The soil used in this study was collected from an oil palm plantation field located at Banting Estate, Malaysia (2°47′3″N, 101°27′55″E). According to USDA (United States Department of Agriculture) soil taxonomy, the soil was classified as a loamy clay texture (26% sand, 31% silt, and 42% clay). After mixing, a portion of the collected soil samples was autoclaved thrice (at 121°C, 15 psi, 45 min for each cycle) to generate an initially pre-sterilized soil. The soil kept in the sealed bags was allowed to cool overnight between each set of autoclaving processes. Pre-sterilized soils were used in order to eliminate microbial communities indigenous to the soil. A complete elimination was confirmed by the absence of colonies on the nutrient agar after 48 h of incubation using spread plate technique.

A total of 80 germinated oil palm commercial seeds (Deli Dura × AVROS Pisifera) were obtained from the Oil Palm Breeding Unit, Sime Darby Plantation Research Sdn. Bhd., Malaysia. These full-sib seeds were derived from a fresh fruit bunch to achieve genetic uniformity. The seeds that qualify for sowing must have a radicle and plumule of approximately 1 cm in length. Each seed was sowed in a polybag (25 cm × 30 cm) filled with an equal weight of either normal or pre-sterilized soils. Seedlings were subjected to four treatments including (i) fertilized normal soil (+FN), (ii) unfertilized normal soil (−FN), (iii) fertilized sterilized soil (+FS), and (iv) unfertilized sterilized soil (−FS), in a complete randomized designed trial with 20 replicates per treatment. The abbreviations +F referred to the polybags applied with fertilizer, while −F referred to the fertilizer-free polybags, either using natural normal soil (N) or pre-sterilized soil (S). Under the normal commercial practice, oil palm seedlings were raised in the main nursery for 12 months, supplemented with compound fertilizer. Hence, the +FN control was served as a positive control to compare against other treatments in this study. The seedlings were raised in an open space nursery owned by Sime Darby Plantation near Banting. Each polybag was placed on top of a concrete slab to prevent the roots from penetrating the earth. After 6 weeks of sowing, the granular compound fertilizer N-P-K-Mg (14:13:9:2.5:trace elements, CCM Fertilizers Sdn. Bhd., Malaysia) was introduced into these seedlings according to the fertilizer schedule and dosage stated in the Sime Darby Plantation’s standard agriculture reference manual (unpublished). The complete fertilizer application throughout the 8 month trial was 100 g of granular NPK fertilizer per polybag. Upon the treatment effects were observed on these palms, a destructive sampling was conducted using 10 randomly selected seedlings from each treatment. Plant height, the number of developed fronds and total dry biomass were recorded accordingly.

In each treatment, five randomly selected seedlings were subjected to soil–root washing to segregate the rhizocompartment. Each oil palm seedling was carefully excavated from the polybag. Loose bulk soil from each seedling was collected and preserved individually at −80°C. The exposed root was vigorously shaken to remove the loosely adhering soil for collecting the rhizosphere microbiome. Soil bound firmly on the root was treated as rhizoplane. Rhizosphere and rhizoplane pellets from the same seedling were obtained by repeated chemical washings and centrifugations as described by Wieland et al. (2001). In total, we performed five replications across three compartments (bulk soil, the rhizosphere and rhizoplane) for +FN control, −FN, +FS, and −FS treatments.

Total genomic DNA extractions for each sample were performed using FastDNA™ Spin Kit for Soil (MP Biomedical, USA) following the manufacturer’s protocol. The quality and quantity of the extracted DNA were measured using agarose gel electrophoresis and Qubit spectrometer (Invitrogen, USA). The V3-V4 hypervariable regions of the prokaryotic 16S rRNA gene were amplified using primer 341F and 806R (Bates et al., 2011). Five replicates of all treatments except the +FN control in rhizoplane (only four replicates due to a problem with one sample) were sequenced individually using paired-end (2 bp × 250 bp) protocol on an Illumina HiSeq platform (Illumina, USA). This has accounted to a total of 59 samples.

