF in its native form, as shown in its ability to capture HDGF from HEK293 cell lysate (Fig. 1A). The captured HDGF migrated on SDS-PAGE with an apparent molecular weight of about 37 kDa (major) and 42 kDa (minor). The specificity of the H3–HDGF interaction is further validated by co-reduction in the intensity of H3-reactive bands in Western blot of HEK293 cells with CRISPR-Cas9–mediated HDGF knockout (Fig. 1B). The H3–HDGF interaction is strong with a dissociation constant (Kd) of 6.14 × 10^–9 mol/L (6.14 nmol/L) when assayed by surface plasmon resonance using GST-HDGF as the analyte (Fig. 1C). By probing a set of GST-HDGF fusion proteins with sequential C-terminal deletion, the epitope recognized by H3 was mapped to between amino acid residue 160 and 170 in a region predominated by an unfolded sequence (Fig. 1D). This stretch of sequences is highly homologous between the human and mouse (Fig. 1E).

The response of EGFRm+ NSCLC tumor to the third-generation EGFR TKI osimertinib was evaluated in established PDX tumors. Continuous osimertinib dosing of mice bearing TM00219 (250–329 mm³) or TM00199 (228–410 mm³) tumors induced rapid tumor regression (Fig. 2A and B). However, the response was partial with the best percent change from baseline (BCB) ranging from 53% to 72% in TM00219 to 43% to 70% in TM00199 (Fig. 2C and D). In both models, tumor growth resumed after about 24 to 31 days (median, 24 days) for TM00219 or 17 to 38 days (median, 29 days) for TM00199, even with continued osimertinib dosing. Similarly, xenograft tumors derived from Hcc827, an EGFR-mutant cell line sensitive to erlotinib in vitro, also had incomplete response to erlotinib in vivo (Supplementary Fig. S1).

Next, we examined whether anti-HDGF antibody can alter the tumor response to osimertinib. In established TM00219 tumors (257–427 mm³), the humanized anti-HDGF antibody H3 was given at 13 mg/kg concurrently with the start of osimertinib dosing and subsequently maintained at twice per week till tumor relapse or up to 200 days. The combination treatment also induces rapid tumor regression (Fig. 2A) at a pace similar to or slightly faster than that of osimertinib monotherapy as shown in the mean percent change from baseline in tumor volume (Fig. 2E). More importantly, tumor regression proceeded to near completion with the BCB reaching 96% to 99.8% (Fig. 2C). Similarly, in established TM00199 tumors (229–410 mm³), osimertinib plus H3 combination also induced rapid tumor regression with BCB ranging from 91% to 99% (Fig. 2B and D) at a pace of tumor regression similar to that of osimertinib monotherapy (Fig. 2F). In both models, there were significant extensions of PFS ranging from 114 to 220 days (median, 128 days) for TM00219 and 98 to 112 days (median, 104 days) for TM00199. The difference in the probability of PFS in the two treatment arms is significant (P = 0.0004; Fig. 2G and H). However, H3 alone has no effect on the growth of osimertinib-naïve PDX tumors (Supplementary Fig. S2A and S2B) or tumor progressed on osimertinib monotherapy (Supplementary Fig. S3A and S3B).

To identify potential mechanisms leading to the limited response in osimertinib monotherapy, we examined the early changes in the major pro-proliferative signaling pathways downstream of EGFR in serially collected TM00219 tumors after initiation of osimertinib treatment by IHC and Western blotting. After oral osimertinib administration, significant suppression of EGFR Y1068 phosphorylation was observed at 6 hours in IHC-stained tumor tissues (Fig. 3A). In the downstream Akt/mTOR and MAPK pathways, phosphorylation of Erk1/2, Akt, and PRAS40 also showed significant suppression by 16 hours in IHC staining. However, 24 to 30 hours after initiation of osimertinib treatment, there was an apparent increase in phosphorylation of these proteins in IHC-stained tumor samples (Fig. 3B–D). In several tumors, positive phosphoprotein staining was most prominent in small loci. However, at the same time, phosphorylation of EGFR Y1068 remains low (Fig. 3A). Western blot examination of tumor protein extracts from the corresponding serially collected samples confirmed suppression of EGFR Y1068 phosphorylation starting from 6 hours after osimertinib dosing and was undetectable by 72 hours (Fig. 3E). The phosphorylation of MEK1/2 and Erk1/2 of the Ras/MAPK pathway and Akt, PRAS40, and p70S6K of the Akt/mTOR pathway began declining at 6-hour post-osimertinib dosing, remaining low at 16 hours. However, there was a significant increase in phosphorylation of these proteins at 24 hours, in line with the increase in phosphorylation observed in IHC-stained tumor tissues (Fig. 3E). During this period, the total level of these proteins likely remains unchanged as exemplified by EGFR, Akt, and 4EBP1.

