Intermittent afatinib treatment suppresses the growth of resistant T790M-H1975 cells in non-small cell lung cancer (NSCLC) co-culture

Introduction

The discovery of tyrosine kinase inhibitors (TKIs) that selectively target activating mutations in the protein kinase epidermal growth factor receptor (EGFR) led to a breakthrough in the treatment of non-small cell lung cancer (NSCLC) [1], [2], [3], [4]. EGFR-TKIs target common activating mutations, such as the exon 19 deletion (E746_A750del) and L858R, which show clinical benefit in the majority of patients with EGFR-mutated NSCLC when administered continuously [5], 6]. However, resistance invariably develops, with the emergence of de novo (pre-existing) T790M resistance reported in approximately half of the treated patients [7]. Previously, in NSCLC, the variant allelic frequency (VAF) of de novo T790M was almost absent in diagnosed patients with predominant common activating mutations [8], [9], [10]. This notion coincides with older detection technologies such as polymerase chain reaction (PCR) clamp, scorpion amplified refractory mutation system methods, and direct sequencing [11], 12]. However, as detection technology advances over time, the prevalence of de novo T790M mutation increases with the aid of better sensitivity in detection methods, which include matrix-assisted laser desorption ionisation-time of flight mass spectrometry (MALDI-TOF MS) [7], 13], TaqMan assay [14], 15], colony hybridisation [16], and peptide nucleic acid (PNA)-clamping PCR [17], 18]. Digital [19], droplet digital PCR, and next-generation sequencing [20], 21] were able to have an analytical sensitivity of approximately 0.001 %, which detected 66 % of the overall incidence of ultra-low de novo T790M [22], 23].

An emerging body of evidence has suggested that intermittent dosing schedules with elongated treatment-free periods might have advantages in delaying the emergence of resistance over continuous treatment [24], [25], [26]. Preclinical studies with H3255 lung cancer xenografts showed that intermittent weekly gefitinib dosing showed better therapeutic outcomes than continuous dosing [27]. Multiple case reports have detailed how allowing an elongated treatment-free period in a dosing regimen prolongs progression-free survival with reduced side effects in erlotinib [28] and afatinib [29], 30].

The success of intermittent treatment is supported by a cell-cell competition treatment regimen called “adaptive therapy.” This concept emphasises a significant proportion of viable treatment-sensitive cells to allow cellular competition with de novo resistant cells for nutrients and space during the treatment-free period [31]. The application of this treatment regimen is aimed at being patient-specific, as tumour growth and heterogeneity vary between individuals. Thus, it is one of the biggest hurdles to having it in the clinical setting. Multiple reports of the successes of adaptive therapy come from mathematical modelling and simulations to predict the treatment outcome [32], 33]. Under adaptive therapy, the same tumour could be managed with an average of just 35 % of the maximum-tolerated dose and a total dose that was 70 % of that given with continuous treatment [34]. In addition, the adaptive therapy cycling demonstrated that control may be durably maintained for up to 20 cycles, which is significantly longer than continuous treatment [35]. Although it is understood that intermittent treatment is shown to be therapeutically inferior to continuous treatment, it is unknown whether optimising intermittent treatment with adaptive therapy principles can improve the treatment regimen.

In the present study, we used an in vitro strategy to examine the effect of afatinib with a treatment-free period in intermittent treatment to spare a significant proportion of sensitive cells and delay de novo resistant cell lines.

Materials and methods

Response of the co-culture of sensitive and resistant cells to afatinib

The experimental design in co-culture containing 0.1 and 0.5 % resistance was carried out with continuous therapy of 100 and 500 pM afatinib up to 96 and 192 h to represent 2 treatment cycles. Meanwhile, in intermittent therapy, co-cultures were exposed to concentrations of 100 and 500 pM afatinib for 24 h and replaced with treatment-free media for 72 h. The process was repeated after 96 h of design, as shown in Figure 1. Following that, the pharmacokinetic calculation from the area under the curve (AUC) to experimental concentrations was done with Equation (1). The AUC has been extrapolated to 96 h via PharmaCalc (Zurich, Switzerland).