Paired-end reads merging, assigning, trimming, and quality filtering were performed using DADA2 pipeline v1.8, from an open source R package (Callahan et al., 2016). All the parameters set were by default. Primer-trimmed sequences with more than 200 bp without ambiguous base calls were quality filtered using the “filterAndTrim” function in the same pipeline. Subsequently, the estimated error rate was learned and calculated using the machine learning algorithm implemented in DADA2 from the subsampled reads. Forward and reverse sequences were merged to obtain amplicon sequence variant (ASV). Chimeric sequences were eliminated, resulting in a sequence length ranging from 400 to 450 nucleotides. Taxonomic classification of each ASV was annotated using SILVA reference database (SILVA 138 release) (Quast et al., 2013). After removing chloroplast and mitochondria, we further removed trace sequences of microbial read singleton, doubleton, and retained only ASVs with more than 10 reads in at least 60% of samples (n = 5), for each treatment independently. Sequencing data were rarefied to the lowest number of reads observed in a single sample, that is 23,124 reads.

Statistical analyses were conducted using R version 4.2.1 (R Core Team, 2020). To compare between treatments, significant differences in the soil, plant nutrient and alpha diversity were evaluated with non-parametric Kruskal–Wallis test and followed by post hoc comparisons conducted in pairwise Wilcoxon tests (p-value < 0.05). The alpha–diversity were represented by Chao1 richness (Chao, 1984) and Shannon index (Shannon, 1948). The treatment effect on the overall soil–root microbiome composition was assessed using permutational analysis of variance (PERMANOVA) with the “adonis2” function, and the treatment centroids was calculated using betadisper, both from the community ecology R “vegan” package (Oksanen et al., 2019). To visualize the microbial composition differences, unconstrained principal coordinate analysis (PCoA) was analyzed using the R package “phyloseq” (Paul and Susan, 2013) and plotted with “ggplot2” function (Wickham, 2016). We performed a differential abundance DESeq2 analysis (Love et al., 2014) to identify the ASVs that were significantly varied in rhizocompartments across treatments by comparing to +FN control. The significant value was reported at adjusted p-value < 0.01 with abundance in log2 fold change greater than 2. Heatmap of the relative abundance of archaea was generated by R “pheatmap” package (Kolde, 2019). The resulting sequencing data are publicly available at NCBI SRA under Bioproject PRJNA818730.

With the same subset of 59 samples, we performed soil and plant chemical analyses. Bulk soils were collected for physicochemical analysis. The soil samples were oven-dried at 70°C, ground and subsequently sieved to pass through a 5 mm sieve. These soil samples were then analyzed for pH, carbon/nitrogen ratio (C/N ratio), total nitrogen (TN), total phosphorus (TP), available phosphorus (AP), exchangeable cations included potassium (K), calcium (Ca), magnesium (Mg) and electrical conductivity (EC). The plant samples were also analyzed separately with TN, TP, K, Mg and Ca. Using an elemental analyser, soil and plant samples were analyzed for total C and N by the dry combustion method (Matejovic, 1993). Total phosphorus was determined by the vanadate-molybdate method (Tandon et al., 1968) using a spectrophotometer at 425 nm. Cations including K, Ca, and Mg were extracted using 1 M ammonium acetate (NH₄OAc) calibrated at pH 7 followed by inductively coupled plasma (Perkin Elmer, USA) (David, 1960). The EC of the soil was measured by mixing soil and water at a 1:2 ratio (w/v), and then using a conductivity meter (Mettler Toledo, USA) to determine the EC of the resulting mixture (Heiniger et al., 2003). All the analyses were completed by Espek Research and Advisory Services (ERAS) laboratory, based on the M.S 678 (soil) method outlined by the Standard and Industrial Research Institute of Malaysia (Malaysian Standard [MS], 1980).

Seedlings grown in normal soils (+FN and −FN) appeared to be healthy after 8 months of planting; however, seedlings grown in pre-sterilized soils (+FS and −FS) were stunted despite fertilizer being applied to the +FS treatment (Figure 1 and Supplementary Figure 1). Distinctive plant height, the number of leaves and total biomass were observed for each treatment (Figures 2A–C). The control +FN seedlings were significantly taller and heavier, with a mean of 94.5 cm and 590 g, as compared with seedlings grown in normal unfertilized soil (−FN) with a mean of 48.9 cm and 148 g (p < 0.001). Seedlings in both +FN and −FN had an average leaf count of 13 and 10, respectively, with lanceolate, bifurcate and pinnae observed and recorded. In general, seedlings that were grown in the −FN treatment displayed slower growth compared to the +FN control. On the other hand, seedlings that were grown in the pre-sterilized soils (−FS and +FS) had stunted growth with a short plant height, remaining within the range of 18–33 cm, and an average total biomass of 20 g. Furthermore, the leaf development of these seedlings stopped at month five after sowing, without any pinnae observed.

