Novel Active Homo-Aza (Lactam) Steroidal Antimetabolites for the Treatment of Human Pancreatic and Colorectal Cancer

Abstract

Background: Colorectal and pancreatic cancers remain therapeutically challenging, with limitations in efficacy and limitations due to toxicity from conventional antimetabolites such as 5-fluorouracil (5-FU), methotrexate (MTX), and gemcitabine (GEM). Steroidal conjugation offers an approach to enhance selectivity and toxicokinetics. Methods: Five novel hybrid homo-aza (lactam) steroidal antimetabolites (GE23, CS18, CS23, KA44, MV16) were synthesized and tested against three pancreatic and four colorectal carcinoma cell lines with distinct molecular characteristics. Antiproliferative activity (MTT), apoptosis (Annexin V/PI), and cell cycle effects were assessed. Thymidylate synthase (TS) and dihydrofolate reductase (DHFR) inhibition was examined via molecular docking, Western blot, and enzymatic assays. Correlations between docking binding scores (DBS) and biological data were analyzed, and effects were compared with reference drugs (5-FU, MTX, GEM). Results: CS23, CS18, and KA44 exhibited the most potent cytostatic activity (mean GI₅₀ 10–80 µM). CS23 also induced high cytocidal effects, strong apoptosis (40% at 72 h), and G1/S arrest. Moreover, docking predicted the high binding affinity of CS23 for both TS (−11.2 kcal/mol) and DHFR (−11.5 kcal/mol), which was validated by Western blot and enzymatic inhibition (IC₅₀ ≈ 20 nM). Correlation analyses showed significant relationships between hybrid steroidal antimetabolites’ cytostatic efficacy and DBS for TS (r = −0.75) and DHFR (r = −0.76), and combined DBS values predicted growth inhibition (r = −0.81, p < 0.01). No simple, universal correlation with single mutations of KRAS, BRAF, PI3K, or TP53 was found. Conclusions: Lactam steroidal antimetabolite hybrids, particularly CS23, act as dual TS/DHFR inhibitors, inducing apoptosis and cell cycle arrest with improved selectivity. Their strong in silico–in vitro concordance provides a compelling preclinical rationale for further evaluation of steroidal antimetabolites as next-generation therapeutics for resistant gastrointestinal malignancies.

1. Introduction

Cancer remains one of the leading causes of morbidity and mortality worldwide, accounting for millions of deaths each year despite substantial advances in diagnosis and therapy. Recent advances in modern medicine have accelerated the development of targeted therapies. However, for many types of cancer, chemotherapy remains the backbone of treatment. Colorectal and pancreatic cancer are both common and lethal, and chemotherapy (including 5-fluorouracil and gemcitabine) is still being used in the neoadjuvant or adjuvant setting [1,2]. The efficacy of conventional chemotherapeutic agents is often hampered by several intrinsic limitations, including poor aqueous solubility, lack of selectivity, systemic toxicity, and the development of multidrug resistance. Consequently, there is a continuous need for novel molecular entities capable of exerting potent antineoplastic activity while displaying improved pharmacological and pharmacokinetic profiles. This includes molecules already used for non-oncologic indications (e.g., GLP-1 receptor agonists [3]) or naturally derived molecules (such as cannabinoids [4]).

Extensive studies conducted over the past few decades have demonstrated the wide applicability of conjugation across various therapeutic domains, ranging from targeted drug delivery and imaging to enzyme inhibition and anticancer therapy [5]. Using this technique, two or more different molecules with usually very diverse profiles can be coupled together, forming a “conjugation”, which in turn carries the effects of both molecules.

Among the most successful classes of anticancer agents are the purine and pyrimidine analogs, which act as antimetabolites, interfering with nucleic acid synthesis and repair [6]. These compounds structurally resemble endogenous nucleobases or nucleosides and are incorporated into cellular metabolic pathways, thereby disrupting essential biochemical processes required for DNA replication and RNA transcription. Their mechanism of action commonly involves the inhibition of key enzymes, such as thymidylate synthase (TS) and dihydrofolate reductase (DHFR) [7].

Pyrimidine analogs, including 5-fluorouracil (5-FU), cytarabine (Ara-C), and gemcitabine, remain cornerstone agents in the treatment of a variety of solid tumors and hematologic malignancies [8]. Their therapeutic efficacy largely stems from the incorporation of modified pyrimidine bases into DNA or RNA, resulting in chain termination or enzyme inhibition. Similarly, purine analogs such as 6-mercaptopurine, 6-thioguanine, cladribine, and fludarabine have shown pronounced activity against leukemias and lymphomas by mimicking purine metabolites and interfering with DNA polymerase function [9]. However, their major drawbacks include systemic toxicity and non-selectivity, which lead to cell death in both cancerous and healthy cells.

To reduce toxicity, the idea of combining these molecules with lipophilic carriers may reduce toxicity and possibly enhance intracellular delivery of chemotherapeutics. Among the numerous scaffolds explored, steroidal conjugates have attracted significant attention due to the intrinsic biological activity of the steroidal backbone and its ability to enhance membrane permeability, metabolic stability, and tissue selectivity. The combination of a steroid nucleus with pharmacologically active moieties has therefore emerged as a promising strategy to generate multifunctional molecules with superior therapeutic potential [10]. Building on this concept, numerous steroidal conjugates have been developed by covalently linking the steroidal framework to various cytotoxic pharmacophores, including nucleoside analogs [11], organometallic complexes (cisplatin [12,13] or others [14,15]) nitrogen mustards [16], nitrosureas [17], taxols [18], and anthracycline [19,20,21] derivatives, as well as carbohydrate derivatives (glycosides [22] or saponins [23]). Hybrid compounds using pyrazoline–thiazole scaffolds have shown promise [24], whereas novel thiazole derivatives hybridized with fluorinated indenoquinoxaline as dual inhibitors targeting VEGFR2/AKT have shown promise against hepatocellular carcinoma [25]. Our team specifically has had extensive experience throughout the years developing different molecules, such as combining steroids with the nitrogen mustards [16,26,27,28] and homo-aza steroids [28,29,30,31].

Thymidylate synthase (TS) is a pivotal enzyme in the de novo synthesis of thymidylate (dTMP), an essential precursor for DNA replication and repair [32]. TS catalyzes the reductive methylation of 2C-deoxyuridine-5C-monophosphate (dUMP) to 2C-deoxythymidine-5C-monophosphate (dTMP), using 5,10-methylenetetrahydrofolate as a cofactor (a source of a one-carbon methyl group). dTMP undergoes two subsequent phosphorylations, generating the precursor of DNA synthesis, 2C-deoxythymidine-5C-triphosphate (dTTP). Since this pathway constitutes the only source of dTTP, TS has emerged as a crucial therapeutic target in cancer chemotherapy. In rapidly proliferating cancer cells, elevated TS activity supports continuous DNA synthesis and contributes to resistance against cytotoxic agents. TS inhibition forms the cornerstone of antimetabolite therapy in gastrointestinal malignancies [33,34]. Agents such as 5-fluorouracil (5-FU) and its prodrug capecitabine exert their cytotoxic effect by forming stable ternary complexes with TS, thereby depleting thymidylate pools and inducing lethal DNA damage.

Another critical therapeutic target in cancer disease is the folate pathway, where the enzyme dihydrofolate reductase (DHFR) undertakes the reduction of dihydrofolate to tetrahydrofolate (THF) using NADPH. The presence of the THF cofactor is of utmost importance, as it contributes to the biosynthesis of purines, pyrimidines (including thymidylate), and several amino acids. DHFR’s role in the synthesis of essential precursors for cell proliferation has established the enzyme as an attractive target for various anticancer drugs, such as aminopterin, methotrexate (MTX), pralatrexate, and trimetrexate [35,36,37].

Building upon these established molecular frameworks, the design of novel homo-aza (lactam) steroidal purine and pyrimidine conjugates represents a promising avenue for developing next-generation chemotherapeutics. By combining the nucleobase-derived antimetabolite activity with the lipophilic, membrane-permeable, and potentially tissue-selective and molecular cancer cell-targeting properties of the lactam steroidal scaffolds, such hybrids may overcome key drawbacks of classical agents—namely, poor selectivity, rapid clearance, and systemic toxicity. In particular, targeting both pivotal enzymes, such as thymidylate synthase (TS) and dihydrofolate reductase (DHFR), through lactam steroidal conjugation could potentiate cytotoxic efficacy while improving pharmacokinetic behavior.

The aim of the present study was to design, synthesize, and biologically evaluate a novel series of homo-aza (lactam) steroidal antimetabolite hybrids targeting key enzymes of nucleotide biosynthesis, namely thymidylate synthase (TS) and dihydrofolate reductase (DHFR). We sought to assess their cytostatic and cytotoxic effects across colorectal and pancreatic cancer models, investigate apoptosis and cell cycle perturbations, and explore whether in silico docking parameters correlate with in vitro biological activity.

2. Results

2.1. The In Vitro Cytostatic and Cytotoxic Activity of the Steroidal Lactam Antimetabolites—MTT Assay

The in vitro drug screening of the novel steroidal lactam antimetabolites was conducted. The cytostatic and cytotoxic effects of the synthesized steroidal lactam antimetabolites (GE23, CS18, CS23, KA44, and MV16) were evaluated in comparison to the reference antimetabolites 5-fluorouracil (5-FU), methotrexate (MTX), and gemcitabine (GEM) using the MTT assay across pancreatic (Hup-T3, Panc 03.27, Panc 08.13) and colorectal (LoVo, HT-29, LS174T, SW403) carcinoma cell lines. Results are depicted in Table 1, Table 2 and Table 3.

Table 1.

In vitro antiproliferative activity of the three reference drug compounds 5-FU, gemcitabine (GEM), and methotrexate (MTX) against seven human cancer cell lines (three pancreatic and four colorectal).

Cell Lines	5-FU			Methotrexate (MTX)			Gemcitabine (GEM)
	GI50	TGI	IC50	GI50	TGI	IC50	GI50	TGI	IC50
Hup-T 3	20 ± 1.5	48 ± 2	>250	26 ± 3.5	158 ± 5.5	>250	11 ± 0.5	23 ± 1	>250
Panc 03.27	18 ± 0.5	58 ± 0.8	128 ± 1.4	34 ± 3	126 ± 7	>250	12.5 ± 0.5	52 ± 3	>250
Panc 08.13	36.5 ± 2.5	75 ± 3.7	229 ± 9.2	55 ± 5.5	>250	>250	19 ± 2.8	64 ± 3.5	>250
LoVo	20 ± 1	250 ± 5	>250	18 ± 2	198 ± 7.5	242 ± 9	12.5 ± 0.5	50 ± 0.8	>250
HT-29	13 ± 1	82 ± 3	>250	8.5 ± 0.5	220 ± 8.5	>250	10 ± 2.5	60 ± 6.0	>250
LS174T	22 ± 1	>250	>250	15 ± 2.5	>250	>250	20 ± 0.5	>250	>250
SW403	16 ± 1	92 ± 2	>250	6.2 ± 0.5	205 ± 8	>250	10 ± 1	18 ± 1	25 ± 5

IC₅₀: half-maximal cytotoxic concentration; GI₅₀: 50% growth inhibition; TGI: total 100% growth inhibition.

Table 2.

In vitro antiproliferative activity of GE23, CS18, and CS23 against seven human cancer cell lines (three pancreatic and four colorectal).

Cell Lines	GE23			CS18			CS23
	GI50	TGI	IC50	GI50	TGI	IC50	GI50	TGI	IC50
HuP-T 3	130 ± 5	235 ± 5	>250	80 ± 5	172 ± 5	238 ± 5	53 ± 0.5	62 ± 0.5	70 ± 1
Panc 03.27	73 ± 2	128 ± 4	>250	50 ± 0.4	95 ± 0.5	245 ± 1	10 ± 0.5	48.5 ± 1	148 ± 3
Panc 08.13	178 ± 7	>250	>250	113 ± 4	203 ± 7	250 ± 8	76 ± 3	128 ± 4	192 ± 7
LoVo	24 ± 2	245 ± 9	>250	32 ± 3	65 ± 2	250 ± 5	30 ± 2	60 ± 2	170 ± 4
HT-29	90 ± 14	>250	>250	32 ± 2	68 ± 4	220 ± 5	65 ± 4	85 ± 4	100 ± 4
LS174T	60 ± 9.8	180 ± 7.5	>250	90 ± 6	172 ± 5	232 ± 5	105 ± 4	140 ± 5	190 ± 5
SW403	100 ± 7.8	172 ± 5	>250	54 ± 3	90 ± 2	200 ± 5	52 ± 1	76 ± 2	114 ± 4

IC₅₀: half-maximal cytotoxic concentration; GI₅₀: 50% growth inhibition; TGI: total 100% growth inhibition.

Table 3.

In vitro antiproliferative activity of KA44 and MV16 against seven human cancer cell lines (three pancreatic and four colorectal).

Cell Lines	KA44			MV16
	GI50	TGI	IC50	GI50	TGI	IC50
Hup-T 3	10 ± 0.5	225 ± 6	>250	42 ± 1	240 ± 7	>250
Panc 03.27	22 ± 0.6	245 ± 8	>250	13.2 ± 0.2	48.5 ± 0.5	54 ± 0.7
Panc 08.13	45 ± 3	>250	>250	69 ± 4.5	198 ± 8.5	>250
LoVo	3.5 ± 0.5	240 ± 4	>250	12.5 ± 0.7	72 ± 1.5	>250
HT-29	4 ± 0.5	>250	>250	200 ± 5	>250	>250
LS174T	100 ± 5	>250	>250	25 ± 2	>250	>250
SW403	150 ± 6.3	>250	>250	12 ± 0.5	30 ± 1	170 ± 3

Among the reference compounds, 5-FU exhibited potent cytostatic activity across the tested gastrointestinal and pancreatic adenocarcinoma cell lines, with GI₅₀ values ranging from 13 to 36.5 μM, and the most pronounced effects were observed in the HT-29 (13 μM) and Panc 03.27 (18 μM) lines (p < 0.01 vs. control). The compound reached total growth inhibition (TGI) at concentrations between 48 and 82 μM in the more sensitive lines, whereas IC₅₀ values generally exceeded 250 μM, indicating a predominantly cytostatic rather than cytotoxic mechanism. The LoVo and LS174T lines demonstrated relative resistance, requiring higher doses for significant growth inhibition (p < 0.05).

MTX showed moderate antiproliferative activity, with GI₅₀ values ranging between 6.2 and 55 μM, indicating variable sensitivity across the panel (p < 0.05). The most pronounced effects were observed in SW403 (GI₅₀ = 6.2 μM) and HT-29 (8.5 μM) cells, followed by LoVo and LS174T (18–20 μM). Complete growth suppression (TGI) was achieved at concentrations from 126–220 μM, with IC₅₀ values exceeding 250 μM in nearly all cell lines. These data confirm that MTX acts primarily through antimetabolic growth arrest, rather than inducing extensive cytotoxicity at the tested exposure period.