We next examined the effect of anti-HDGF antibody on these signaling pathways in tumors treated with osimertinib for 10 to 13 days. At this point of treatment, there was approximately 30% tumor size reduction. In IHC staining of phosphorylated Erk1/2, Akt, PRAS40, and 4EBP1, there were regions of strong positive staining in tumors from the osimertinib monotherapy arm. In contrast, the staining intensities of these phosphorylated proteins in tumors from the osimertinib plus anti-HDGF antibody combination arm were typically much weaker (Fig. 4A–D). The difference in the staining intensity of P-Erk1/2, P-Akt, and P-PRAS40 between the two arms was significant as shown in the Cleveland dot graph.

We further examined the level of phosphorylation in the Akt/TOR and MAPK pathways in protein extracted from these tumors by Western blotting (Fig. 4E). Stronger staining of phosphorylated MEK1/2, Erk1/2, p70S6K, and 4EBP1 was observed in the osimertinib monotherapy arm comparing with the osimertinib plus H3 combination arm, whereas changes in Akt and PRAS40 were not significant. It is interesting to note that the level of EGFR phosphorylation in both treatment arms was increased compared with the P-EGFR level at 72 hours (Fig. 3E). However, the level of P-EGFR in a particular tumor did not always correlate well with the level of phosphorylation in the downstream pathways. In addition, there was a noticeable reduction in the total 4EBP1 level in both treatment arms compared with naïve samples.

To evaluate the contribution of MET activation in the incomplete response to osimertinib of the PDX tumors, we examined the level of MET and P-MET expression in TM00219 in comparison with Hcc827 cells. In medium containing 10% FBS, Hcc827 cells showed a high level of MET and P-MET expression that can be abolished by treatment with crizotinib. However, the TM00219 PDX tumors displayed significantly lower levels of MET expression and negligible levels of P-MET expression in naïve tumor and in arms treated with osimertinib or osimertinib plus H3 for 13 days. More importantly, the level of MET expression in the PDX tumor did not correlate with the drug treatment regimen (Supplementary Fig. S4A).

We next examined if the C797X secondary mutation is involved in the incomplete response to osimertinib of these PDX tumors. Genomic DNA fragments flanking EGFR gene exon 20 were amplified from naïve tumors and tumors treated with osimertinib or osimertinib plus H3 for 13 days and then sequenced. The sequencing chromatograms of the two treated arms were similar to naïve tumors, and no indication of C797X mutation was found (Supplementary Fig. S4B).

Next, the relationship of HDGF expression and TKI tolerance was examined in tumors from different treatment arms. In IHC of naïve tumor, HDGF staining was seen in the nucleus of most of the tumor cells with intensity ranging from weak to medium to strong. Weak cytoplasmic staining of HDGF was also observed in naïve tumor cells (Fig. 5A and B). Interestingly, in both osimertinib monotherapy and combination therapy arms, all surviving tumor cells displayed nearly uniform medium strong to strong nuclear staining and negligible cytoplasmic staining (Fig. 5A and C). However, we did not observe significant differences in the mean intensity between osimertinib monotherapy and osimertinib plus H3 combination arms despite some variations within the treatment arms. In addition, we did not observe significant changes in HDGF staining intensity in Western blot analysis of tumor proteins extracted from different treatment arms (Fig. 4E).

EGFR TKIs are the preferred treatment options for patients with advanced NSCLC harboring EGFR mutations. However, despite the 10 to 19 months of PFS achieved in many patients, most lesions eventually progress after acquiring resistance to TKI. Furthermore, less than 10% of the patients treated with EGFR TKI achieved a complete response. Most patients experience a partial response with 20% to 80% tumor regression.

In contrast, cultured EGFRm+ NSCLC cells lacking concurrent bypassing mutation, such as Hcc827, can be effectively killed within days by TKI at a fraction of the clinically achievable plasma drug concentration, typical submicromolar to nanomolar. However, xenograft tumors derived from Hcc827 showed partial response to TKI (Supplementary Fig. S1). This suggests that in vivo environment promotes tumor survival when exposed to TKI. The surviving tumor cells, often referred to as “drug-tolerant persister,” exhibit transient and reversible drug tolerance. Conceivably, the ability of tumor cells to rapidly adapt to the drug-tolerant state upon initiation of TKI administration is crucial for the survival of tumor cells during the early stage of EGFR-targeted therapy. Subsequent evolution and acquisition of resistant drivers within the drug-tolerant cells ultimately give rise to bona fide resistant clones, leading to disease progression. The incomplete response and the duration of PFS in many patients undergoing TKI therapy suggest that this drug-tolerance phase may exist prior to overt disease progression.