(1) AUC ng / mL = concentration   nmol / L × 1000 molecular   weight   g / mol

(2) Doubling time  hours = Hours × ln 2 ln Final   viable   cell   number Initial   viable   cell   number

Figure 1:

Illustration representation of the experimental setup for continuous and intermittent treatments.

Results

In the absence of afatinib, both cell lines were evaluated for their proliferation rate (Figure 2), followed by the calculated doubling time listed as Equation (2). In Table 1, the mean doubling time of the HCC827 was longer than that of the H1975, indicating that resistant cells (H1975) proliferate faster than sensitive cells (HCC827) by approximately 10 h.

Figure 2:

Growth kinetics of EGFR-mutant NSCLC cell lines over 96 h. Cell proliferation of HCC827 (afatinib-sensitive) and NCI-H1975 (afatinib-resistant) cell lines was monitored over 96 h. Graphs represent the average of triplicate wells ±SD.

The sensitivity of sensitive, resistant, and mixed co-cultures with varying resistance percentages to afatinib was identified in Figure 3 via MTT assay as described in the Section “In vitro growth inhibitory assay”. At 50 % growth inhibition, the fold resistance between the two was calculated to be more than 1000-fold (Table 2). This is evident from the presence of the EGFR T790M mutation in the H1975 cell line. The percentage of resistant cells in the co-culture is necessary to mimic the heterogeneous tumour sub-populations in patients. This ensures the efficacy of continuous and intermittent treatment cycles. Resistant cells were mixed with sensitive cells at a known percentage, and the cell mixture’s drug sensitivity was measured by growth inhibition assays (Figure 3). Co-cultures with resistance percentages of 0.1 and 1.0 displayed the closest sensitivity to pure sensitive cells (0 %), whereas sensitivity was gradually reduced when resistant cells made up >10 % of the population. Therefore, monitoring the progression of resistant cells with subsequent experiments was done with resistance percentages of 0.1% and 0.5 %. Additionally, this low resistance percentage has been documented to be present in metastatic prostate cancer patients [6], 9]. The in vitro resistance percentages of 0.1% and 0.5 % were selected to mimic the clinical real-world tumour composition diagnosed in tumour-sensitive patients harbouring minute drug-resistant cells.

Figure 3:

Dose-response curve with varying resistance percentages to afatinib. Mixed populations of cells were treated with increasing concentrations of afatinib for 96 h, at which point growth inhibition was measured. The total cell number was determined after 96 h and graphed as the percent growth compared to the DMSO control ± SD.

The concentrations used were kept below 500 pM to observe the role of sensitive cells in competing with resistant cells. Thus, according to Figure 3, specific final concentrations of 100 and 500 pM were used alongside the resistance as mentioned above percentages in continuous and intermittent treatments. It is important to note that in Figure 4A and B, the cell count for continuous treatment at 100 pM is almost similar to that for intermittent treatment at 500 pM. This suggests that under these conditions, equal drug efficacy can be achieved, as reported in case reports, particularly in high-dose intermittent treatment regimens in clinics [36], 37]. In both resistance percentages, it was observed that the treatment exposure of 96 h in intermittent treatment significantly (p<0.05) sustained a higher viable cell count than continuous treatment in the concentration of 100 pM. A similar trend was observed for 500 pM but with fewer viable cells. In 192 h, the growth trend for continuous and intermittent treatments at 100 and 500 pM was similarly replicated in 96 h, with more cell growth. This implied that at these concentrations, afatinib only slows the growth of the cells rather than killing them.

Figure 4:

Cell growth of co-culture cells treated with DMSO control (0.01 %), continuous and intermittent treatments for 100 and 500 pM in 96 and 192 h. (A) 0.1 %, and (B) 0.5 % resistance. A similar growth pattern was shown for both 0.1% and 0.5 % resistance, with intermittent treatment having a higher cell count than continuous treatment across 96 hours and 192 h. Data are shown as means ± SD of triplicate cultures of three independent experiments. Significant differences between control and treatment groups were evaluated using one-way ANOVA, and significant differences between treatment groups (continuous and intermittent) were evaluated using an independent sample t-test (*p<0.05).