Gemcitabine (GEM) demonstrated the highest overall cytostatic potency among the reference drugs, with GI₅₀ values as low as 10 μM in HT-29 and SW403 cells, and moderate activity in other lines (11–20 μM). The TGI values ranged between 18 and 64 μM, while cytotoxic thresholds (IC₅₀) were consistently high (>250 μM), except for SW403 (25 μM), where gemcitabine displayed notable cytotoxicity (p < 0.01 vs. control and vs. MTX). These findings align with gemcitabine’s known mechanism as a nucleoside analog, inducing both DNA synthesis blockade and apoptosis at higher concentrations.

When considered collectively, the reference drugs displayed a ranked cytostatic potency of GEM > 5-FU > MTX, consistent across most colorectal and pancreatic lines. Statistical comparison confirmed gemcitabine’s superiority in reducing cellular proliferation (p < 0.001 vs. MTX, p < 0.05 vs. 5-FU). However, 5-FU exhibited broader efficacy across the cell spectrum, while methotrexate’s effects were more selective. The absence of pronounced cytotoxicity (IC₅₀ > 250 μM in most cases) for all three suggests that their dominant action under the tested conditions was antiproliferative, not lytic.

b. Among the newly synthesized compounds, GE23 displayed moderate antiproliferative activity, with GI₅₀ values ranging from 73 μM (Panc 03.27) to 130 μM (Hup-T3) in pancreatic cells and 24–100 μM in colorectal lines. Its effect was mainly cytostatic, as IC₅₀ values exceeded 250 μM in all cases. The modest activity in Panc 03.27 and LoVo suggests some sensitivity in cells harboring KRAS G12V/G13D mutations, but the overall inhibitory potency remained lower compared with CS18 or CS23 (p < 0.05). These findings indicate that GE23, though active, requires higher concentrations for comparable growth suppression and may serve as a structural baseline in the series.

CS18 demonstrated improved activity over GE23, with GI₅₀ values between 50 and 80 μM in pancreatic lines and 32–90 μM in colorectal models. Notably, CS18 showed potent inhibition in HT-29 (GI₅₀ = 32 μM) and SW403 (GI₅₀ = 54 μM), both harboring KRAS and PIK3CA mutations, suggesting enhanced affinity toward rapidly proliferating or signaling-deregulated cells. Although cytotoxicity remained limited (IC₅₀ > 200 μM), the compound achieved significantly lower GI₅₀ values than GE23 (p < 0.05), indicating superior cytostatic efficiency.

CS23 emerged as the most active compound of the steroidal lactam series. It exhibited low GI₅₀ values across nearly all lines—10 μM in Panc 03.27, 30 μM in LoVo, and 52 μM in SW403—outperforming both CS18 and GE23 (p < 0.01). In addition, its consistent activity across both pancreatic and colorectal cancer cell lines suggests broad-spectrum antiproliferative potential and cytotoxic activity. While IC₅₀ values remained > 100 μM, its TGI values (48–128 μM) indicated effective suppression of cell growth. Overall, CS23 exhibits high cytostatic potency, with potent cytotoxic effects observed at higher concentrations, making it a promising lead molecule.

KA44 displayed cell-selective but very potent inhibition, particularly in LoVo (GI₅₀ = 3.5 μM) and HT-29 (GI₅₀ = 4 μM), surpassing 5-FU and GEM (p < 0.001). Its efficacy decreased markedly in other lines, indicating molecular selectivity toward specific genotypes or microenvironmental conditions (e.g., BRAF mutant profiles). Despite its strong cytostatic activity, IC₅₀ values remained > 250 μM, confirming a non-cytotoxic mechanism at therapeutic levels. KA44’s distinct potency pattern suggests a different mode of action from the steroidal lactams, possibly linked to DNA or enzyme interaction kinetics rather than structural conjugation.

MV16 exhibited variable but notable activity in Panc 03.27 (GI₅₀ = 13.2 μM) and SW403 (GI₅₀ = 12 μM), showing selective inhibition of KRAS-mutated lines, while activity decreased in others (>40 μM). The TGI range (30–230 μM) and absence of marked cytotoxicity suggest reversible, cytostatic suppression. Compared to KA44, MV16 displayed a wider but shallower inhibition profile, consistent with multi-target interference rather than strong binding to a single enzymatic site. The differences between MV16 and KA44 were statistically significant in LoVo and HT-29 (p < 0.05).

When compared across all compounds, CS23 and KA44 demonstrated the highest overall antiproliferative activity. CS23’s consistent low GI₅₀ values across multiple pancreatic and colorectal lines underscore its broad and balanced cytostatic and cytotoxic effect, while KA44’s exceptionally low GI₅₀ values in select lines point to target-specific potency. Statistical analysis confirmed that both compounds exhibited significantly higher inhibition compared to the reference drugs (p < 0.01 vs. 5-FU; p < 0.05 vs. GEM).

CS18 ranked next, offering a favorable intermediate potency with broad applicability and moderate selectivity, whereas GE23 served as a less active structural analog. MV16, though variable, showed promising selectivity toward KRAS-driven tumors and may warrant further exploration in this genetic context.

Overall, the new steroidal lactam analogs demonstrated strong, reproducible cytostatic effects comparable or superior to classical antimetabolites, while maintaining low cytotoxicity profiles

2.2. Therapeutic Index and Selectivity Toward Cancer Cells

To further evaluate the selectivity of the tested compounds toward malignant versus non-transformed cells, GI₅₀ values were determined in three normal human cell lines (FHC, CCD-112CoN, and MRC-5), and therapeutic index (TI) ratios were calculated.

The known antimetabolites MTX, 5-FU, and GEM, as well as the compounds GE23, CS18, CS23, KA44, and MV16, were tested for cytostatic and cytotoxic activity against all three human normal cell lines (FHC, CCD-112CoN, and MRC-5) and showed TGI and IC₅₀ > 250 μΜ.

As shown in Table 4, all steroidal hybrids exhibited markedly higher mean GI₅₀ values in normal cells (468–568 µM) compared with classical antimetabolites, indicating reduced cytostatic activity in non-transformed tissues under the same experimental conditions.

Table 4.

The GI50s (μΜ) generated by each of the tested compounds for each experimental normal human cell line are shown.

Normal Cell Lines			FHC	CCD-112CoN	MRC-5	Mean GI50 ± SE
Test Compounds	GE23	GI50 (μΜ)	601	506	554	553.7 ± 21.4
	CS18		582	502	568	550.7 ± 25.8
	CS23		597	532	574	567.7 ± 20.2
	KA44		516	421	467	468 ± 17.6
	MV16		535	455	501	497 ± 23.5
	5-FU		213	171	194	192.7 ± 11.5
	MTX		238	178	201	205.7 ± 12.2
	GEM		145	103	128	125.3 ± 7.9

Therapeutic index (TI) values were calculated as the ratio of the mean GI₅₀ in normal cells to the GI₅₀ in each tumor cell line. This analysis revealed that several steroidal hybrids displayed favorable and, in some cases, markedly elevated TI values in specific cancer models (Table 5). In particular, KA44, CS23, and MV16 exhibited the highest mean TI values across the tumor panel, reflecting preferential cytostatic activity in malignant cells relative to normal cells.

Table 5.

The Tis demonstrated by each of the tested compounds for each experimental human cancer cell line are presented.

Tumor Cell Lines			Hup-T3	Panc 03.27	Panc 08.13	LoVo	HT-29	LS174T	SW403	Mean TI
Test Compounds	GE23	TI ratio	4.3	7.6	3.1	23.1	6.2	9.2	5.5	5.9
	CS18		6.9	11.0	4.9	17.2	17.2	6.1	10.2	8.5
	CS23		10.7	56.8	7.5	18.9	8.7	5.4	10.9	10.2
	KA44		56.8	25.8	12.6	162.2	141.9	5.7	3.8	11.9
	MV16		11.8	37.7	7.2	39.8	2.5	19.9	41.4	9.3
	5-FU		9.6	10.7	5.3	9.6	14.8	8.8	12.0	9.3
	MTX		7.9	6.0	3.7	11.4	24.2	13.7	33.2	8.8
	GEM		11.4	10.0	6.6	10.0	12.5	6.3	12.5	9.2

The experimental compounds KA44, CS23, and MV16 exhibited the highest therapeutic indexes among all tested compounds (in that order). Based on the very high Tis were generated from KA44, CS23, and MV16 in specific tumor cell lines, indicating targeted specificity and high sensitivity—it is of high therapeutic interest to explore the molecular mechanisms responsible for this phenomenon. Notably, exceptionally high TI values observed in certain cell lines (e.g., KA44 in LoVo cells) resulted from a combination of very low tumor cell GI₅₀ values and high GI₅₀ values in normal cells, indicating pronounced cell line-specific sensitivity rather than uniform selectivity across all models.

2.3. The Effect of the Lactam Steroidal Antimetabolite CS23 Derivative on Apoptosis

Flow cytometric analysis was conducted to determine whether the cytotoxic effects of the steroidal pyrimidine analog CS23 were associated with apoptosis induction. HT-29 and HUPT3 cells were treated with CS23 at an IC50 concentration (μΜ) for 48 and 72 h. Comparative experiments were performed with 5-fluorouracil (5-FU) and gemcitabine (GEM) for 48 and 72 h. 5-FU, which was administered at twice the TGI concentration for the same time intervals, while GEM was given at 250 μΜ (Hup-T3) and 500 μΜ (HT-29).

In the HT29 colorectal cancer cell line, CS23 markedly increased the population of early apoptotic cells, reaching 28.69% after 48 h and 40.29% after 72 h (Figure 1). These values were significantly higher than those observed for 5-FU (17.22% and 27.75%, p < 0.01) and for GEM (10.56% and 19.95%, p < 0.001). The overall distribution pattern indicated that CS23 primarily triggers apoptosis rather than necrosis, as the necrotic cell fraction remained minimal at both time points.

Figure 1.

Apoptotic induction in HT-29 cancer cells upon treatment with CS23, 5-FU, and gemcitabine at different concentrations for 48 and 72 h, respectively, compared with those of untreated cells (control groups). The percentages of viable early apoptotic, late apoptotic, and necrotic cells are demonstrated.

In the Hup-T3 pancreatic cancer cell line, CS23 also demonstrated a distinct pro-apoptotic effect, with early apoptotic populations of 19.16% at 48 h and 38.92% at 72 h (Figure 2). These results were significantly higher than those induced by gemcitabine (12.24% and 19.95%, p < 0.01) but did not differ significantly from 5-FU (43.27% and 49.21%, p > 0.05). The time-dependent increase in apoptotic cells confirms that CS23 acts through a progressive apoptotic mechanism.

Figure 2.

Apoptotic induction in HUPT3 cancer cells upon treatment with CS23 and 5-FU at IC₅₀ and a double concentration of TGI (μΜ) for 48 and 72 h, respectively, compared with those of untreated cells (controls). The percentages of viable early apoptotic, late apoptotic, and necrotic cells are demonstrated. Student’s t-test was used to compare the level of significance between the experimental groups. Differences with p < 0.05 were considered significant.

Overall, these findings demonstrate that CS23 functions as a potent apoptotic inducer, exhibiting superior activity to gemcitabine and comparable or enhanced efficacy relative to 5-FU, depending on the cancer type. Its pronounced apoptotic effect, particularly in colorectal cancer cells, supports the hypothesis that pyrimidine–steroid hybridization enhances the ability of nucleobase analogs to initiate programmed cell death while minimizing necrosis.

2.4. The Effect of CS23 on Cell Cycle Progression

The effect of CS23 on cell cycle progression was evaluated in two cancer cell lines, HT-29 and HUPT3, following treatment at the IC₅₀ and TGI concentrations (μΜ) for 24 and 48 h, respectively. The reference compounds, 5-FU and GEM, were administered at 250 μΜ for 48 h. In all cases, untreated cells were also cultured, thereby acting as control groups.

According to flow cytometric analysis, the majority of untreated HT-29 cells, cultured for 48 h, were distributed in the G1 phase (71.88%), while 8.34% and 5.93% of cells were observed in the S and G2/M phases, respectively. However, treatment with CS23 at the IC50 concentration (μΜ) for 24 h induced significant cell cycle arrest in the G1 phase (68.43%; p < 0.001), along with a significant decrease in the G2/M phase (1.91%; p < 0.001). At the 48 h time point, the untreated HT-29 cancer cells were particularly arrested in the G1 phase (53.54%), with a lower cell population being observed in the S (8.77%) and G2/M (23.01%) phases. Compared to the control group, the administration of CS23 at the TGI concentration (μΜ) resulted in a significant decrease in the G2/M phase population (9.68%; p < 0.001), along with a concomitant increase in the S phase (14.89%; p < 0.001). Both reference compounds, GEM and 5-FU, at 250μΜ for 48 h, significantly reduced the rates in the G2/M phase, decreasing to 0.36% (p < 0.001) and 4.98% (p < 0.001), respectively. At the same time, the cell population was increased in the S phase (11.54% (p < 0.001) for GEM and 17.03% (p < 0.001) for 5-FU) (Table 6, Figure 3). It should be mentioned that under all treatment conditions, the rates in the SubG1 phase were remarkably increased, indicating the apoptotic effect of all compounds, including CS23, GEM, and 5-FU.

Table 6.

Cell cycle distribution in untreated (control) and treated HT-29 cancer cells with CS23 at IC50 and TGI concentrations (μΜ) for 24 and 48 h, respectively. As reference compounds, GEM and 5-FU were administered at 250 μΜ for 48 h.

	% of HΤ-29 Cells in Each Phase of the Cell Cycle
Phases	Control 24 h	CS23 IC50 24 h	Control 48 h	CS23 TGI 48 h	GEM 250 (μΜ) 48 h	5-FU 250 (μΜ) 48 h
SG	2.24%	19.64%	11.14%	31.66%	46.42%	21.81%
G1	75.88%	68.43%	53.54%	39.12%	41.33%	53.55%
S	8.34%	9.47%	8.77%	14.89%	11.54%	17.03%
G2/M	5.93%	1.91%	23.01%	9.68%	0.36%	4.98%

Figure 3.

Cell cycle growth arrest in HT-29 cancer cells induced by CS23, GEM, and 5-FU. HT-29 cells were arrested in the S phase after being treated at the IC₅₀ and TGI concentrations (μΜ) for 24 and 48 h, respectively. Treatment with CS23, GEM, and 5-FU at the respective treatment conditions significantly increased the SubG1 phase. Student’s t-test was used to compare the level of significance between the experimental groups. Differences with p < 0.05 were considered significant.