Recently, there was heightened interest in understanding the mechanisms of drug tolerance (40–42). Activation of RTK bypassing signaling pathways is frequently observed in drug-tolerant persisters due to the highly convergent nature of RTK signaling, such as in other ERBB family of kinases, MET and AXL (43). Co-targeting these kinases is a rational approach to break tumor cell tolerance, either by using inhibitors with a wilder specificity (e.g., afatinib) or a combination of inhibitors (e.g., EGFR TKI plus crizotinib). However, the success of these strategies varies. Afatinib, a pan-ERBB inhibitor, exhibits a similar efficacy profile to the first-generation TKIs (44). Dacomitinib, another pan-ERBB inhibitor, had shown extended PFS and overall survival (45), whereas amivantamab, a bispecific antibody targeting EGFR and MET, has shown improved efficacy in combination with lazertinib as a first-line treatment of EGFRm+ NSCLC (46). These results underscore that complex mechanisms may exist to promote tumor cell survival and drive the transition from TKI sensitive to resistant.

In this study, we investigated the role of HDGF in promoting tumor tolerance to osimertinib using two PDX tumor models of NSCLC: TM00199 and TM00219. These models were derived from patients who had progressed on single-agent or combination erlotinib treatment (38).

We observed that the PDX tumors are susceptible to EGFR inhibition. Oral osimertinib administration rapidly and profoundly inhibited phosphorylation at EGFR Y1068 and downstream AKT/MTOR and MAPK pathways. These pathways play a critical role in cell survival and are often associated with secondary mutations, leading to resistance against various anticancer therapies. The swift inhibition of these pathways upon osimertinib administration explained the rapid tumor shrinkage.

However, the response of the PDX tumors to osimertinib was incomplete, with maximum regression ranging from 50% to 70% that achieved around 30 days after treatment initiation, followed by progression. Notably, even at the early stage of treatment with approximately 30% tumor shrinkage, significant recovery of phosphorylation in AKT/MTOR and MAPK pathways was observed. Serially collected samples revealed that near-maximum inhibition of EGFR Y1068 and downstream phosphorylation occurred as early as 6 hours, but subsequent recovery of phosphorylation could be detected within 24 hours. This rapid transition from initial suppression to partial recovery suggests that factors other than drug penetration or novel mutations likely contribute to this phenomenon. Reactivation of EGFR downstream pathways was also observed in other studies using patient-derived tumors (38, 47).

Surprisingly, concurrent administration of an anti-HDGF antibody at the start of osimertinib treatment led to tumor regression exceeding 98%. On average, the progression-free time was four to five times longer than osimertinib monotherapy. This suggests that most persistent tumor cells prior to progression were just transiently tolerant to TKI, as opposing to stable resistance. Their tolerance can be overcome by blocking HDGF.

Analysis of tumors harvested ∼10 days after treatment initiation revealed significantly lower levels of activating phosphorylation in AKT/MTOR and MAPK pathways in the combination arms compared with osimertinib monotherapy. This suggests that HDGF participates in reactivating the prosurvival pathways. The anti-HDGF antibody attenuated this reactivation during osimertinib treatment, enhancing overall efficacy. However, when administrated alone, anti-HDGF antibody did not affect the growth of osimertinib-naïve or post-progression tumors, suggesting that HDGF signaling is relatively weak compared with the dominant oncogenic driver.

HDGF, identified as a mitogen for several cell types (39, 48), contains two nuclear localization signals. In most cell types, HDGF is predominantly localized to the nucleus, where it likely functions as a transcription factor or participates in RNA biogenesis. In this study, negative to strong nuclear staining and weak cytoplasmic staining were observed in treatment-naïve tumor cells. In osimertinib-treated tumors, however, all the surviving tumor cells showed strong nuclear staining, suggesting an increase in the level of HDGF expression. Corroborating with our study, biopsy samples of patients with NSCLC treated with EGFR TKI demonstrated a considerable increase in HDGF staining intensity after relapse (49). Furthermore, HDGF expression in cultured cells with EGFR mutation and in xenograft tumors derived from them significantly increased after gefitinib or osimertinib treatment (49). These findings suggest that higher levels of HDGF expression are associated with increased TKI tolerance. As HDGF could promote the expansion of hematopoietic or cancer stem cells, it is possible that tumor cells with higher HDGF expression exhibit more stem cell–like features and are thus more tolerant to TKI. This is particularly interesting considering that chemotherapy could extend the benefit of osimertinib (50), and stem-like cancer cells tend to be more resistant to chemotherapy (51). Targeting HDGF could potentially suppress resistance to chemotherapy too and further enhance the benefit of osimertinib plus chemotherapy. In a previous study, the administration of an anti-HDGF antibody resulted in reduced stem cell features in tumors and enhanced the efficacy of combination chemotherapy in NSCLC tumors (37).