To quantify the respective sensitive and resistant levels in the treated cell counts in Figure 4A and B, qPCR was used with cell line-specific primers that target the gene of interest, which is indistinguishable between both cell types, namely, exon 19 deletion in sensitive cells and L858R in T790 M-resistant cells. Although the primer sets were designed from other publications (Supplementary Table S1) [38], optimisation was done with evident similarity in both cell numbers increasing linearly with total RNA extracted, followed by a standard curve with specific primers targeting cell line-specific cDNA sequences showing R-values of more than 0.99 for both cell lines (Supplementary Figure S1). Both Figure 5A and B showed similar expression trends, with genes of interest for sensitive cells all below normalised control, in contrast to resistant cells, which were all above normalised control. In Figure 5A and B, upon normalisation at 96 h of DMSO control (0.01 %), the gene expression for sensitive cells reduces gradually with increasing concentration. This correlates with a gradual increase in resistant cells. This trend was similarly replicated between the two gene expressions at 192 h, which is statistically significant compared to the DMSO control. In intermittent treatment, at 96 h, gene expression for sensitive cells is higher, while, interestingly, the expression for resistant cells in intermittent treatment is lower than in continuous treatment at both concentrations. This trend is replicated at 192 h, with the gene expression of resistant cells in intermittent treatment being lower than in continuous treatment, which is statistically significant (p-value < 0.05).

Figure 5:

Gene expression analysis of co-culture cells treated with DMSO and both continuous and intermittent treatments at 100 and 500 pM in 96 and 192 h. (A) 0.1 %, and (B) 0.5 % resistance. The PCR product for sensitive cells is the exon 19 deletion, and the PCR product for resistant cells is L858R. Significant differences between control and treatment groups were evaluated using one-way ANOVA, and significant differences between treatment groups (continuous and intermittent) were evaluated using an independent sample t-test (ns, no significance; *p<0.05).

While it has been shown that gene expression for resistant cells in intermittent treatment is lower than in continuous treatment, resistant cells were fluorescently labelled in a 0.1 % resistance co-culture to observe the cellular growth pattern and to support the qPCR data. The treatment period lasted only 96 h because the fluorescent signals would become even lower beyond this period. Images in Figure 6A show the proliferation of the resistant cells directly in response to the treatment regimen used in this study. In continuous treatment, fluorescently labelled resistant cells proliferate with less sensitive cells surrounding them, while a growing cluster of resistant cells was “caged” in intermittent treatment, showing a slight difference in morphology for resistant cells in both treatments. Both fluorescent cell count (Figure 6B) and fluorescent total co-culture read (Figure 6C) showed that fluorescent resistant cells were higher in continuous treatment as compared to intermittent treatment, and the efficacy of growth inhibition posed by intermittent treatment was calculated to inhibit 50.74 % growth resistance as compared to continuous treatment (Figure 6D).

Figure 6:

Fluorescently labelled resistant cells analysis (labelled with CellTracker™ Orange CMRA) in 0.1 % resistance co-culture. (A) Fluorescent images were captured at 100 × magnification at 96 h. Brightfield images were visualised by an inverted phase-contrast microscope. (B) Total cell count of resistant cells at 96 h (labelled with CellTracker™ Orange CMRA). The fluorescent cell count was done from the top left, top right, middle, bottom left, and bottom right to obtain a representative count of the total co-culture from the microscope glass cover per well. (C) Total fluorescence read at 96 h. The absorbance of the cell lysate was measured using a fluorescent microplate reader. (D) The calculated resistance inhibition efficacy is achieved by intermittent suppression of the growth of resistant cells. Significant differences between control and treatment groups were evaluated using one-way ANOVA, and significant differences between treatment groups (continuous and intermittent) were evaluated using an independent sample t-test (*p<0.05).