Following analysis on HUPT3 cells, 59.87% of untreated cells, cultured for 24 h, were primarily distributed in the G1 phase, whereas 24.89% were recorded in the S phase, and 6.45% were recorded in the G2/M phase. Compared to the untreated HUPT3 cells at 24 h, treatment with CS23, at an IC50 concentration for the same time point, led to a significant increase in the rates of the SubG1 phase (14.21%; p < 0.001), accompanied by an increase in the S phase, as well (26.23%; p < 0.001) (Table 7, Figure 4). For the untreated HUPT3 cells, grown for 48 h, the highest percentage was recorded in the G1 phase (53.65%). Compared to the control group, treatment with CS23 at the TGI concentration for 48 h significantly reduced the cell population in the G2/M phase (3.69%; p < 0.001), while concurrently increasing the rates in the SubG1 (19.38%; p < 0.001) and S phases (29.54%; p < 0.001). Treatment with GEM and 5-FU at 250 μΜ for 48 h also induced an increase in the SubG1 phase, reaching 15.86% (p < 0.001) and 27.29% (p < 0.001), respectively. Similar to CS23, the two reference compounds increased the cell distribution in the S phase, simultaneously reducing the rates in the G2/M phase. More precisely, the cell population in the S phase reached 30.54% and 29.09% for GEM and 5-FU, respectively, while the cell distribution in the G2/M phase dropped to 7.6% and 6.26%, respectively.

Table 7.

Cell cycle distribution in untreated (control) and treated HUPT3 cancer cells with CS23 at IC₅₀ and TGI concentrations (μΜ) for 24 and 48 h, respectively. As reference compounds, GEM and 5-FU were administered at 250 μΜ for 48 h.

	% of HUPT3 Cells in Each Phase of the Cell Cycle
Phases	Control 24 h	CS23 IC50 24 h	Control 48 h	CS23 TGI 48 h	GEM 250 (μΜ) 48 h	5-FU 250 (μΜ) 48 h
SG	5.04%	14.21%	6.29%	19.38%	15.86%	27.29%
G1	59.87%	49%	53.65%	43.38%	47.19%	30.59%
S	24.89%	26.23%	27.90%	29.54%	30.54%	29.09%
G2/M	6.45%	6.74%	8.44%	3.69%	7.6%	6.26%

Figure 4.

Cell cycle growth arrest in HUPT3 cancer cells induced by CS23, GEM, and 5-FU. HUPT3 cancer cells were arrested in the S phase after being treated with CS23 at the IC₅₀ and TGI concentrations (μΜ) for 24 and 48 h, respectively. Treatment with CS23, GEM, and 5-FU at the corresponding treatment conditions significantly increased the SubG1 phase. Student’s t-test was used to compare the level of significance between the experimental groups. Differences with p < 0.05 were considered significant.

2.5. Computational Studies to Investigate the Effects of the Steroidal Lactam Antimetabolites on TYSY

To investigate the interaction of the hybrid steroidal–pyrimidine derivatives with thymidylate synthase (TYSY or TS), molecular docking studies were performed using the three-dimensional structure of the enzyme (PDB ID: 1JU6), retrieved from the RCSB Protein Data Bank (PDB). This structure was selected based on its high crystallographic resolution and its biological relevance to the class of compounds under study.

To validate the docking protocol, the co-crystallized dUMP was first removed and then re-docked into the binding site of thymidylate synthase (PDB ID: 1JU6; RMSD = 0.228 v) (Figure 5A). This validation step was performed to assess the reliability of the docking methodology [37].

Figure 5.

(A) Comparison models of the crystallized dUMP (purple) and docked/re-docked dUMP (orange) in TS. The superposition of the crystallographic pose of dUMP took place with the third energy docked pose of dUMP in the binding site of TS (PDB ID: 1JU6; RMSD = 0.228 Å). (B) MTX docking models on thymidylate synthase. Molecular docking studies revealed two-dimensional interactions of the tested drug molecules in the TYSY protein active site, and 3D docking models indicated the positions of the tested compounds in the TYSY active site. (C) 5-FU docking model on thymidylate synthase. (D) GE23 docking model on thymidylate synthase. (E) CS18 docking model on thymidylate synthase. (F) CS23 docking model on thymidylate synthase. (G) MV16 docking to thymidylate synthase. (H) KA44 docking to thymidylate synthase. Hydrogen bonds are indicated by green dotted lines.

The analysis of the docking results revealed that the steroidal hybrids GE23, CS18, and CS23 exhibited particularly strong binding affinities toward thymidylate synthase, with docking binding scores (DBS) of −11.8, −11.4, and −11.2 kcal/mol, respectively. The overall ranking of binding strength followed the order (Figure 5A–G, Table 8) GE23 (−11.8) > CS18 (−11.4) > CS23 (−11.2) > MV16 (−11.0) > MTX (−9.7) > KA44 (−6.7) > 5-FU (−5.1).

Table 8.

Docking scores, interactions, and interacting residues between the chosen compounds and TS.

Compound	Docking Score (kcal/mol)	Interactions (Hydrogen Bonds)	Interactions (Miscellaneous)
KA44	−6.7	Asp218, Leu221, Lys308	Val79 (alkyl), Phe80 (π-alkyl), Trp109 (alkyl), His196 (alkyl), Leu221 (π–σ), Phe225 (π-alkyl), Met311 (π-sulfur)
GE23	−11.8	Gln214 (carbonyl-H), Gly217 (C-H), Asp218 (carbonyl-H), Asn226 (carbonyl-H, H-H)	Ile108 (alkyl), Cys195 (π-alkyl), Leu221 (alkyl), Phe225 (π–σ), Met311 (alkyl)
MV16	−11.0	Arg50, Arg215, His256, Tyr258 (π-donor and conventional H bond)	Arg215 (halogen), Phe80 (π-alkyl), Ile108 (alkyl), Cys195 (π-sulfur), Leu221 (alkyl), Phe225 (π–σ, π-alkyl), Met311 (alkyl)
CS18	−11.4	Tyr135 (carbonyl-H), Gln214 (carbonyl-H), Ser216 (C-H), Asp218 (carbonyl-H), His256 (C-H), Tyr258 (π-donor H)	Arg50 (π-sulfur), Ile108 (alkyl), Leu192 (alkyl), Cys195 (alkyl), Leu221 (alkyl), Phe225 (π-alkyl), Met311 (π-sulfur, alkyl)
CS23	−11.2	Cys195 (π-donor H), Asp218, Lys308 (carbonyl-H)	Phe80 (π-alkyl), Leu192 (π-alkyl), Cys195 (π-sulfur), Phe225 (π-alkyl), Leu221 (π-alkyl, alkyl), Met311 (π-alkyl)
5-FU	−5.1	Cys195, Gln214, Gly217 (C-H), Ser216, Tyr258	Asp218 (π-anion)
MTX	−9.7	Phe80 (carbonyl-H), Cys195 (π-donor H), Ser216, Gly220, Phe225 (π-donor H), Tyr258	Ile108 (π-alkyl), Asp218 (π-anion), Leu221 (π-alkyl), Phe225 (π–π T-shaped), Met311 (π-sulfur),

These hybrid molecules were stabilized within the enzyme’s active site primarily through multiple hydrogen bonds with Asp218, Asn226, and Gln214, as well as extensive π–alkyl and π–sulfur interactions involving Leu221, Phe225, and Met311, which collectively contributed to the enhanced energetic stability of the complexes.

In contrast, KA44, with a lower docking score (−6.7 kcal/mol), formed a limited number of hydrogen bonds and exhibited weaker hydrophobic contacts, indicating reduced conformational stability and a less favorable fit within the binding pocket.

The reference inhibitor, methotrexate (MTX), validated the reliability of the docking model, displaying a docking score of −9.7 kcal/mol—consistent with potent TS inhibition and comparable to those of the active hybrid molecules. Conversely, 5-fluorouracil (5-FU) showed markedly lower binding affinity (−4.9 kcal/mol), which is consistent with its simple chemical structure and the absence of extensive hydrophobic or π-stacking interactions.

Overall, the results indicate that the steroidal hybrids GE23 and CS18 form the most stable inhibitor–enzyme complexes, characterized by an optimal combination of hydrogen bonding and stabilizing π–alkyl and π–sulfur interactions within the catalytic pocket. Their energetically favorable binding surpasses that of conventional reference compounds, suggesting superior bioactivity and selectivity toward the thymidylate synthase target.

To the same extent, methotrexate (MTX) (docking score: −9.7 kcal/mol), as a classical antifolate, showed significant affinity for TS, forming hydrogen bonds with Phe80, Cys195, Ser216, Gly220, Phe225, and Tyr258. The π–π and π–anion interactions with Asp218, Phe225, and Met311 contributed to the overall stabilization.

Although the docking score was strong (−9.7 kcal/mol), MTX displayed slightly fewer hydrophobic contacts than the hybrid derivatives, reflecting its highly polar nature. This supports the hypothesis that incorporating a steroidal scaffold in the hybrid molecules improves lipophilicity and strengthens binding via nonpolar contacts, which may translate into better pharmacokinetic behavior.

The reference antimetabolite 5-FU, a well-known thymidylate synthase inhibitor, demonstrated a relatively weak docking score under the same conditions. It interacted with Ser216, Gly222, and Asn226 through carbonyl and hydrogen bonds, reflecting its known but transient binding mode to the catalytic site. The absence of strong hydrophobic or aromatic contacts likely explains the less favorable binding energy compared with the hybrid analogs. This finding is consistent with experimental data indicating that 5-FU’s activity arises from intracellular conversion to FdUMP, which forms a covalent ternary complex with TS and folate rather than through direct, stable noncovalent binding. 5-FU forms a hydrogen bond between the NH group of the uracil scaffold and Ser216. In the meantime, the carbonyl group at C2 also forms a hydrogen bond with the same amino acid. Additionally, the C4 carbonyl group forms a hydrogen bond with Asn226, while stacking π-stacking interactions were formed between the aromatic region and the amino acid residues Leu221 and Phe225. In our model, the NH at the third position also forms a hydrogen bond with Ser216, and the C2=O group forms a hydrogen bond with the same amino acid. However, the carbonyl group on C4 forms its hydrogen bond with the amino acid Gln214 instead of Asn226 and also forms a C-H bond with Gly217. Despite the differences between the interacting amino acids, our model suggests that 5-FU binds in the same catalytic region as previously reported [37,38,39,40].

Compound GE23, a conjugate of 5-FU and the modified testosterone, is also found to be strongly bound to TS and actually displayed the strongest binding affinity for thymidylate synthase, reflected by its most favorable docking energy, with a docking score of −11.8 kcal/mol. The steroid moiety of the conjugate forms hydrophobic van der Waals interactions with the amino acids Ile108, Leu221, and Met311, and a π-σ interaction is formed between C10 of the cyclopentane ring and Phe225. The 5-fluorouracil moiety binds to the same pocket as the docked model of 5-FU, forming two hydrogen bonds between C2=O, the NH group at the third position, and Asn226. In addition, the C4=O group forms three hydrogen bonds (two conventional and one carbon–hydrogen) with Gln214, Asp218, and Gly217 (C-H). These results suggest that the molecular architecture of GE23 allows optimal accommodation within the catalytic cavity, providing strong anchoring interactions comparable to or exceeding classical TS inhibitors.

Compounds CS18 and CS23 exhibit a partially different binding motif. In the case of CS18 (−11.4 kcal/mol), the steroid part of the conjugates forms alkyl and π-alkyl interactions with the amino acids Ile108, Leu221, and Phe225, which are the three important amino acids of TS’s catalytic site. In addition, the pyrazolepyrimidine moiety forms one π-donor hydrogen bond between the aromatic pyrimidine ring and Tyr258 and π-alkyl interactions with the amino acids Leu192 and Cys195, a π-sulfur interaction between the pyrimidine ring and Met311, and a π-cation interaction with Arg50, but it is also observed that an unfavorable donor–donor interaction is formed between the aniline NH and Arg50, which suggests a loss of binding affinity. Interestingly, the conjugate forms hydrogen bonds with the catalytic region via the introduced linker. In detail, two hydrogen bonds are formed between the ethane region of the linker that is bound to the NH at position 1 of the pyrazole ring and the amino acids Ser216 and His256. Finally, the carbonyl groups of the succinic linker form three hydrogen bonds with the amino acids Tyr135, Gln214, and Asp218.

CS23 (−11.2 kcal/mol) is a modified conjugate of CS18 bearing a phenyl ring at C3 of the pyrazole ring. The testosterone moiety of the conjugate forms a hydrogen bond between the carbonyl group and Lys308, whereas the cyclohexane rings form π-alkyl interactions with Phe80, and the cyclopentane ring forms a π-alkyl interaction with Phe225 of the catalytic site. In the meantime, the aniline NH of the pyrimidine ring forms a hydrogen bond with the amino acid Asp218, and the pyrazole ring forms a hydrogen bond with Cys195. With this amino acid, two more interactions are formed: a π-donor hydrogen bond with the phenyl ring and a π-sulfur interaction with the pyrazole ring. The binding of the pyrazolepyrimidne moiety is further stabilized via π-alkyl interactions between the pyrimidine ring and the amino acids Leu221 and Met311 and between the phenyl ring and Leu192.

MV16 achieved a docking score of −11.0 kcal/mol, forming hydrogen bonds with Arg50, Arg215, His256, and Tyr258 and halogen interactions with Asp21 and Ser59. The ligand’s aromatic rings engaged in π–π and π–σ interactions with Phe31 and Phe225, while Ile108, Cys195, Leu221, and Met311 provided hydrophobic stabilization. The combined polar and hydrophobic contributions imply a multi-anchor binding mode that may translate into efficient enzyme inhibition and slower dissociation kinetics.

KA44 (docking score: −6.7 kcal/mol) exhibited a moderate affinity toward the TS active site. Hydrogen bonding interactions were observed with Asp218, Leu221, and Lys308, stabilizing the compound within the catalytic pocket. Additional hydrophobic interactions with Val79, Phe80, Trp109, His196, and Leu221 (π–σ), as well as π–alkyl contacts with Phe225 and π–sulfur bonding with Met311, contributed to the stabilization of the ligand–enzyme complex. Although the overall docking score was lower than those of other derivatives, KA44’s interaction profile indicates a moderate fit within the active pocket, mainly driven by hydrophobic and alkyl-type contacts rather than polar interactions. The limited number of hydrogen bonds may explain its weaker inhibitory potential relative to the other analogs.

2.6. Computational Studies to Investigate the Effects of the Steroidal Lactam Antimetabolites on DHFR

Molecular docking simulations were also performed using the crystallographic structure of human DHFR (PDB: 1KMS) to elucidate the binding interactions and predict inhibitory potential of the synthesized steroidal lactam derivatives (GE23, CS18, CS23, KA44, and MV16) in comparison with the reference antimetabolites 5-fluorouracil (5-FU) and methotrexate (MTX). Validation of the docking protocol was carried out with SRI-9439 (RMSD = 1.853) (Figure 6A). The docking results (Table 9, Figure 6A–G) revealed strong binding affinities for all steroidal derivatives, with docking scores ranging from −10.9 to −12.2 kcal/mol—significantly exceeding those of the classical inhibitors MTX (−9.5 kcal/mol) and 5-FU (−4.9 kcal/mol)—indicating enhanced energetic stability and improved active site complementarity.

Figure 6.

(A) Comparison models of the crystallized SRI-9439 (magenta) and docked/re-docked SRI-9439 (cyan) in DHFR. The superposition of the crystallographic pose of SRI-9439 took place with the second energy docked pose of SRI-9439 in the binding site of DHFR (PDB ID: 1KMS; RMSD = 1.853 Å). (B) 5-FU docking models for dihydrofolate reductase. Molecular docking studies revealed two-dimensional interactions of the tested drug molecules in the DHFR protein active site, and 3D docking poses indicated the positions of the tested compounds in the DHFR active site. (C) Methotrexate (MTX) docking model for DHFR. (D) GE23 docking model for DHFR. (E) CS18 docking model for DHFR. (F) CS23 docking model for DHFR. (G) MV16 docking model for DHFR. (H) KA44 docking to DHFR.