In order to ensure autoclaving did not alter the soil chemistry, the pH and nutrient contents of the soil were measured before and after autoclaving. As shown in Table 1, the pH, macronutrient and micronutrient content in normal and pre-sterilized soils did not differ significantly (p > 0.05). After 8 months, the effect of fertilizer application on the soil nutrient content was remarkable, where TP, AP, K and Mg showed at least a two-fold increase with the treatment +FN and +FS treatments compared to the unfertilized counterpart soils of −FN and −FS (Table 1). Moreover, the chemical fertilizer raised the EC slightly to 75 μS/cm in the +FN control, whereas a sharp increase in the EC of 336 μS/cm took place in the +FS treatment.

The effect of the fertilizer treatment on the plant nutrient content was examined after 8 months of planting (Table 2). As a result of the low plant biomass obtained with the pre-sterilized soil treatment, plant nutrient analysis was not possible for the +FS and −FS treatments. Therefore, only the +FN and −FN treatments were analyzed in this study. The TN content was not significantly different between these two treatments; however, the parameters of TP, K, Mg and Ca content were significantly lower in the −FN treatment.

In this study, both +FS- and −FS-treated oil palm seedlings without pinnae suffered severely, whereas −FN seedlings grew slower than +FN control seedlings. Interestingly, despite similar environmental conditions and the same amount of NPK granules being supplied, palm growth in +FS was markedly different from that in +FN. Our results showed that the autoclave did not significantly change the soil macronutrient content. However, a few studies suggested that heat sterilization increased soil nutrient availability (Mahmood et al., 2014; Oruru et al., 2018; Li et al., 2019). As reported in the oil palm soil fertility classification guidelines by Goh and Rolf (2003), the soil macronutrients (NPK concentration) applied in our study were at the optimum level for oil palm growth before and after 8 months of planting. Therefore, the nutrient availability of N, P and K was not the limiting factor causing stunted growth in both +FS and −FS treatments. We observed a high accumulation of soil nutrient content such as K, Ca and Mg cations, resulting from the release of compound fertilizer in the +FS treatment relative to the +FN control. This was evidenced by the two-fold increase in the soil EC value in the +FS treatment. A high EC value can affect plant nutrient availability and soil microorganism activity. A previous study showed that osmotic pressure primarily caused by excess K, Mg, Ca, Na and Cl ions, may affect microbial activity or even suppress the growth of certain microbes by drying or lysing their cells (Yan et al., 2015). Although the sterilization process eliminated indigenous microbes from the soil, living microbes re-established themselves in polybags during seedling cultivation in the pre-sterilized soil. However, the microbial diversity in the +FS treatment remained low compared to the +FN control. The acquisition of this new microbial community in pre-sterilized soil was a random natural incidence which may result from irrigation, rain, human activity or insect activity. The results suggest that the stunted palm growth in the +FS treatment could be due to this excessive accumulation of soluble salts in the root zone that hindered the plant roots from withdrawing water from the surrounding soil (Sogoni et al., 2021), thus limiting plant nutrient uptake and affecting the microbial assemblage around the root zone at the end of the planting.

Salts and nutrients released from compound fertilizer were readily absorbed by the plant, as evidenced by the high amount of total dry biomass and optimal soil K, Ca and Mg cations in the +FN control. Although compound fertilizer had a significant impact on plant growth, PCoA and taxonomy plots revealed that the microbiota in the normal soil +FN control and −FN treatment differed slightly. This is similar to the results previously reported in woodland soil (Karki et al., 2021), where a low pH and low nutrient status in −FN favored the growth of oligotrophs such as Acidobacteriota, whereas fertilizer inputs increased the relative abundance of copiotroph in the +FN control, such as Bacteroidota and Pseudomonadota (formerly known as Proteobacteria), which prefer nutrient-rich conditions for growth. Copiotrophic bacteria are specialized in degrading complex polysaccharides (Spring et al., 2015). Nakayama et al. (2021) reported that microbial species of Pseudomonadota and Bacteroidota that shared a similar niche in forest soil could perform net nitrogen nitrification and mineralisation. Similarly, the phosphorus (P) applied via fertilization was quickly fixed into a stable form in the soil, requiring the help of phosphate solubilising bacteria to release the fixed P into a plant usable form. In this study, the +FN control was found to have high abundances of bacteria from the genera Rhizobium and Enterobacter, which may function as potential P solubilisers. Altogether, this suggests that the fertilizer provision in the +FN control altered the function of indigenous soil microbes toward improved nutrient acquisition or nutrient transformation for better plant growth. However, further validation is needed by checking the function of these native soil microbes.