Although qPCR analysis and fluorescent analysis showed that intermittent treatment has been shown to delay the emergence of resistance over 192 h, the concentrations of 100 and 500 pM are not what is being experienced in patients. To emulate the plasma drug concentration in patients, the pharmacokinetic profile of afatinib in 24 h was reported as the area under the plasma concentration-time curve (AUC) [39]. The AUC has been extrapolated to 96 h to make the treatment regimen change from once daily (continuous treatment) to once every 4 days (intermittent treatment) (Figure 6). The extrapolated AUC was converted to experimental concentrations in nanomolar according to Equation (1). The experimental plan of both continuous and intermittent treatments was remodelled with high concentrations to analyse their efficacy with pharmacokinetic parameters (Figure 7). While the treatment-free period from the previous experimental setup in Figure 2 was replaced with drug concentrations, this supports the residual concentration in the plasma drug during the clearance period in the pharmacokinetics of afatinib.

Figure 7:

Extrapolated plasma concentration-time curve of afatinib in 96 h using PharmaCalc (Zurich, Switzerland). (Refer to Equation (2)).

Cell count analysis revealed a significant reduction in the number of viable cells between 0 and 96 h. Interestingly, the viable cell count increased from 96 to 192 h for both 0.1% and 0.5 % resistance percentages in continuous and intermittent treatments based on Figure 8. From 0 to 192 h, intermittent treatment sustained a higher cell count than continuous treatment (Figure 9A). In the qPCR analysis, intermittent treatment showed higher gene expression for resistant cells as compared to continuous treatment at 96 h, and it is significantly replicated at 192 h for both resistance percentages, 0.1% and 0.5 % (Figure 9B). A statistically significant reduction of approximately 85 % of viable cells was observed for both 0.1 and 0.5 % resistance at 96 h, followed by an increase in viable cells at 192 h, particularly for 0.5 % resistance (Figure 9A). The gene expression of sensitive cells reduces over time, reaching approximately a 50 % reduction at 192 h (Figure 9B). Inversely, the expression increased by more than 1.5-fold for resistant cells for both 0.1% and 0.5 % resistance percentages, while being statistically significant at 96 and 192 h.

Figure 8:

Illustration representation of the experimental setup for ‘once daily’ and ‘once every 4 days’ treatment regimen.

Figure 9:

Progression of cell growth and gene expression analysis of co-culture cells with 0.1 and 0.5 % resistance treated in 96 and 192 h. (A) The cell count for continuous and intermittent for 0.1% and 0.5 % resistance at 96 hours and 192 h. (B) The gene expression of sensitive and resistant cells, respectively, at 0.1% and 0.5 % resistance. Significant differences between control and treatment groups were evaluated using one-way ANOVA, and significant differences between treatment groups (continuous and intermittent) were evaluated using an independent sample t-test (ns, no significance; *p<0.05).

Discussion

Although adding an elongated treatment-free period to intermittent treatment delayed the emergence of resistance cells, this study showed it required specific conditions, such as pM concentrations and less than 0.5 % de novo resistance. Not only can resistance be delayed, but it is also paired with a significantly higher cell count than continuous treatment. This means that patients who undergo intermittent treatment have a larger tumour burden than those who undergo continuous treatment [40], [41], [42], [43]. Maintaining a potentially large tumour burden while focusing on shrinking tumours as quickly as possible can make both clinicians and patients uneasy. However, the results contradicted each other upon applying a similar concept with nM concentrations converted from pharmacokinetic parameters. Interestingly, resistant cells grow more with lower drug pressure intermittently as opposed to growing slowly with higher concentrations. This might explain the ineffectiveness of intermittent therapy in clinics, as allowing such an extensive period between treatments can lower the plasma drug concentration [44]. One of the attempts to employ intermittent treatment involves high-dose intermittent strategies as opposed to using a similar dosage for continuous treatment [45], [46], [47]. Reports suggested this strategy was mainly used to reduce the toxicity effect of the drug while trying to keep the drug concentration within the therapeutic window [48], [49], [50].