Table 9.

Docking scores, interactions, and interacting residues between the chosen compounds and DHFR.

Compound	Docking Score (kcal/mol)	Interactions (Hydrogen Bonds)	Interactions (Miscellaneous)
KA44	−12.2	-	Leu22 (alkyl), Pro26 (π-alkyl), Phe31 (π–π stacked, π-alkyl), Phe34 (π-alkyl), Ile60 (alkyl), Pro61 (π-alkyl, alkyl), Lys63 (π-cation)
GE23	−11.7	Ala9 (carbonyl-H), Glu30 (C-H), Asn64 (carbonyl-H)	Phe31 (Halogen), Leu22 (alkyl), Phe31 (π-alkyl, π-σ, π–π stacked), Phe34 (π-alkyl), Ile60 (alkyl), Pro61 (alkyl), Leu67 (alkyl), Val115 (alkyl)
MV16	−10.9	Asp21, Lys63, Asn64, Val115	Asp21 (halogen), Ser59 (halogen), Phe31 (π–σ), Phe34 (π-alkyl), Ile60 (alkyl), Pro61 (π-alkyl), Leu67 (alkyl)
CS18	−11.6	Ala9 (π-donor), Gly30, Ser59 (carbonyl-H), Tyr121 (C-H)	Val8 (π-alkyl), Ala9 (π–σ), Leu22 (π-alkyl), Phe31 (π-alkyl), Phe34 (π-alkyl), Ile60 (alkyl), Leu67 (alkyl)
CS23	−11.5	Gly116 (C-H), Tyr121	Leu22 (alkyl), Arg28 (π-alkyl), Arg32 (π-alkyl), Phe31 (π–σ, π–π T-shaped), Phe34 (π-alkyl), Ile60 (alkyl), Leu67 (alkyl), Val115 (alkyl)
5-FU	−4.9	Ile7, Val115	Glu30 (Halogen), Val8 (π-alkyl), Ala9 (π-alkyl)
MTX	−9.5	Glu30, Thr56 (C-H), Ser59 (C-H), Asn64, Arg70, Val115, Tyr121	Ala9 (π-alkyl), Phe31 (π–π stacked), Ile60 (π–σ)

As expected, 5-FU exhibited the weakest binding affinity (−4.9 kcal/mol) and formed only limited hydrogen and halogen interactions (Ile7, Val115, Glu30), along with weak hydrophobic contacts (Val8, Ala9). The lack of extended aromatic or nonpolar domains prevents 5-FU from achieving deep binding into the DHFR pocket. Its known activity in cells arises primarily through indirect interference with thymidylate synthase, not through DHFR inhibition, which is consistent with the present computational results.

The rank order of predicted DHFR binding affinities was KA-44 (−12.2) > GE23 (−11.7) > CS18 (−11.6) > CS23 (−11.5) > MV16 (−10.9) > MTX (−9.5) >> 5-FU (−4.9).

All hybrid molecules surpassed the standard inhibitors in predicted binding energy, confirming the enhancing role of the steroidal scaffold in achieving deeper and more stable engagement with the DHFR catalytic pocket. KA44 emerged as the most energetically stable ligand, likely due to its dense π-stacking and cationic interactions, whereas GE23, CS18, and CS23 combined hydrophobic anchoring with multiple hydrogen bonds, reflecting multimodal enzyme inhibition potential.

Collectively, these results support the hypothesis that steroidal hybridization improves both DHFR affinity and binding stability, offering a rational design framework for developing dual-target inhibitors capable of disrupting both folate metabolism and DNA synthesis pathways in resistant colorectal and pancreatic cancer.

The classical antifolate MTX displayed a strong docking score of −9.5 kcal/mol, engaging in multiple hydrogen bonds with Glu30, Thr56, Ser59, Asn64, Arg70, Val115, and Tyr121, as well as π–π stacking with Phe31. Additional π–σ and hydrophobic interactions with Ala9 and Ile60 stabilized the binding orientation. Although MTX remains a high-affinity DHFR ligand, the superior docking energies of the steroidal hybrids (−10.9 to −12.2 kcal/mol) suggest that lipophilic conjugation substantially enhances enzyme affinity through improved pocket occupancy and nonpolar complementarity.

GE23 demonstrated a highly favorable docking score (−11.7 kcal/mol), comparable to KA-44, supported by multiple hydrogen bonds (Ala9, Glu30, Asn64) and halogen/aromatic interactions (Phe31, Leu22, Phe34, Ile60, Pro61). The hybrid structure allows the 5-fluorouracil moiety to occupy the canonical substrate pocket, while the steroidal backbone reinforces binding through hydrophobic anchoring along Leu67 and Val115. This dual anchoring behavior mirrors the bifunctional inhibition pattern seen in hybrid antifolate analogs, suggesting that GE23 can mimic both the substrate and cofactor regions of DHFR.

CS18 achieved a docking score of −11.6 kcal/mol, characterized by hydrogen bonding with Gly30, Ser59, and Tyr121 and π-donor interactions with Ala9. Its steroidal moiety formed multiple π-alkyl and alkyl contacts with Val8, Leu22, Phe31, Phe34, Ile60, and Leu67, resulting in extensive hydrophobic stabilization throughout the catalytic pocket. The strong contribution of π-donor hydrogen bonding with Tyr121 and carbonyl hydrogen bonding at Ser59 supports a mixed polar–nonpolar binding profile, which may facilitate durable enzyme inhibition and slower off-rates compared with classic DHFR inhibitors.

CS23 (−11.5 kcal/mol) bound tightly within the active site, forming hydrogen bonds with Gly116 and Tyr121 while exhibiting extensive π-stacking and hydrophobic interactions with Phe31, Phe34, Ile60, Leu67, Val115, and Arg28/Arg32. The π–π T-shaped interaction with Phe31 suggests aromatic stabilization similar to MTX’s pteridine ring system. The phenyl extension of CS23 at C3 likely enhances van der Waals complementarity, improving binding affinity without relying on strong charge pairing.

MV16 (−10.9 kcal/mol) showed substantial binding affinity mediated by four hydrogen bonds (Asp21, Lys63, Asn64, Val115), supported by halogen interactions with Asp21 and Ser59. Hydrophobic contacts with Phe31, Phe34, Ile60, and Pro61 stabilized the ligand conformation. The binding mode indicates a hybrid anchoring mechanism combining polar interactions at the NADPH-binding region and hydrophobic stabilization near the catalytic site. The pattern resembles known gemcitabine–steroid conjugates that modulate redox cofactors and interfere with enzyme turnover.

KA44 exhibited the strongest binding affinity (−12.2 kcal/mol) among all tested compounds. Although no conventional hydrogen bonds were formed, the molecule was stabilized by an extensive network of hydrophobic and aromatic interactions. More specifically, five π-alkyl interactions were formed; two of them are between the pyrimidine ring of reversine and the amino acids Pro26 and Pro61 of the catalytic center, one is formed between the pyrazole ring of reversine and Leu22, and two are formed between the cyclohexane ring of the steroidal moiety of the conjugate and Phe31 and Phe34. In addition, the benzyl ring of reversine plays a central role in the docking of the conjugate by forming a π-π T-shaped interaction with Phe31 and a π-cation interaction with the charged side chain of Lys63. Finally, the conjugate forms multiple alkyl interactions with the hydrophobic catalytic center; the cyclohexylamine portion of reversine forms an alkyl interaction with Pro61. Two alkyl interactions are formed between the methyl groups of the steroidal moiety and Ile60, and finally, one alkyl interaction is formed between Pro61 and the cyclopentane ring of the steroid. These results suggest that KA-44 fits optimally into the hydrophobic cleft of DHFR, functioning as a potent non-classical inhibitor.

2.7. Western Blot Results: GE23, CS18, and CS23 as Potential TYSY Inhibitors

As previously mentioned, three steroidal lactam antimetabolites (GE23, CS18, and CS23) were tested as potential TYSY inhibitors in HT-29 and HUPT3 cells following a 6 h treatment at the TGI concentration (μΜ). To compare their effects, 5-FU was included as a reference compound. According to the Western blot analysis, in HT-29 cells, all three derivatives inhibited TYSY’s activity similar to 5-FU. Specifically, the second band in the blot points out the inhibitory effect of the derivatives and 5-FU, while the specific band is absent in the control group (Figure 7A). The upper bands of TYSY represent TYSY in ternary complexes composed of TYSY, CH2THF, and FdUMP. The density of the upper bands is correlated with the intracellular concentration of FdUMP [38]. Similar results are shown in Figure 7B, where the second band appears upon treatment with the three derivatives and 5-FU, in contrast to its absence in untreated cells.

Figure 7.

The three steroidal lactam antimetabolites, GE23, CS18, and CS23, were tested in HT-29 and HUPT3 cancer cells to investigate their impact on the activity of TYSY and Western blot analysis to determine TS protein expression was performed. The lower band (thick blue arrow) denotes the free form of active TS, and the upper band (thin blue arrow) denotes the ternary complex of inactive TS. β-Actin was used as internal control for protein expression. (A) In HT-29 cancer cells, the second upper band (thin blue arrow) indicates the inhibitory effect of all derivatives on TYSY, in a way similar to 5-FU. The second upper band appears upon treatment with the three derivatives and 5-FU, in contrast to its absence in the untreated cells. (B) Similar results are presented in HUP-T3 tumor cells.

2.8. Inhibition of Dihydrofolate Reductase (DHFR) Activity

DHFR inhibition screening for the tested compounds GE23, CS18, CS23, KA44, MV16, 5-FU, GEM, and MTX was performed in a cell-free DHFR enzymatic activity colorimetric assay. MTX was included as a positive control and reference drug compound. All compounds were tested at a range of concentrations between 5 and 500 nM. All of the results are presented in Figure 8 and Figure 9 as dose–response curves, as well as in Table 10. All of the tested homo-aza steroidal antimetabolites showed a significant inhibitory effect on DHFR activity (higher than MTX). 5FU and GEM demonstrated a low inhibitory effect. The most potent DHFR inhibitor proved to be the KA44. The inhibitory effect of the tested compounds in descending order from the strongest inhibitor to the weakest was KA44 > GE23 > CS18 ≥ CS23 > MV16 > MTX > GEM > 5FU.

Figure 8.

Non-linear dose–response curves of the tested compounds MTX, GE23, 5FU, and CS23 at concentrations of 5–500 nM against DHFR activity.

Figure 9.

Non-linear dose–response curves of the tested compounds CS18, KA44, GEM, and MV16 at concentrations of 5–500 nM against DHFR activity.

Table 10.

Concentrations of the tested compounds that reduce the DHFR activity by 50% (IC50) and related statistics on the corresponding dose–response effects.

Compound	50% DHFR Inhibition (IC50) (nM) ± SE	Adj. R-Square	Significance Level (p)
5-FU	242 ± 4.2	0.952	<0.006
GEM	140 ± 2.2	0.986	<0.003
MTX	30 ± 2.6	0.968	<0.009
GE23	16 ± 0.9	0.984	<0.008
CS18	19 ± 3.0	0.961	<0.01
CS23	21 ± 1.3	0.967	<0.007
MV16	26 ± 3.0	0.953	<0.009
KA44	12 ± 1.1	0.986	<0.01

2.9. Data and Result Correlation

2.9.1. Cumulative Cytostatic/Cytotoxic Effects of Tested Molecules

Cumulative data regarding the cytostatic effects (GI₅₀) of the tested compounds on the screened human pancreatic and colorectal cancer cell lines are presented in Figure 10. The corresponding mean cytostatic potential of the tested compounds in descending order from the strongest to the weakest against the pancreatic cancer cell lines was

Figure 10.

GI₅₀s produced by the tested compounds (GE23, CS18, CS23, KA44, MV16, 5-FU, GEM, and MTX) versus the human pancreatic cell lines HUPT3, Panc 03.27, and Panc 08.13 and the colorectal cancer cell lines LoVo, HT-29, LS174T, and SW403.

GEM > 5-FU ≥ KA44 > MV16 > CS23 > CS18 > MTX > GE2,

while against the colorectal cancer cell lines, it was

KA44 > 5-FU ≥ CS23 > GEM > CS18 > MV16 > MTX > GE23 (p < 0.05).

When the total growth inhibition (TGI) values were considered, reflecting the cytotoxic behavior of the compounds, the mean cytotoxic potency followed a similar but less pronounced pattern (Figure 11). For the pancreatic cancer cell lines, the order was as follows:

Figure 11.

TGIs generated by the tested compounds (GE23, CS18, CS23, KA44, MV16, 5-FU, GEM, and MTX) against the human pancreatic cell lines HUPT3, Panc 03.27, and Panc 08.13 and the colorectal cancer cell lines LoVo, HT-29, LS174T, and SW403.

GEM > 5-FU > CS23 > CS18 ≥ MV16 > KA44 > MTX > GE23.

And for the colorectal cancer cell lines, it was as follows:

CS23 > GEM ≥ CS18 > 5-FU ≥ MV16 > KA44 > GE23 > MTX (p < 0.05).

The mean cytostatic activity (mean GI₅₀) induced by the tested compounds against the four colorectal cancer cell lines was significantly superior to that of the three pancreatic cancer cell lines (mean GI50 = 19.06 and 49.84 μΜ, respectively (p < 0.01)). On the other hand, the mean cytotoxic activity of the tested compounds was similar for both colorectal and pancreatic cell lines (TGI = 147.5 and 145.5 μΜ, respectively).

2.9.2. Correlation of Cytostatic/Cytotoxic Effects of Tested Molecules with Docking Binding Scores on Thymidylate Synthase

The linear correlation between the mean cytostatic effects (GI₅₀) produced by the tested compounds against the seven human cancer cell lines and the respective in silico docking binding scores on TYSY revealed a significant moderate correlation, with Pearson’s correlation index (r) = −0.75 (Adj. R-square = 0.488, p < 0.05) (Figure 12).

Figure 12.

Linear correlation between the mean GI₅₀ (microM) induced by the tested compounds against the 7 human cancer cell lines and the respective in silico docking binding scores (DBS, kcal/mol) on thymidylate synthase. The Pearson correlation coefficient (r) test was used to compare the level of significance between the experimental groups. Differences with p-values less than 0.05 were considered significant.

The linear correlation between the mean cytostatic/cytotoxic effects generated, as they were demonstrated by the corresponding TGIs, by the tested compounds against the seven human cancer cell lines and the respective in silico docking binding scores on TYSY proved to be zero, with Pearson’s correlation index (r) = −0.008 (Adj. R-square = −0.16, p < 0.02).