By the end of the study, microorganisms had built up in the pre-sterilized soil. Despite this, PCoA and PERMANOVA tests revealed a newly established soil microbiota in the −FS and +FS treatments that were distinct from the original indigenous microbiota found in −FN and +FN. Soil sterilization, regardless of the presence of fertilizer, also caused a reduction in microbial diversity across the bulk soil and rhizocompartments. Pre-sterilization may have reduced the number of native soil microbes, which may have influenced the stunted seedling growth observed in the palms under the −FS and +FS treatments. Several studies have shown that soil sterilization can eliminate soil-borne pathogens and promote changes in soil microbe communities that lead to improved plant growth and increased beneficial microbes (Mahmood et al., 2014; Zhang et al., 2016; Li et al., 2019). However, in our experiment, the soil used did not contain Ganoderma boninense, a fungal pathogen that infects oil palm trees in South East Asia (Jeffery Daim et al., 2015) (ITS data not shown). The benefits of soil sterilization, specifically the elimination of soil-borne pathogens, were therefore not applicable to this experiment.

A study by Oruru et al. (2018) on cowpeas found that plants grown in non-sterilized soil performed better than those that were grown in pre-sterilized soil. The findings suggest that early root colonization of mycorrhizae or beneficial microbes from the bulk soil is critical for maintaining plant and soil health. Therefore, we speculate that the pre-sterilized soil in our experiment may have eliminated functional microbes that were necessary for colonizing oil palm seedlings’ roots during the early stages of growth and development (Figure 8). This is evidenced by a high number of depleted bulk soil ASVs in the −FS (599) and +FS (283) treatments compared to the −FN treatment (31) (Table 3). Despite the fact that plant roots can regulate the diversity and abundance of the rhizospheric soil microbiome via root exudate secretion, the depletion of microbial sources in the pre-sterilized bulk soil system can disrupt the microbial network and cause some degree of stress to the plant. As a result, the plant may not be able to maintain its health through rhizosphere recruitment. This finding has also been reported in various crop studies where changes in the composition of the bulk soil microbiome significantly affected the root-associated rhizosphere and rhizoplane microbial communities (Edwards et al., 2015; Liu et al., 2019; Zhang et al., 2019; Han et al., 2020).

A significant number of enriched ASVs were found in the soil–root compartments of both the −FS and +FS treatments (Table 3). Among the enriched ASVs were Rhodanobacter, Roseiarcus, Chujaibacter, Acidibacter, Alicyclobacillus, Bryobacter, and Occallatibacter. None of these are pathogenic toward plants. The enrichment of these species may be attributed to the “priority effect,” which occurs when an early-arriving microbial community competes for space and resources at the plant roots, thereby preventing the establishment of a late-arriving community (Debray et al., 2022). Similarly, the −FN treatment did not detect any of the plant pathogenic bacteria that cause disease in economically important crops (Mansfield et al., 2012), including Pseudomonas syringae pathovars, Ralstonia solanacearum, Agrobacterium tumefaciens, Xanthomonas oryzae pv. oryzae, and Erwinia amylovora.

In addition to the bacterial community, soil archaeal communities were significantly affected in all of the treatments. Crenarchaeota, and Thermoplasmatota were the predominant archaeal phyla in the −FN and +FN treatments. These phyla are associated with nitrogen cycling (Nitrosopumilaceae, Nitrososphaeraceae, and Nitrosotaleaceae) (Wan et al., 2021), and carbon metabolism (Marine Group I) (Swan et al., 2014). Intriguingly, a noticeable reduction in Nitrosotaleaceae (Candidatus Nitrosotalea) and Nitrososphaeraceae was observed in pre-sterilized rhizosphere soil. This may be due to competitive interactions with bacterial nitrogen nitrifiers, particularly since Nitrospira abundance was approximately four times higher in the +FS and −FS treatments. While some clades in archaeal Crenarchaeota contribute significantly to nitrification and carbon cycling, we cannot conclude that enriched or depleted archaeal-associated ASVs are exclusively related to the health of oil palm seedlings.