One of the notable reasons that delays the emergence of resistant cells is the presence of sensitive cells, which imposes a survival competition during the treatment-free period for nutrients and space. This effect has been proven at pM concentrations. Although the concentration used was similar for both continuous and intermittent treatments, with a fixed experimental timepoint as illustrated in Figure 1, the presence of a 72-h treatment-free interval appeared to play a role in delaying the growth of resistant cells (Figure 6). According to adaptive therapy, to achieve control in tumour burden that is able to suppress the growth of resistance, it is necessary to have sensitive cells that grow faster than resistant cells because the latter is likely to acquire a resistant mutation. Although Figure 2 has shown that the expected doubling time is the complete opposite, resistance can still be managed below 0.5 % resistance, and this notion has been supported by multiple reports [51], 52]. In Figure 6A, it can be seen in intermittent treatment that a cluster of growing resistant cells is “covered” by surrounding sensitive cells, which stops the growth of resistance via contact inhibition. This observation is contradictory to continuous treatment, as vacant spaces left unattended by dying sensitive cells only allow resistant cells to occupy them and continuously grow without hindrance. This phenomenon has been coined “competitive release” [53], 54]. The importance of keeping the majority of sensitive cells was further emphasised in Figure 9A and B, where higher concentrations that drastically kill sensitive cells allow for the resistance to become the dominant sub-population and drive the tumour growth without any suppression from sensitive cells. Therefore, we are led to believe that continuous treatment at high concentrations promotes the competitive release of de novo resistance cells [55].

Even though intermittent treatment with adaptive therapy principles can delay the emergence of resistance, the question becomes: how long can it delay the inevitable? In this study, the experimental design of intermittent treatment was developed to represent two treatment cycles against continuous treatment (Figure 1). This is to prevent the co-culture from reaching more than 80 % confluence, which would have prevented contact inhibition. It is unknown how many cycles it took for the sensitive cells to no longer be able to suppress the resistant cells. Adaptive therapy is primarily studied based on mathematical modelling to simulate the treatment cycles until tumour progression using the cell-competitive Lotka–Volterra [51], 56] and agent-based models [34]. It has been reported that even with such models, de novo resistance tends to increase slightly with each treatment cycle, eventually leading to therapeutic failure. Thus, making an important point about changing treatment strategies for cancer can only delay it, but not eradicate it, as summarised in Figure 10. However, mathematical models demonstrate the effectiveness of an elongated treatment-free period, and low-dose treatment strategies may be durably maintained for up to 20 cycles – significantly longer than continuous therapy in metastatic castrate-resistant prostate cancer [27]. In melanoma treated with BRAF/MEK inhibitors, it was predicted in a patient-calibrated Lotka–Volterra model that it would have delayed time to progression by 6–25 months with dose rates of 6–74 % as compared to continuous therapy [56].

Figure 10:

Illustrated summarisation of experimental results.

Some notable limitations of the study needed to be mentioned, such as the absence of a tumour microenvironment in the co-culture experimental setup. The presence of immune cells and fibroblasts is said to affect the outcome of adaptive therapy [57], 58]. In addition, the setup was done entirely with a 2-dimensional setup instead of a 3-dimensional one. The dimension of the co-culture does not represent the actual cellular interactions that might be present in vivo or patient tumours [59]. Although we were able to quantify the gene expression level of sensitive and resistant cells via specific primers in qPCR, we were not able to quantify the level of housekeeping genes that are specific for each cell line in the co-culture. This method poses difficulty in establishing a relative fold of gene expression normalised with housekeeping genes via the delta-delta Ct method; instead, the relative readings are based on normalising based on Ct values. Multiple better detection technologies, such as genetically encoded fluorescent tags and fluorescent in situ hybridisation (FISH), are suggested. Besides that, the treatment setup in this study is based on a fixed concentration for 24 h instead of a constantly changing plasma drug concentration-time curve. Even though the AUC of the curve has been converted directly to experimental concentrations, the constant-changing concentrations, which are affected by maximum drug concentration in plasma (C_max), clearance of the drug from plasma (CL/F), terminal elimination half-life (t_1/2), time to reach C_max (t_max), and volume of distribution (V) over 24 h, were not replicated.

Conclusions

Collectively, intermittent treatment with 72 h of treatment-free period suppresses the growth of de novo resistance with 0.1% and 0.5 % resistance percentages at specific concentrations as compared to continuous treatment. These findings proved to be effective in vitro in NSCLC cell lines.