The cytotoxicity that was induced against the seven human cancer cell lines by the tested compounds was not significantly correlated with their inhibitory activity and docking binding affinity scores on TYSY. On the other hand, the cytostatic effects of the tested compounds were significantly correlated with the induced inhibition of TYSY. The hybrid lactam steroidal antimetabolites presented a very good and significant correlation between their cytostatic activity and docking binding scores in the following order within 95% confidence intervals: CS18 > GE23 ≥ CS23 ≥ MV16 > KA44. These results can be explained by the more complex molecular mechanisms and that the tested compounds induce cytotoxicity on cancer cells beyond the TYSY activity inhibition.

2.9.3. Correlation of Cytostatic/Cytotoxic Effects of Tested Molecules with Docking Scores on Dihydrofolate Reductase

The linear correlation between the mean cytostatic effects (GI₅₀) produced by the tested compounds against the seven human cancer cell lines and the respective in silico docking binding scores on DHFR revealed a moderate correlation, with Pearson’s correlation index (r) = −0.76 (Adj. R-square = 0.512, p < 0.03) (Figure 13).

Figure 13.

Linear correlation between the mean GI₅₀ (microM) induced by the tested compounds against the 7 human cancer cell lines and the respective in silico docking binding scores (DBS, kcal/mol) on DHFR. The Pearson correlation coefficient (r) test was used to compare the level of significance between the experimental groups. Differences with a p-value less than 0.05 were considered significant.

The linear correlation between the mean cytostatic/cytotoxic effects generated, as they were demonstrated by the corresponding TGIs, by the tested compounds against the seven human cancer cell lines and the respective in silico docking binding scores on DHFR proved to be low/moderate and non-significant, with Pearson’s correlation index (r) = −0.44 (Adj. R-square = 0.06, p < 0.12). The weak correlation between DBS/TGI suggests that cytotoxicity involves other pathways, such as off-target molecular effects or general cellular stress from the lipophilic steroid conjugate.

The linear correlation between the 50% inhibition of the DHFR activity (IC₅₀) induced by the tested compounds in a cell-free enzymatic assay, in vitro, and the respective in silico docking binding scores on DHFR showed a very strong correlation, with Pearson’s correlation index (r) = 0.96 (Adj. R-square = 0.915, p < 0.001) (Figure 14).

Figure 14.

Linear correlation between the tested drug concentrations that inhibit DHFR activity by 50% (IC₅₀) and the respective in silico drug docking binding scores (DBS, kcal/mol) on DHFR. The Pearson correlation coefficient (r) test was used to compare the level of significance between the experimental groups. Differences with p-values less than 0.05 were considered significant.

Thus, computational docking studies on the binding affinity of the tested compounds on the DHFR active site can predict the induced inhibition of DHFR activity with high accuracy.

The linear correlation between the mean cytostatic effects (GI₅₀) produced by the tested compounds against the seven human cancer cell lines and the respective drug concentrations that inhibit DHFR activity by 50% (IC₅₀) showed a significant moderate correlation, with Pearson’s correlation index (r) = −0.65 (Adj. R-square = 0.322, p < 0.03) (Figure 15).

Figure 15.

Linear correlation between the tested drug concentrations that inhibit DHFR activity by 50% (IC₅₀, nM) and the mean GI₅₀ (μM) induced by the tested compounds against the 7 human cancer cell lines. The Pearson correlation coefficient (r) test was used to compare the level of significance between the experimental groups. Differences with p-values less than 0.05 were considered significant.

The linear correlation between the mean cytostatic/cytotoxic effects generated, as they were interpreted by the corresponding TGIs, by the tested compounds against the seven human cancer cell lines and the respective drug concentrations that inhibit DHFR activity by 50% (IC₅₀) showed a non-significant weak/moderate correlation, with Pearson’s correlation index (r) = −0.44 (Adj. R-square = 0.059, p < 0.13).

In conclusion, the produced cytotoxicity against the seven human cancer cell lines by the tested compounds was not significantly correlated with their docking binding affinity scores on DHFR, nor with the generated inhibition of DHFR enzymatic activity. On the other hand, the cytostatic effects of the tested compounds were significantly correlated with either their docking binding affinity scores on DHFR or the induced inhibition of DHFR enzymatic activity. The hybrid lactam steroidal antimetabolites presented a very good and significant correlation of their cytostatic activity with the inhibition of DHFR enzymatic activity in the following order within 95% confidence intervals: CS18 > MV16 ≥ CS23 > KA44 > GE23.

The cytotoxic effects of the tested drugs are probably exerted with the participation of other cellular mechanisms beyond the DHFR activity inhibition.

The linear correlations between the mean cytostatic (GI₅₀s) and cytostatic/cytotoxic effects generated, as they were interpreted by the corresponding TGIs, by the tested compounds against the seven human cancer cell lines and the respective combined docking binding scores (DBS DHFR + DBS TYSY) are presented in Figure 16. Significant correlations appeared between cytostatic activity (GI₅₀) of the tested compounds and their respective combined DBS on DHFR and TYSY (Pearson’s correlation index, r = −0.81, Adj. R-square = 0.604, p < 0.01). The hybrid lactam steroidal antimetabolites presented a very good and significant correlation of their cytostatic activity with their respective combined docking binding affinity in DHFR and TYSY binding active sites in the following order within 95% confidence intervals: CS18 > KA44 > MV16 ≥ CS23. These results indicated that the cytostatic anticancer activity of the tested compounds and, more specifically, of the steroidal antimetabolites CS18, KA44, CS23, and MV16 on colorectal and pancreatic cancer can be safely predicted by their combined DBSs on DHFR and TYSY. On the other hand, no significant correlation was observed in the tested compounds’ activity between their combined DBSs on DHFR and TYSY and respective cytotoxic effects, as indicated by the corresponding TGIs (Pearson’s correlation index, r = −0.24, Adj. R-square = −0.097, p < 0.23).

Figure 16.

Linear correlation between the combined docking binding scores (DBS DHFR + DBS TYSY) of the tested compounds on DHFR and TYSY and the mean GI50s (μM) induced by the tested compounds against the 7 human cancer cell lines. The Pearson correlation coefficient (r) test was used to compare the level of significance between the experimental groups. Differences with p-values less than 0.05 were considered significant.

2.9.4. Correlations of the Sensitivity of the Tested Compounds to the Known Actionable and Clinically Significant Mutations in the Tested Cell Lines

Actionable and clinically significant oncogene mutations were detected in the tested human pancreatic and colorectal cancer lines using next-generation sequencing (NGS) technology. Frequent and clinically significant oncogenic mutations in colorectal and pancreatic cancer of the RAS-RAF-MEK-ERK and PI3K-AKT-mTOR signaling pathways, as well as in TP53, are shown in Table 1.

As shown in Figure 17 and Figure 18, conclusions for a specific or dependent anticancer activity of the tested compounds on the screened cancer cell lines, related to the respective mutations in the KRAS, BRAF, PI3K, and TP53 genes they bear, cannot be extracted. No simple universal correlation with individual single KRAS, BRAF, PI3K, or TP53 mutations was observed, although cellular context (potentially defined by a combination of mutations or other factors and combined pathway alterations) influences sensitivity. PANC 08.13 cancer cells that bear the mutation p.G12D c.35G>A in KRAS and bear non-mutated BRAF, PI3K, and TP53 showed the most resistance to the treatment with the tested compounds. On the other hand, LS174T cells that bear the same mutation in KRAS and, in addition, p.D211G c.632A>G mutation in BRAF and p.H1047R c.3140A>G mutation in PIK3CA, appeared more sensitive. Vice versa, PANC 03.27 cells with mutated KRAS (p.G12V c.35G>T) and TP53 (Splice Variant p.?c.375+5G>T) but with wild-type (wt) BRAF and PI3K (wt) are more sensitive to the tested anticancer antimetabolites than SW403 cancer cells that bear the same KRAS mutation; mutated TP53; and, in addition, mutated PIK3CA (p.Q546K c.1636C>A) and wild-type BRAF. Additionally, HT-29 cells with wild-type KRAS, mutated BRAF (p.V600E c.1799T>A), mutated PIK3CA (p.P449T c.1345C>A), and mutated TP53 (p.R273H c.818G>A) present an intermediate sensitivity to the tested drugs. However, it is worth noting that regarding homo-aza steroidal antimetabolites, (a) the KA44 is the most active against HT29 colorectal cancer cells (wt. KRAS; V600E mutBRAF; P449T mutPIK3CA) and LoVo colorectal cancer cells that bear mutated KRAS (p.G13D c.38G>A) and PIK3CB (p.E1051K c.3151G>A), as well as wild-type BRAF and TP53. On the other hand, LS174T cells with mutated PIK3CA (p.H1047R c.3140A>G), KRAS (p.G12D c.35G>A), BRAF (p.D211G c.632A>G), and wild-type TP53 (wt) presented low sensitivity to KA44. Maybe cancer cells with specific mutations of PI3K or specific mutation combinations are more sensitive to the anti-proliferative effects of the KA44. (b) CS23 induces the higher cytostatic and cytotoxic antitumor effect against PANC 03.27 pancreatic cancer cells that carry mutated KRAS (p.G12V c.35G>T), whereas TP53 (Splice Variant p.?c.375+5G>T) has wild-type BRAF and PI3K. On the contrary, SW403 cancer cells bear the same mutation as the PANC 03.27 KRAS mutation and have mutated PIK3CA (p.Q546K c.1636C>A) and TP53 (p.E51* c.151G>T), whereas BRAF is non-mutated and presented significantly lower sensitivity to CS23. In general, regarding KRAS (wild-type and different mutations), LoVo colorectal cancer cells that carry the p.G13D c.38G>A mutation are the most sensitive to the tested compounds.

Figure 17.

The cytostatic effects (GI) of the tested compounds on the seven colorectal (four) and pancreatic (three) cancer cell lines are shown, and the respective mutations of KRAS, BRAF, PI3K, and TP53 in each cell line are presented.

Figure 18.

The cytostatic and cytotoxic effects (TGI) of the tested compounds on the seven colorectal (four) and pancreatic (three) cancer cell lines are shown, and the respective mutations of KRAS, BRAF, PI3K, and TP53 in each cell line are presented.

3. Discussion

The present study introduces and systematically evaluates a new family of homo-aza steroidal antimetabolites designed to exploit the pharmacodynamic vulnerabilities of colorectal and pancreatic cancer cells. These compounds—GE23, CS18, CS23, KA44, and MV16—were conceptualized as dual-target inhibitors that integrate the mechanistic properties of nucleobase analogs (targeting thymidylate synthase, TS) and antifolates (targeting dihydrofolate reductase, DHFR) into a single molecular framework. The conjugation of these pharmacophores onto a steroidal backbone represents a strategic hybridization that merges cytostatic potency with improved lipophilicity, cellular penetration, and potentially enhanced selectivity. Furthermore, this cytostasis could be quantitatively predicted from in silico binding to TS and DHFR.

The steroidal hybrid antimetabolites exhibited moderate but consistent antiproliferative effects, with mean GI₅₀ values spanning 10–80 μM across multiple colorectal and pancreatic cancer models, and this was combined with dual TS/DHFR inhibition. GI₅₀ is the primary potency metric for these compounds, with cytotoxicity (IC₅₀) occurring at significantly higher concentrations. The compounds induced a pronounced cytostatic effect, particularly against colorectal cancer cell lines, while maintaining limited cytotoxicity at TGI levels exceeding 100 μM. CS23 and CS18 emerged as the most active members of the series, with KA44 demonstrating exceptional potency in DHFR inhibition despite more moderate antiproliferative effects in certain cellular contexts.

To further assess selectivity toward malignant cells, GI₅₀ values were determined in three normal human cell lines (FHC, CCD-112CoN, and MRC-5), and therapeutic index (TI) ratios were calculated as the mean GI₅₀ in normal cells divided by the GI₅₀ in tumor cells. All steroidal hybrids exhibited substantially higher GI₅₀ values in normal cells (≈468–568 µM) compared with classical antimetabolites, indicating reduced cytostatic activity in non-transformed cells. TI analysis revealed that KA44, CS23, and MV16 displayed the most favorable selectivity profiles across the tumor panel, with particularly high TI values observed in specific cancer models. These elevated TI values reflect pronounced cell line-dependent sensitivity resulting from low tumor cell GI₅₀ values combined with minimal activity in normal cells, rather than uniform tumor specificity. Collectively, these findings support an improved therapeutic window for selected steroidal hybrids while highlighting the influence of cellular context on compound responsiveness.

Comparisons with reference antimetabolites were intended to provide contextual benchmarks rather than dose-equivalent pharmacodynamic comparisons. However, the cytostatic–cytotoxic profile of these molecules sharply contrasts with classical antimetabolites such as 5-fluorouracil (5-FU) and methotrexate (MTX). While 5-FU and GEM act primarily as DNA-damaging agents that induce both replication arrest and cell death, the steroidal hybrids appear to preferentially induce sustained cytostasis through reversible S-phase arrest and progressive apoptosis rather than acute necrosis, with the exception of CS23, which exhibits high cytostatic potency, with also potent cytotoxic effects observed at higher concentrations. This behavior aligns with the design rationale of targeting metabolic checkpoints—TS and DHFR—rather than direct DNA scission or nucleoside incorporation [1,7,33].

Flow cytometry confirmed that CS23 triggers apoptosis in a time-dependent and dose-dependent manner, with early apoptotic fractions reaching 40% in HT-29 cells at 72 h—significantly exceeding both gemcitabine and 5-FU. Early apoptosis kinetics are detectable earlier, but the 72 h timepoint was selected for consistency with proliferation/MTT timing. The minimal necrotic fraction underscores a regulated death pathway. Moreover, the observed accumulation in the S phase and concomitant increase in the Sub-G1 fraction strongly support a mechanism of replication stress due to thymidylate depletion, consistent with TS inhibition [32,39]. Parallel findings in HUP-T3 pancreatic cells reveal that apoptosis induction is conserved across tissue types. Apoptosis and cell cycle analyses were focused on the lead compound CS23, and broader comparative analyses across all derivatives and time points remain an area for future investigation.

The docking and Western blot data converge to establish TS as a bona fide target of the steroidal derivatives. The compounds exhibited docking energies between −11.2 and −11.8 kcal/mol, outperforming the classical 5-FU (−5.1 kcal/mol). The hydrogen bonds formed with residues Asp218, Gln214, and Asn226, combined with π–alkyl and π–sulfur contacts with Leu221, Phe225, and Met311, emulate the binding geometry of the FdUMP–TS–CH₂THF ternary complex that underlies 5-FU cytotoxicity [32,37].

Western blot validation confirmed this mechanism: the characteristic disappearance of the lower, active TS band and the emergence of upper ternary complex bands in CS23-, CS18-, and GE23-treated cells mirrored the biochemical behavior of 5-FU-treated samples.

The parallelism between in silico docking predictions and biochemical inhibition strongly supports that the steroidal moiety enhances pocket occupancy and stability, possibly through increased hydrophobic interactions and reduced ligand desolvation costs. Importantly, the correlation (r = −0.75, p < 0.05) between TS docking scores and mean GI₅₀ values across all seven cancer cell lines signifies that binding affinity is a strong predictor of cytostatic potency. This correlation implies that TS inhibition is a principal contributor to the observed growth arrest, though not the sole determinant of overall cytotoxicity.