As a protein without a classical secretion signal, HDGF is readily detected in media of cultured NSCLC cells (36). The medium level of HDGF in abnormal fibroblast culture is also elevated (52). It is not entirely clear how HDGF enters the extracellular media. One study suggests that HDGF N-terminal sequence is involved in secretion (53). Stress or apoptosis of cells could increase the level of HDGF released into the media, a condition that is relevant to cancer therapy (54). In patients with NSCLC treated with gefitinib or osimertinib, there were statistically significant increases in plasma concentrations of HDGF (49). Exogenous HDGF could bind to cell surface–displayed nucleolin and activate MAPK/Erk and PI3K/AKT pathways (30). In cultured NSCLC cells or xenograft tumor, the expression of HDGF is positively correlated with the activation of these pathways and the resistance to EGFR TKI; knocking down HDGF diminishes the activation of these pathways and sensitizes the cells to TKI inhibition (49).

Besides the malignant cells, the tumor microenvironment also contains various nonmalignant cells, including vasculature cells, mesenchymal-derived cells such as fibroblast, and immune cells in immune-competent animals and human patients. These cells interact with each other through secreted growth factors, cytokines, or surface-displayed ligands and their receptors. HDGF released by tumor cells could signal to the nonmalignant cells in the tumor microenvironment to promote the survival and malignancy of tumor cells.

Although HDGF-dependent osimertinib-tolerant cells constitute 20% to 30% of the mass in tumors of the osimertinib arm prior to progression, analysis of TM00219 tumors collected approximately 13 days after treatment initiation revealed a low level of MET expression compared with Hcc827 and an insignificant level of P-MET expression. Notably, there were no meaningful differences in the expression of MET and P-MET between osimertinib-naïve and -treated tumors. Furthermore, no secondary mutations at C797 were detected after sequencing EGFR exon 20 in these tumors. These data suggest that MET amplification/activation and EGFR C797X secondary mutation, which are commonly associated with acquired resistance to osimertinib in post-progression tumor in human patients, were unlikely the primary mechanism of osimertinib tolerance at the early stage of osimertinib treatment.

Taken together, our data demonstrate that during the early stage of EGFR-targeted therapy, tumor cells can rapidly develop tolerance to osimertinib without acquiring new resistant mutations or activating MET. This process likely involves both tumor cells and nonmalignant cells in the tumor microenvironment (Fig. 6). Subsequent emergence of tumor cells harboring new drug-resistant alterations from the drug-tolerant cells ultimately leads to disease progression while continuing osimertinib treatment. The drug-tolerant cells serve as a vital link in the transition of tumor from a sensitive state to a resistant one. In this intricate process, HDGF plays a crucial role by inducing tumor tolerance, partly through the activation of MAPK/ERK and PI3K/AKT pathways. Blocking HDGF promotes tumor regression and significantly extends survival in tumor-bearing mice.

The current study also showed the complexity of tumor cell signaling under TKI inhibition. For instance, in the first 72 hours of osimertinib treatment, there was near-complete inhibition of EGFR Y1068 phosphorylation. However, rephosphorylation at Y1068 was observed on 10 to 13 days in tumors that had about 30% tumor shrinkage. Surprisingly, this increase in Y1068 phosphorylation did not consistently correlate with downstream activation in the MAPK/ERK and PI3K/AKT pathways. This phenomenon may be explained by cross-talk between EGFR and other kinases. Furthermore, anti-HDGF antibody could not fully prevent the emergence of osimertinib resistance, suggesting that additional bypass pathways may coexist.

Our study demonstrates that tumor tolerance to TKIs could be a crucial mechanism contributing to incomplete responses during the early phase of EGFR-targeted therapy. HDGF plays a critical role in this process. Concurrent inhibition of HDGF with the initiation of TKI treatment can re-sensitize drug-tolerant cells to TKI, resulting in enhanced tumor regression and delayed disease progression. These findings suggest that targeting drug-tolerant cells early in the treatment process could be a promising strategy for improving EGFR-targeted therapy in NSCLC. Nevertheless, further advanced investigations are necessary to fully elucidate the HDGF-mediated signaling pathway and determine the clinical utility of anti-HDGF antibodies in NSCLC treatment.

H. Ren reports a pending patent application with the USPTO. No disclosures were reported by the other authors.

C.Q. Zhou: Investigation. A. Li: Formal analysis, investigation, writing-review and editing. K. Ri: Investigation. A. S. Sultan: Formal analysis, investigation, writing-review and editing. H. Ren: Conceptualization, data curation, formal analysis, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.

This work is supported in part by the Department of Oncology and Diagnostic Sciences, University of Maryland School of Dentistry (H. Ren); Maryland Technology Development Corporation; Maryland Innovation Initiative Project #0517-011 (H. Ren); and University of Maryland Baltimore Institute for Clinical and Translational Research grant #1UL1TR003098 (H. Ren). A. Li is supported in part by National Institute of Dental and Craniofacial Research grant R25032530.