In silico analysis of DHFR binding revealed a uniform trend of strong affinity among the steroidal analogs, with energies from −10.9 to −12.2 kcal/mol—markedly higher than MTX (−9.5 kcal/mol). The leading compound, KA44, achieved extensive π–π stacking with Phe31 and π-cation coupling to Lys63, providing a dense hydrophobic network reminiscent of optimized antifolate scaffolds [35].

These computational predictions were empirically validated through the DHFR enzymatic inhibition assay, which demonstrated low-nanomolar IC₅₀ values for all tested hybrids. KA44 (12 nM) and GE23 (16 nM) were particularly potent, exceeding MTX’s inhibitory capacity (30 nM). The extraordinary linear correlation (r = 0.96, p < 0.001) between docking binding scores and enzymatic IC₅₀ values confirms that the docking model accurately represents the real biochemical environment of DHFR. This one-to-one correspondence establishes computational docking as a predictive surrogate for enzyme inhibition in this compound class.

Interestingly, while all hybrids inhibited DHFR effectively in vitro, the cellular cytostatic effects (GI₅₀) showed only a moderate correlation (r = −0.65, p < 0.03) with DHFR IC₅₀ values, suggesting partial mediation of antiproliferative activity through this pathway. This discrepancy likely arises from pharmacokinetic constraints, such as cellular uptake and subcellular distribution, or from the multifactorial dependence of cytostasis on both TS and DHFR blockade.

A major conceptual outcome of this study is the demonstration that the combined inhibition of TS and DHFR acts synergistically to predict cytostatic outcomes. When the docking binding scores of both enzymes were combined (DBS_TS + DBS_DHFR), a highly significant correlation with the mean GI₅₀ values was obtained (r = −0.81, Adj. R² = 0.604, p < 0.01). This robust relationship indicates that cellular proliferation inhibition depends not merely on the strength of binding to a single enzyme but on the integrated suppression of both folate cycle checkpoints.

Although strong correlations were observed between docking scores, enzymatic inhibition, and cytostatic activity, these associations do not imply direct causality. Future studies will investigate downstream metabolic adaptations and compensatory pathway responses following TS/DHFR inhibition by steroidal hybrids. Rescue experiments using thymidine or folinic acid supplementation would further delineate the relative contribution of TS and DHFR inhibition, which could be pursued in future mechanistic studies.

Mechanistically, this dual blockade disrupts both dTMP synthesis (via TS inhibition) and the regeneration of tetrahydrofolate (via DHFR inhibition), leading to collapse of the nucleotide pool balance and arrest in the S phase. The tight correlation with cytostatic rather than cytotoxic parameters further suggests that this combined interference produces metabolic stasis without triggering extensive necrosis—an effect desirable for maintaining therapeutic index.

The poor correlation between combined DBS values and TGIs (r = −0.24, p > 0.2) provides a crucial mechanistic distinction: cytostasis (GI₅₀) reflects direct enzyme inhibition, whereas cytotoxicity (TGI or IC50) involves secondary events such as mitochondrial failure, oxidative stress, autophagic collapse, etc., that occur beyond the scope of the TS–DHFR axis. These findings substantiate that the primary molecular determinant of antiproliferative potency lies in coordinated TS/DHFR suppression, while additional downstream pathways govern irreversible cytotoxicity.

The hierarchical structure of correlations (DHFR IC₅₀ vs. DBS r = 0.96 > combined DBS vs. GI₅₀ r = −0.81 > individual enzyme DBS vs. GI₅₀ r ≈ −0.75) delineates an information cascade from molecular docking to cellular outcome. The highest correlation between biochemical inhibition and docking reflects near-perfect structural congruence at the enzyme level. The subsequent decrease in correlation strength when cellular endpoints are considered likely reflects biological noise introduced by cell-specific uptake, efflux, and compensatory signaling.

An essential mechanistic element emerging from the current work is the realization that the lactam (omo-aza) nucleus constitutes the primary structural determinant of biological activity within these steroid-based conjugates. While the overall hybridization of a steroidal framework with an antimetabolite pharmacophore provides a favorable topological balance between lipophilicity and polar functionality, it is the aza-lactam domain that seems to drive the high docking affinities and cytostatic efficacy observed across the tested compounds. The cyclic amide introduces both rigidity and a directional hydrogen-bonding pattern that enables an optimal geometric fit within catalytic pockets such as those of thymidylate synthase (TS) and dihydrofolate reductase (DHFR). In contrast to open-chain amides or non-aza steroidal derivatives, the lactam ring imposes a constrained conformation that minimizes entropic penalties upon binding, enhancing enthalpic complementarity and residence time within the enzymatic cavity.

At the electronic level, the presence of a nitrogen atom within the lactam ring increases local electron density and dipole moment, fostering more stable interactions with the polar residues of the target enzyme’s active site. This enhanced polarizability can facilitate π–π or cation–π stacking with adjacent aromatic amino acids (e.g., Phe, Tyr, Trp), while the carbonyl oxygen of the lactam can act as a dual hydrogen bond acceptor—critical for anchoring the conjugate in regions rich in hydrogen bond donors. Furthermore, the omo-aza substitution introduces a partial resonance delocalization that redistributes charge along the conjugated system, improving orbital overlap with key catalytic residues. These quantum-level features provide a plausible structural rationale for the consistently superior docking scores obtained for lactam-containing analogs compared with their non-lactam or simple amide counterparts.

From a medicinal chemistry standpoint, this insight shifts the interpretation of the compounds’ potency from being merely the additive result of steroid–antimetabolite conjugation to reflecting the intrinsic bioactive contribution of the lactam pharmacophore. The lactam ring should therefore be viewed not as a neutral linker but as a dynamic, electronically active pharmacophoric core that orchestrates the spatial orientation and binding energy of the conjugate. This structural innovation represents the principal novelty of the current molecular design: the integration of a lactam-based aza-heterocycle that modulates both steric alignment and electronic interactions, amplifying the efficacy of hybrid molecules beyond what conventional conjugation strategies achieve. Collectively, these findings highlight the omo-aza nucleus as a key driver of selectivity, affinity, and cytotoxic performance, thereby establishing a new conceptual foundation for subsequent optimization of steroidal antimetabolites.

Nevertheless, the persistence of statistically significant correlations across three biological scales (enzyme → cell → combined enzyme pair) is remarkable, demonstrating that the biochemical reality of enzyme inhibition is faithfully transmitted into phenotypic behavior. In pharmacological modeling, this level of translational coherence validates the underlying design hypothesis and underscores the predictive power of integrated computational–experimental approaches [40,41].

Mutation profiling of KRAS, BRAF, PI3K, and TP53 across the seven tested lines revealed no strict genotype–response correlation, indicating that the activity of the steroidal antimetabolites is mechanistically independent of the classical MAPK and PI3K signaling alterations. In other words, no clear association with individual KRAS, BRAF, PI3K, or TP53 mutations was observed, although cellular context appears to influence sensitivity. This possible independence distinguishes them from pathway-specific kinase inhibitors and underscores their potential efficacy across genetically heterogeneous tumors.

Subtle trends, however, were notable. For example, PANC 08.13 cells harboring a single KRAS G12D mutation with wild-type BRAF and TP53 were among the most resistant, suggesting that intact apoptotic and checkpoint machinery may modulate sensitivity thresholds. Conversely, LS174T cells with triple mutations in KRAS, BRAF, and PIK3CA displayed higher sensitivity, possibly reflecting additive metabolic vulnerability. These observations imply that the metabolic blockade imposed by TS/DHFR inhibition exerts a dominant effect that overrides oncogenic signal transduction, rendering these compounds broadly active in mutation-diverse settings.

The divergent correlation profiles between GI₅₀ and TGI values illuminate an important pharmacodynamic distinction. Cytostasis (growth inhibition) is primarily governed by metabolic enzyme inhibition, whereas cytotoxicity (total cell kill) involves secondary network effects, e.g., mitochondrial depolarization, reactive oxygen species, and checkpoint failure. The lack of correlation between TGI and docking energies (r ≈ 0) underscores that beyond a threshold of metabolic disruption, cell fate becomes dominated by stochastic downstream responses rather than direct enzymatic blockade.

This separation suggests that the aza steroidal hybrids act as “metabolic brakes” rather than outright cytotoxins. Their ability to sustain long-term growth inhibition with minimal necrosis could translate into favorable safety profiles in vivo that have been demonstrated in our previous work [26,28]. Furthermore, selective cytostasis may sensitize tumors to subsequent insults—radiation, DNA alkylators, or immunotherapies—positioning these molecules as candidates for combination regimens.

From a drug design perspective, the correlation between combined docking scores and cytostatic activity offers a quantitative predictive metric. The regression equation derived from the current dataset could serve as a calibration model for future analogs, where a combined DBS_TS + DHFR ≤ −22 kcal/mol predicts a GI₅₀ below 30 μM with 95% confidence. This kind of mathematical predictability is exceedingly rare in pharmacochemistry and provides a tangible framework for virtual high-throughput screening.

In essence, the current results exemplify a closed mechanistic loop: computational docking predicts biochemical inhibition, which in turn predicts phenotypic cytostasis. Such continuity validates the theoretical premise of multi-target rational design and places these molecules within the emerging paradigm of network pharmacology, where simultaneous modulation of interconnected nodes (TS, DHFR, folate metabolism) yields superior therapeutic control.

The steroidal nucleus, beyond serving as a lipophilic carrier, likely facilitates membrane permeability and intracellular retention. Its nonpolar surface complements the hydrophilic nucleobase moiety, yielding an amphiphilic balance optimal for passive diffusion and subcellular localization. Moreover, conjugation may confer protection from deamination and efflux, prolonging intracellular exposure relative to classical antimetabolites.

From a pharmacokinetic standpoint, the lipophilic skeleton may also favor binding to plasma proteins and reduced renal clearance, extending half-life. These advantages could offset the need for high dosing, an enduring limitation of 5-FU and MTX therapies. The modest intercellular variability in GI₅₀ across the seven lines (SD < 25%) supports uniform bioavailability within the experimental system.

Despite its comprehensive design, the present study has several inherent limitations that warrant cautious interpretation. First, all biological experiments were conducted in vitro using established cell lines; therefore, pharmacokinetic, metabolic, and immunologic factors influencing drug behavior in vivo remain uncharacterized. The lack of in vivo validation limits the extrapolation of cytostatic and enzymatic findings to clinical relevance. Second, while the molecular docking and correlation analyses demonstrated strong statistical associations, these relationships cannot fully account for dynamic conformational changes of enzymes or solvent effects that occur under physiological conditions. Third, this study employed a limited panel of pancreatic and colorectal cell lines, which—although genetically diverse—may not capture the full heterogeneity of human tumors, particularly regarding transporters, efflux mechanisms, and metabolic enzymes. Fourth, no direct structural confirmation (e.g., crystallography or molecular dynamics) was performed to verify the predicted binding modes of the hybrids. Finally, while correlations between docking scores and biological activity were robust, they rely on mean group values rather than individual cellular kinetics, introducing potential oversimplification. Future studies incorporating animal models, kinetic modeling, and co-crystal structures are necessary to substantiate the translational validity of these promising steroidal antimetabolites.

However, the translational relevance of this work lies in demonstrating that dual-target hybridization can achieve predictably synergistic inhibition at the enzymatic and cellular levels. In clinical terms, such molecules could potentially overcome resistance mechanisms associated with isolated TS or DHFR inhibition—most notably, upregulation of TS or folate transporters following 5-FU or MTX exposure [1,8,38]. By simultaneously depriving cells of thymidylate and tetrahydrofolate, the hybrids render compensatory metabolic adaptation infeasible.

The present study is limited to in vitro and in silico analyses and, therefore, does not address in vivo pharmacokinetics, efficacy, or systemic toxicity. Building upon the mechanistic and correlative framework established herein, these results should be regarded as a preclinical proof of concept, providing a mechanistic and biological rationale for future research, which should extend beyond in vitro and in silico validation toward comprehensive preclinical evaluation. The next critical step is the assessment of the pharmacokinetic, metabolic, and toxicological profiles of the steroidal antimetabolites in vivo to determine bioavailability, organ distribution, and stability within the folate metabolic network. Advanced molecular dynamics simulations and crystallographic studies of compound–enzyme complexes could further refine the understanding of dynamic binding events and conformational adaptability within the TS and DHFR active sites. In parallel, testing on patient-derived organoids or xenografts bearing specific mutational constellations (KRAS, PIK3CA, BRAF, TP53) may clarify the mutation-independent efficacy observed in vitro and guide biomarker-driven therapy design. Finally, structural optimization through rational modification of the steroidal moiety could improve solubility, cellular uptake, and selectivity, while combination studies with immune or DNA repair modulators may enhance therapeutic synergy. Collectively, these directions chart a translational pathway from predictive modeling to precision oncology applications.

4. Materials and Methods

4.1. Chemistry

Our group aimed to generate hybrid molecules bearing a modified testosterone steroid, coupled with the antimetabolites 5-FU, gemcitabine, aza-reversine, and a mimic of the DNA base adenine, 1H-pyrazolo[3,4-d]pyrimidin-4-amine, through ester bonds. As linkers, succinate and a simple acetate linker were used (Figure 19). The purity of the synthesized compounds was tested using NMR and LC-MS analysis. The supporting documentation includes the synthetic procedures, the 1H-NMR, 13C-NMR, and the LC-MS analysis of the final compounds (Supplementary Material S1).

Figure 19.

The chemical structures of 5-FU, gemcitabine, azareversine, 1H-pyrazolo[3,4-d]pyrimidin-4-amine, 3-phenyl-1H-pyrazolo[3,4-d]pyrimidin-4-amine, steroidal derivatives CS18 and CS23, GE23, KA44, and MV16.

4.1.1. Synthesis of the Modified Testosterone Steroid

The synthesis of the conjugates started with the preparation of homo-aza-testosterone 7, according to the established experimental procedure developed by our group [31]. Initially, the hydroxyl group (-OH) is protected using acetic anhydride in the presence of pyridine and 4-dimethylaminopyridine (DMAP) (Figure 20). The ketone reacts to form the corresponding oxime in the presence of hydroxylamine hydrochloride. The Beckmann rearrangement of the oxime group in the presence of SOCl₂ in dioxane leads to lactam 6 in 62% yield. Deprotection of -OH group using aqueous LiOH yields homo-aza-testosterone (7) in 75% yield.

Figure 20.

Synthesis of homo-aza-testosterone (7).

4.1.2. Synthesis of KA44, an Azareversine Derivative

1. Synthesis of the azareversine

The synthesis of azareversine was initiated with the reaction of 6-chloro-2-fluoropurine and benzylamine, as previously described by our group [42]. Afterwards, a nucleophilic aromatic substitution with tert-butyl-(4-aminophenyl) piperazine-1-carboxylate, followed by acidic hydrolysis, yielded azareversine 11 (Figure 21) [9].

Figure 21.

Synthesis of azareversine 11.

• b.Synthesis of the steroid hybrid of azareversine, conjugate KA44

The homo-aza-testosterone 7 reacts with succinic anhydride, forming acid 12. This reaction is followed by the coupling of 12 with azareversine in the presence of EDCI and DMAP in DMF solvent. The hybrid KA44 (13) was isolated in 73% yield (Figure 22).

Figure 22.

Synthesis of the steroid hybrid of azareversine, KA44 (13).

4.1.3. Synthesis of Conjugate GE23, a 5-FU/Homo-Aza-Testosterone Hybrid Derivative

The hybrid molecule carrying the antimetabolite 5-FU was also synthesized from the modified steroid 7. 5-FU reacted with bromoacetic acid in order to introduce an acetate linker, which is then esterified with homo-aza-testosterone 7. Conjugate GE23 was isolated in 92% yield (Figure 23).

Figure 23.

Synthesis of the 5-FU hybrid derivative GE23.

4.1.4. Synthesis of Pyrimidine Conjugates CS18 and CS23

1. Synthesis of 1H-pyrazolo[3,4-d]pyrimidin-4-amine (18)

The synthesis of derivative 18 began with the preparation of the amino-carbonitrile 17 based on a known experimental procedure [43,44]. Compound 17 was cyclized with formamide, yielding 1H-pyrazolo[3,4-d]pyrimidin-4-amine 18 in 70% yield (Figure 24).

Figure 24.

Synthesis of 1H-pyrazolo[3,4-d]pyrimidin-4-amine 18.

• b.Synthesis of the steroidal derivative CS18

To produce the steroidal derivative CS18, the following synthetic route was carried out. Compound 18 was first treated with NaH in DMF solvent, followed by the addition of chloroethanol at room temperature. The mixture was then heated, yielding the alkylated pyrazolo[3,4-d]pyrimidin-4-amine 19. The latter reacted with acid 12 in an esterification reaction, using 4-dimethylaminopyridine (DMAP) and ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDCI) as coupling agent, and a catalytic amount of 4-dimethylaminopyridine (DMAP) in dry DMF. The hybrid steroidal derivative CS18 was isolated using column chromatography in 48% yield (Figure 25).

Figure 25.

Synthesis of the steroidal derivative CS18.

• c.Synthesis of the steroidal derivative CS23

For the synthesis of the steroidal derivative CS23, the 3-phenyl-1H-pyrazolo[3,4-d]pyrimidin-4-amine was first prepared. The procedure starts with an iodination reaction using N-iodosuccinimide (NIS) in dry DMF, leading to the iodinated derivative 19 at the 3-position. The alkylation of compound 19 in dry DMF was followed by the addition of chloroethanol and heating, leading to the formation of alcohol 20. Subsequently, a palladium-catalyzed Suzuki coupling was performed. Compound 20 reacted with phenylboronic acid in the presence of Na₂CO₃, Pd(OAc)₂, and PPh₃ (with a Pd(OAc)₂/PPh₃ ratio of 1:4) in a 1,4-dioxane/water (5/1) solvent system, yielding the desired 3-phenyl-1H-pyrazolo[3,4-d]pyrimidin-4-amine 21 in quantitative yield (Figure 26).

Figure 26.

Synthesis of the 3-phenyl-1H-pyrazolo[3,4-d]pyrimidin-4-amine (21).

The final step in the synthetic route of the steroidal derivative CS23 involves the esterification reaction between derivative 21 and testosterone derivative 12 in the presence of EDCI and a catalytic amount of DMAP in dry DMF. The hybrid steroidal derivative CS23 was isolated in 53% yield (Figure 27).

Figure 27.

Synthesis of the hybrid steroidal derivative CS23.

4.1.5. Synthesis of the Conjugate MV16, a Gemcitabine Steroidal Hybrid

Homo-aza-testosterone (7) reacted with succinic anhydride to form acid 12. The conjugation methodology was carried out according to our group’s synthetic protocol [45]. Compound 12 was then conjugated with gemcitabine in the presence of (benzotriazol-1-yloxy)tris(dimethylamino)phosphonium (BOP) and 4-dimethylaminopyridine (DMAP) in dry DMF at 36 °C for 48 h. The hybrid MV16 was isolated by column chromatography in 46% yield (Figure 28).

Figure 28.

Synthesis of the gemcitabine steroidal hybrid MV16.

4.2. Drug Preparation

Five novel lactam steroidal antimetabolites, CS18, CS23, GE23, KA44, and MV16 (Figure 19), were synthesized according to the previously described experimental procedure. Stock solutions were prepared by dissolving the required amount of the substances in DMSO solvent in order to reach the desired concentration of 15 mM. The final volume of DMSO did not exceed 1% of the culture medium. In addition, stock solutions of methotrexate (MTX), 5-FU, and gemcitabine (GEM) dissolved in normal saline (NaCl 0.9%) at a concentration of 15 mM were prepared.

4.3. Cell Lines

Our study was carried out in seven human cancer cell lines with distinct molecular characteristics, of which three were of pancreatic cancer origin (HUPT3, Panc 03.27, and Panc 08.13) and four were of colorectal cancer origin (LoVo, HT29, LS174T, and SW403) (Table 1). The selected colorectal and pancreatic cancer cell lines represent a spectrum of clinically relevant molecular backgrounds and intrinsic drug sensitivities, allowing assessment of compound activity across heterogeneous cellular contexts. All cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in different culture mediums according to the supplier’s instructions. Panc 03.27 and Panc 08.13 cells were cultured in RPMI 1640 medium, supplemented with 15% heat-inactivated fetal bovine serum (FBS) and 10 µg/mL insulin. HUPT3 cells were cultured in EMEM medium, supplemented with 10% heat-inactivated FBS, 2 mM of glutamine, 1% non-essential amino acids, and 1% sodium pyruvate. LoVo and LS174T cells were cultured in RPMI 1640 medium, supplemented with 10% heat-inactivated FBS. HT29 and SW403 cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% heat-inactivated FBS. All growth mediums were supplemented with 1% penicillin/streptomycin. All cancer cell lines were cultured as monolayers and maintained at 37 °C in a humidified 5% CO₂ atmosphere.

For the assessment of cytostatic and cytotoxic activity of the new lactam steroid hybrid antimetabolites (GE23, CS18, CS23, KA44, and MV16) in normal human cells, three human cell lines, the FHC (CRL-1831; normal human epithelial colonocytes), CCD-112CoN (CRL-1541; normal human colon fibroblasts), and MRC-5 (CCL-171; normal human lung fibroblasts) were tested. Cell lines were provided by ATCC. CCD-112CoN and MRC-5 cells were cultured in Eagle’s Minimum Essential Medium (EMEM), supplemented with 10% non-heat inactivated fetal bovine serum (FBV). FHC cells were cultured in DMEM:F12 medium, supplemented with 10 mM HEPES (for a final conc. of 25 mM), 10 ng/mL cholera toxin, 0.005 mg/mL insulin, 0.005 mg/mL transferrin, 100 ng/mL hydrocortisone, 20 ng/mL human recombinant EGF, and 10% FBS. All cell lines were cultured as monolayers and maintained at 37 °C in a humidified 5% CO₂ atmosphere. All cell culture materials were obtained from Sigma-Aldrich (Merck KGaA, Darmstadt, Germany).

Actionable and clinically significant oncogene mutations were detected in the tested human pancreatic and colorectal cancer lines using next-generation sequencing (NGS) technology. Frequent and clinically significant oncogenic mutations in colorectal and pancreatic cancer of signaling pathways, as well as in TP53, are shown in Table 11.

Table 11.

Actionable and clinically significant oncogene mutations in KRAS, BRAF, PI3K, and TP53 in the tested human pancreatic and colorectal cancer cell lines.

	Cell Lines	KRAS	BRAF	PI3K	TP53
Pancreatic cancer	PANC 08.13	p.G12D c.35G>A	wt	wt	wt
	PANC 03.27	p.G12V c.35G>T	wt	wt	Splice Variant p.?c.375+5G>T
	HUP-T3	p.G12R c.34G>C	wt	wt	p.R282W c.844C>T
Colorectal cancer	LoVo	p.G13D c.38G>A	wt	PIK3CB p.E1051K c.3151G>A	wt
	LS174T	p.G12D c.35G>A	p.D211G c.632A>G	PIK3CA p.H1047R c.3140A>G	wt
	SW403	p.G12V c.35G>T	wt	PIK3CA p.Q546K c.1636C>A	p.E51* c.151G>T
	HT-29	wt	p.V600E c.1799T>A	PIK3CA p.P449T c.1345C>A	p.R273H c.818G>A

4.4. In Vitro Anticancer Activity—MTT Assay

The cells were plated in 96-well plate at a density of 1 × 10⁴ cells/mL per well and maintained for 72 h in incubator and grown as monolayers. After 24 h, they were treated with 0.1–100 μM of the compounds GE23, CS18, CS23, KA44, and MV16 (10–200 μM for LT and LE) for 48 h. 5-FU, gemcitabine, and MTX were included as reference drug compounds. The viability of cultured cells was estimated by (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) metabolic assay, as described previously [27,42,46]. MTT (Sigma, St Louis, MO, USA) was dissolved in PBS at a concentration of 5 mg/mL, filter-sterilized, and stored at 4 °C. MTT (0.2 mL of stock solution) was added to each culture (per mL) and incubated for 3 h at 37 °C to allow metabolization. Formazan crystals were solubilized by acidic isopropanol (0.04 N HCl in absolute isopropanol in a ratio of 1:3 v/v). Absorbance of the converted dye was measured at 540 nm wavelength on an ELISA reader (Versamax, Orleans, LA, USA). The mean concentrations of each drug that generated 50% or total (100%) growth inhibition (GI₅₀ and TGI, respectively) and the drug concentrations that produced cytotoxicity against 50% of the cultured cells [half-maximal inhibitory concentration (IC₅₀)] were calculated using the linear regression method. Using seven absorbance measurements [time 24 h (Ct24), control growth 72 h (Ct72), and test growth in the presence of drug at 5 concentration levels (Tt72x)], the percentage of growth was calculated at each level of the drug concentrations. The percentage growth inhibition was calculated according to National Cancer Institute (NCI) as follows: [(Tt72x) − (Ct24)/(Ct72) − (Ct24)] × 100 for concentrations for which Tt72x > Ct24 and [(Tt72x) − (Ct24)/Ct24] × 100 for concentrations for which Tt72x < Ct24. GI₅₀ was calculated from [(Tt72x) − (Ct24)/(Ct72) − (Ct24)] × 100 = 50, TGI from [(Tt72x) − (Ct24)/(Ct72) − (Ct24)] × 100 = 0, and IC₅₀ from [(Tt72x) − (Ct24)/Ct24] × 100 = 50. All the experiments were carried out in triplicate.

4.5. Analysis of Apoptosis by Flow Cytometry

Apoptosis was assessed using the FITC Annexin V Apoptosis Detection Kit with 7-AAD (Biolegend, San Diego, CA, USA). For the purpose of this study, two cell lines were tested, HT-29 and HUPT3 cancer cells, seeded (1.0 × 10⁶ cells/well) in a 6-well plate and maintained for 24 h at 37 °C in humidified 5% CO₂ atmosphere. After 24 h of cell growth, culture medium was replaced with fresh medium, and cells were exposed to CS23 at their respective IC₅₀ concentrations from the MTT experiments (for HT29, IC₅₀ = 100 μΜ; for HUPT3, IC₅₀ = 70 μΜ). The reason for this choice was that it was found to be consistently the most cytotoxic of the molecules. As reference drug compounds, 5-FU and gemcitabine were used for 48 and 72 h. 5-FU was tested at twice the concentration of its TGI (for HT29, 165 μΜ; for HUPT3, 95 μΜ) for 48 and 72 h. In both cases, untreated cells served as controls. In every case, the experiment was performed in triplicate. Annexin V/PI was intentionally performed at 72 h to assess cumulative apoptotic burden.

For the apoptosis assay, cells were washed with ice-cold PBS (pH 7.4) and detached enzymatically with standard trypsinization. All required centrifugations were carried out at 1500 rpm for 5 min, including culture medium discard and two washing steps with 2 mL of cold cell staining buffer (BioLegend, San Diego, CA, USA). Subsequently, pellets were re-suspended with 150 μL of Annexin V Binding Buffer, and 100 μL of cell suspension was transferred. Cells were stained with 5 μL of FITC Annexin V and 5 μL of 7-AAD Viability Staining Solution and then incubated for 15 min at room temperature in the dark. Afterwards, 400 μL of Annexin V Binding Buffer and 400 μL of cell staining buffer were added before analyzing on the flow cytometer (CyFlow^®, SL, Partec, GmbH, Münster, Germany). For each sample, 1 × 10⁴ events were acquired, and analysis was carried out in triplicate. Flow cytometric analysis was performed using the Partec FloMax software (v2.9 SL, Partec, GmbH, Münster, Germany).

4.6. Cell Cycle Growth Arrest Analysis by Flow Cytometry

Cell cycle progression was analyzed according to the CyStain PI Absolute T reagent kit (Sysmex, Partec, GmbH, Münster, Germany). HT29 and HUPT3 cancer cells were cultured in a 6-well plate and incubated for 24 h at 37 °C in a 5% CO₂ atmosphere. After 24 h, the growth medium was replaced with fresh medium. Untreated cells served as controls, while cells were treated with CS23 at their respective IC₅₀ concentrations for 24 h and 48 h. Similarly, both cell lines were treated with CS23 at their TGI concentrations for 48 h. As reference drug compounds, 5-FU and gemcitabine were included at their respective TGI concentrations and treated for 48 h. To perform cell cycle analysis, cells were washed with ice-cold PBS (pH 7.4) and detached using standard trypsinization. Centrifugation steps were performed at 1500 rpm for 5 min, with washing steps of 2 mL of cold cell staining buffer (BioLegend, San Diego, CA, USA). Pellets were then resuspended in 150 μL of nuclei extraction buffer and incubated in the dark for 15 min with gentle shaking. Once DNA was extracted, DNA content was stained with 2 mL of cell staining solution (containing staining buffer, PI, and RNAse) and incubated for 30 to 60 min in the dark before samples were analyzed in the flow cytometer (CyFlow^®, SL, Partec, GmbH, Münster, Germany). For each sample, 1 × 10⁴ events were acquired, and analysis was carried out in triplicate. Flow cytometric analysis was performed using the Partec FloMax software.

4.7. Western Blot Analysis—Thymidylate Synthase (TYSY) Activity Inhibition

To further explore the inhibitory impact of the novel lactam steroidal antimetabolites on thymidylate synthase’s activity, Western blot was performed in two cell lines, HT-29 and HUPT3, which were treated with GE23, CS18, and CS23 at TGI concentration (10 μΜ) for 6 h, while for a reference drug compound, 5FU was included. After treatment, cell lysates were collected and centrifuged at 20,000× g for 10 min at 4 °C. Supernatants were transferred to a new tube, and the total protein concentration was determined according to the Bradford assay. All the antibodies (Table 12), biotinylated protein ladder (#7727, dilution 1:1000), nitrocellulose membranes (#12369), and 20× lumiGLO reagent and 20× peroxide (#95538S), were purchased from Cell Signaling Technology (Danvers, MA, USA). Equal amounts of protein (40 μg) and ladder were separated using a 12% sodium dodecyl sulfate–polyacrylamide gel. The separated protein bands were transferred onto nitrocellulose membranes. Subsequently, the membranes were incubated for 1 h at room temperature with a blocking solution of 5% bovine serum albumin in Tris-buffered saline with 0.1% Tween. After the blocking step, membranes were incubated overnight at 4 °C on a shaker with the primary antibody against thymidylate synthase (TYSY) (purified rabbit polyclonal antibody, catalog #AP6682b, Abcam). Following the 24 h of incubation, the primary antibody was removed, and the membranes were incubated with a secondary antibody (anti-rabbit IgG) at 37 °C for 1 h. The antibody-binding bands were visualized using 20× lumiGLO reagent and 20× peroxide (dilution 1:20). The band density was quantified by the Image-Pro Plus software version 6.0 (Media Cybernetics, Rockville, MD, USA) and normalized against loading controls (β-actin). The inhibitory forms against TYSY were normalized relative to total TYSY.

Table 12.

Antibodies used for Western blot analysis of thymidylate synthase.

Antibody	Dilution	Catalogue No.	RRID	Supplier
Primary Antibodies
Rabbit β-Actin (loading control) polyclonal antibody	1:2000	#4967	AB_330288	Cell Signaling Technology
Rabbit TYMS (thymidylate synthase) polyclonal antibody	1:500	#AP6682b	AB_1968389	Abcepta
Secondary Antibody
Anti-rabbit IgG horseradish peroxidase- conjugated secondary polyclonal antibody	1: 2000	#7074	AB_2099233	Cell Signaling Technology

4.8. Assessment of Dihydrofolate Reductase (DHFR) Inhibition

DHFR inhibition screening for the tested compounds GE23, CS18, CS23, KA44, MV16, 5-FU, GEM, and MTX was performed in a cell-free DHFR enzymatic activity colorimetric assay (AssayGenie, Dublin, Ireland; Catalog #BN00512). MTX was included as a positive control and reference drug compound. The testing and DHFR inhibition screening were performed according to the manufacturer’s instructions. The DHFR activity was monitored by the reduction in absorbance reading at OD340 nm, while potential inhibitors arrest this decrease according to the following general reactions:

$$\mathrm{Dihydrofolate}+\mathrm{NADPH}+\mathrm{H}+\stackrel{\mathit{D}\mathit{H}\mathit{F}\mathit{R}}{\to }\mathrm{Tetrahydrofolate}+\mathrm{NADP}+(\mathrm{decrease}\mathrm{in}\mathrm{OD}340\mathrm{nm})$$

$$\mathrm{Dihydrofolate}+\mathrm{NADPH}+\mathrm{H}+\underset{\mathit{I}\mathit{n}\mathit{h}\mathit{i}\mathit{b}\mathit{i}\mathit{t}\mathit{o}\mathit{r}}{\stackrel{\mathit{D}\mathit{H}\mathit{F}\mathit{R}}{\to }}\mathrm{Tetrahydrofolate}+\mathrm{NADP}+(\mathrm{no}\mathrm{decrease}\mathrm{in}\mathrm{OD}340\mathrm{nm})$$

All compounds were tested at a concentration range of 5–500 nM. Briefly, methotrexate (MTX) and screening compounds were diluted 100-fold in DHFR assay buffer at the desired final concentrations, and 2 μL was added into wells in a 96-well clear ELISA plate with flat bottom, assigned as Sample Screening (S), Enzyme Control (EC), and Inhibitor Control (IC), respectively. A 400-fold diluted DHFR in DHFR assay buffer, enough for the number of wells to be analyzed, was prepared. A total of 98 µL of diluted DHFR was added into the desired well(s) containing the test samples, EC, and IC, with a partial volume of 100 µL. Then, a 100 µL DHFR assay buffer was added to the desired well(s), as the Background Control. A 40-fold dilution of NADPH stock solution was prepared, vortexed briefly, and kept on ice. A total of 40 µL of diluted NADPH was added to each well containing the test samples, EC, IC, and Background Control, and mixed well. Incubation at room temperature for 15 min followed. Exposure to light was avoided. A 15-fold dilution of DHFR substrate in DHFR assay buffer was prepared, vortexed briefly, and kept on ice. A total of 60 µL of diluted DHFR substrate was added to each well containing the test samples, EC, IC, and Background Control, and mixed well. The total volume was 200 µL. Absorbance at 340 nm was measured in kinetic mode for 15 min at room temperature immediately in an ELISA reader. Two time points (t1 = 4 min and t2 = 10 min) in the linear range of the plot were chosen, and the corresponding values for the absorbance (OD1 and OD2) were observed.

The slope for all test inhibitor samples [S] and [EC] was calculated by dividing the net ∆OD (A1 − A2) values by the time ∆t (t2 − t1). The Solvent Control or Inhibitor Background Control readings from its paired sample readings were subtracted. DHFR % relative activity and inhibition induced by the treatment of the test compounds were calculated by the following equations, and the respective curves were demonstrated.

$$\mathrm{Relative}\mathrm{Inhibition}(\%)={\displaystyle \frac{Slope of \left[EC\right]-Slope of \left[S\right]}{Slope of EC}}\times 100$$

$$\mathrm{Relative}\mathrm{Activity}(\%)={\displaystyle \frac{Slope of \left[S\right]}{Slope of EC}}\times 100$$

4.9. Molecular Docking Studies

All synthesized conjugates were evaluated in silico for their binding interactions with human thymidylate synthase (TS, also referred to as TYSY; PDB ID: 1JU6) and dihydrofolate reductase (DHFR; PDB ID: 1KMS). The selected crystal structures were retrieved from the RCSB Protein Data Bank based on their high resolution and biological relevance to antimetabolite binding.

Prior to docking, protein structures were prepared by removal of crystallographic water molecules and co-crystallized ligands, followed by addition of polar hydrogen atoms and assignment of Kollman charges. Ligand structures were energy-minimized and converted to the appropriate format using PyRx 0.8 software. Docking simulations were performed using AutoDock Vina (version 1.1.2), with grid boxes defined to fully encompass the catalytic binding pockets as determined by the coordinates of known active-site residues and reference inhibitors [40,41].

Docking parameters were kept constant across all ligands to allow comparative evaluation of binding affinities within the same enzymatic system. For each compound, multiple binding poses were generated, and the lowest-energy conformation, as ranked by the Vina scoring function, was selected for further analysis. Protein–ligand interactions were visualized and analyzed using BIOVIA Discovery Studio Visualizer (v24), focusing on hydrogen bonding, hydrophobic interactions, and π–π or π–alkyl contacts within the active site [47].

Ligand Preparation: Ligands were sketched in ChemDraw Professional 15 and energy-minimized using the Universal Force Field (UFF) to reach a local minimum. All ligands were then converted to PDBQT format, ensuring all rotatable bonds were defined.

Receptor Preparation: The crystal structure of Thymidylate Synthase PBD 1JU6 was obtained from the RCSB Protein Data Bank. All water molecules and co-crystallized ligands were removed. Polar hydrogen atoms were added, and Kollman charges were assigned using AutoDockTools.

The active site of the enzyme was identified according to the co-crystallized inhibitor present in the 1JU6 structure and verified through the localization of key amino acid residues known to participate in ligand binding. The catalytic pocket is characterized by three hydrogen bonds involving Asn226 and Ser216, as well as two cation–π interactions with Phe225 and Leu221, which contribute to the stabilization of the ligand–receptor complex [36].

To define the docking region, a grid box was generated with the following parameters:

Center (Å): X = 18.9714, Y = 4.6788, Z = 17.9478

Dimensions (Å): X = 25.3507, Y = 19.4917, Z = 17.9478

The grid field was designed to encompass the entire active cavity of the enzyme, ensuring accurate identification of the most favorable binding poses of the tested compounds.

To gain a deeper understanding of the interactions between the target molecules and thymidylate synthase, molecular docking studies were performed using AutoDock Vina.

The “exhaustiveness” parameter was set to 128 to ensure a thorough search of the conformational space. The docking results were ranked based on their binding affinity scores (kcal/mol).

Molecular docking simulations were also performed using the crystallographic structure of human DHFR (PDB: 1KMS) to elucidate the binding interactions and predict inhibitory potential of the synthesized steroidal lactam derivatives

The active site, comprising key residues Ile7, Ala9, Glu30, Phe31, Phe34, Ile60, Pro61, and Val115, was enclosed within a grid box with the following parameters:

Center (Å): X = 13.04, Y = 20.98, Z = 35.77

Dimensions (Å): 23.97 × 20.61 × 22.74

To validate and assess the reliability of the docking protocol, the co-crystallized dUMP was first removed and then re-docked into the binding site of thymidylate synthase (PDB ID: 1JU6) (Figure 5A). Similarly, SRI-9439 was used in dihydrofolate reductase (PDB: 1KMS) (Figure 6A) [37].

It should be noted that molecular docking provides a predictive, static approximation of ligand–enzyme interactions and does not account for protein flexibility, solvent effects, or intracellular metabolism. Therefore, docking scores were interpreted comparatively rather than as absolute predictors of inhibitory potency and were used to support structure–activity relationship analysis in conjunction with experimental enzymatic and cellular data.

4.10. Statistical Analysis

Student’s t-test and Pearson’s correlation coefficient (r) test were used to compare the level of significance between the experimental groups. Differences with p-values less than 0.05 were considered significant. Microsoft Excel (version 16.19, Microsoft Hellas, Athens, Greece) and Origin(Pro), Version 2016 (OriginLab Corporation, Northampton, MA, USA), were used for analysis and figures.

5. Conclusions

Ultimately, these findings position homo-aza steroidal antimetabolites as lead scaffolds for next-generation multi-target cytostatics, bridging classical antimetabolite pharmacology and modern structure-based design. Among the series, CS23 emerged as the most balanced and potent hybrid, combining strong binding affinity to both TS (−11.2 kcal/mol) and DHFR (−11.5 kcal/mol) with pronounced induction of apoptosis and S-phase arrest. The statistically significant correlations between docking energies, biochemical inhibition (r = 0.96), and cytostatic activity (r = −0.81 for combined TS/DHFR scores) establish, for the first time, a predictive continuum between in silico affinity and in vitro efficacy. These findings validate the dual-target design principle as an effective strategy to overcome the metabolic plasticity that underlies resistance to single-pathway antimetabolites such as 5-FU and methotrexate.

In conclusion, this work not only introduces a promising class of dual-action antimetabolites but also provides a robust methodological blueprint—linking computational chemistry, enzymology, and cellular pharmacodynamics—for the rational design of future anticancer therapeutics aimed at translational precision and resistance circumvention.

Abbreviations

The following abbreviations are used in this manuscript:

5-FU	5-Fluorouracil
ATCC	American Type Culture Collection
Ara-C	Cytarabine
BRAF	B-Raf proto-oncogene
BRCA1	Breast cancer gene 1
CO2	Carbon dioxide
Ct24	Control cell growth at 24 h
Ct72	Control cell growth at 72 h
DBS	Docking binding score
DHFR	Dihydrofolate reductase
DMAP	4-Dimethylaminopyridine
DMEM	Dulbecco’s Modified Eagle Medium
DMSO	Dimethyl sulfoxide
DNA	Deoxyribonucleic acid
dTMP	2′-Deoxythymidine-5′-monophosphate
dTTP	2′-Deoxythymidine-5′-triphosphate
dUMP	2′-Deoxyuridine-5′-monophosphate
EDCI	Ethyl-3-(3-dimethylaminopropyl)carbodiimide
EGF	Epidermal growth factor
ELISA	Enzyme-linked immunosorbent assay
EMEM	Eagle’s Minimum Essential Medium
FACS	Fluorescence-activated cell sorting
FBS	Fetal bovine serum
FdUMP	5-Fluoro-2′-deoxyuridine-5′-monophosphate
FGFR	Fibroblast growth factor receptor
FITC	Fluorescein isothiocyanate
GI50	Concentration causing 50% growth inhibition
GEM	Gemcitabine
HEPES	4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid
IC50	Half-maximal inhibitory concentration
KRAS	Kirsten rat sarcoma viral oncogene homolog
LC-MS	Liquid chromatography–mass spectrometry
MAPK	Mitogen-activated protein kinase
MTT	3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide
MTX	Methotrexate
NADPH	Nicotinamide adenine dinucleotide phosphate (reduced form)
NGS	Next-generation sequencing
NIS	N-Iodosuccinimide
NMR	Nuclear magnetic resonance
OD	Optical density
PBS	Phosphate-buffered saline
PDB	Protein Data Bank
PI	Propidium iodide
PI3K	Phosphoinositide 3-kinase
PK	Pharmacokinetics
RMSD	Root mean square deviation
RNA	Ribonucleic acid
RPMI	Roswell Park Memorial Institute medium
SAR	Structure–activity relationship
SG	Sub-G1 phase
SDS-PAGE	Sodium dodecyl sulfate–polyacrylamide gel electrophoresis
SOCl2	Thionyl chloride
TGI	Total growth inhibition
THF	Tetrahydrofolate
TP53	Tumor protein p53
TS	Thymidylate synthase
TYSY	Thymidylate synthase (gene/protein notation)
wt	Wild type

Supplementary Materials

The following are available online at https://www.mdpi.com/article/10.3390/ph19020331/s1, Supplementary Material S1: Synthesis of steroidal conjugates; Figure S1: ¹H-NMR and 13C-NMR spectra for 12; Figure S2: ¹H-NMR and 13C-NMR spectra for KA44; Figure S3: LC/ESI-MS analysis for KA44; Figure S4: ¹H-NMR and 13C-NMR spectra for GE23; Figure S5: LC/ESI-MS analysis for GE23; Figure S6: ¹H-NMR and 13C-NMR spectra for CS18; Figure S7: LC/ESI-MS analysis for CS18; Figure S8: ¹H-NMR and 13C-NMR spectra for CS23; Figure S9: LC/ESI-MS analysis for CS23; Figure S10: ¹H-NMR and 13C-NMR spectra for MV16; Figure S11: LC/ESI-MS analysis for MV16.

Author Contributions

Conceptualization, D.T.T. and K.E.A.; synthesis of molecules, V.S., M.V., K.A. and C.S.; methodology, D.T.T. and K.E.A.; software, D.T.T., K.E.A., P.D. and V.S.; validation, K.E.A., D.T.T., P.D. and V.S.; formal analysis, K.A., D.T.T. and P.D.; investigation, K.E.A., D.T.T., P.D., I.A.A., and V.S.; resources, D.T.T. and K.E.A.; data curation, D.T.T., K.A., P.D., V.S., K.A. and M.V.; writing—original draft preparation, K.E.A. and D.T.T.; writing—review and editing, K.E.A., D.T.T., V.S., K.O., M.P., M.D., I.A.A. and S.S.; visualization, K.E.A., D.T.T., V.S., P.D. and N.S.; supervision, D.T.T.; project administration, D.T.T.; funding acquisition, D.T.T., K.E.A. and V.S. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

The original contributions presented in this study are included in the article and Supplementary Materials. Further inquiries can be directed to the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This research received no external funding. APC was funded by K.E.A.