Acta Physiologiae Plantarum

, 39:18

Recombinant pharmaceutical protein production in plants: unraveling the therapeutic potential of molecular pharming

  • Vijaya R. Dirisala
  • Rahul R. Nair
  • Krupanidhi Srirama
  • Prakash Narayana Reddy
  • K. R. S. Sambasiva Rao
  • N. Satya Sampath Kumar
  • Giridhar Parvatam

DOI: 10.1007/s11738-016-2315-3

Cite this article as:
Dirisala, V.R., Nair, R.R., Srirama, K. et al. Acta Physiol Plant (2017) 39: 18. doi:10.1007/s11738-016-2315-3


There is an increasing demand for the generation of recombinant pharmaceutical proteins for a wide array of therapeutic applications. In comparison to bacterial, yeast and animal cells, the production of recombinant proteins in plants with economic and therapeutic importance has only started recently. The most important prerequisite of any expression systems is that it should be simple and inexpensive. In this regard, plant-based expression has emerged an as accepted alternative to conventional expression platforms due to economic feasibility, rapid scalability, higher stability of recombinant proteins, safety due to lack of harmful substances (human, animal pathogens and pyrogens) and capability of producing proteins with desired secondary modifications. Heterologous expression using plants has played a pivotal role in the development of a myriad of recombinant proteins, including neutraceuticals and monoclonal antibodies being utilized in various therapeutic approaches. This paper presents an overview about the current status, various strategies and advantages of pharmaceutical protein production in plant expression systems. We also present a summary of expression of therapeutic monoclonal antibodies, vaccines, clinical trials and the regulatory aspects of plant-based expression. Furthermore, the challenges encountered in plant expression such as costs associated with existing purification strategies are discussed.


Therapeutic proteins Heterologous expression Transgenic plants Edible vaccines Clinical trials 



Antibiotic-associated diarrhea


Monoclonal antibody


Single chain variable fragment


Plant-made pharmaceuticals


Therapeutic protein production in plants has proven as an attractive alternative over other expression systems such as bacterial, mammalian cell culture, yeast and transgenic animals. The pharmaceutical industries have capitalized on generating recombinant proteins using the above-mentioned expression systems, and each format has its own advantages. However, there are also some major disadvantages associated with these expression systems (Egelkrout et al. 2012). These impediments make heterologous expression in transgenic plants and plant cell cultures an extremely attractive alternative (Tusé et al. 2014). Plants perform a majority of the post-translational modifications required for the activity of complex eukaryotic proteins (Webster and Thomas 2012). Protein expression in planta provides tremendous flexibility in terms of cost, biocontainment, scalability and regulatory issues (Fischer et al. 2013). The successful launch of the first recombinant protein and antibody using a plant platform almost 25 years ago paved the way for an increase in the number of plant-produced protein therapeutics. These are expected to replace the therapeutic proteins and antibodies produced using conventional expression systems when patent protection is lost 2017 (Altman and Hasegawa 2011).

Overview of recombinant protein production in plants

The model plant Nicotiana tabacum and N. benthamiana (tobacco), along with other host plants such as Medicago sativa (alfalfa), Musa paradisiaca (banana), Daucus carota (carrot), Lactuca sativa (lettuce), Zea mays (maize), Solanum tuberosum (potato), Oryza sativa (rice), Glycine max (soybean), Solanum lycopersicum (tomato), and Triticum aestivum (wheat), has become the preferred choice for the production of a wide range of therapeutic proteins (Sack et al. 2015). Protein production can be achieved in the whole plant or in a specific portion of a plant such as leaf, fruit and seed (Merlin et al. 2014). By default, protein accumulation in plants takes place in the cytoplasm, which is not generally preferred due to its high level of protease activity. To circumvent this issue, proteins are targeted to the chloroplast (Scotti and Cardi 2014). Proteins from the cell can also be extracted through guttation fluid, as demonstrated in the case of tobacco (Komarnytsky et al. 2000).

Strategies for heterologous protein expression in plants

Production of recombinant proteins from plants is a multi-step process. Initially, it involves the optimization of the coding sequence of the exogenous gene prior to introduction into the host. This is followed by the creation of the expression construct containing the gene of interest and its integration into the plant genome (stable or transient), the propagation of plants expressing the desired protein and the purification of the desired protein (Egelkrout et al. 2012). Stable transformants can be obtained by introducing exogenous DNA into the nuclear or the plastid genome of plant cells. This is achieved via vehicles such as A. tumefaciens (Leuzinger et al. 2013), or by particle bombardment (Chen and Lai 2014). Stable transformation into the chloroplast genome allows the transgenes to be precisely targeted to a specific locus, avoiding accidental gene knockouts and ensuring that foreign DNA does not disseminate into the environment via pollen (Bock 2015). A noted disadvantage of the chloroplast genome is that it lacks suitable machinery to perform typical eukaryotic post-translational modifications such as glycosylation (Rigano et al. 2012). The other strategy for heterologous protein production in plants relies on replicating plant viruses (Hefferon 2013). The small size of viral genomes, along with their ease of manipulation and their infection process, makes them an attractive alternative to stable transgenic systems for rapid transient expression (Gleba et al. 2014). The gene of interest is inserted into viral replicating elements, episomically amplified, and subsequently translated in the plant cell cytosol (Werner et al. 2011). Prior to the existence of magnifection (Klimyuk et al. 2014), transient expression by virus particles was of limited importance due to their low infectivity and their inability to carry larger transgenes. Magnifection technology relies on the transfection efficacy of A. tumefaciens to deliver viral replicons, in combination with the advantages of a high expression yield obtained with viral vectors and post-translational capabilities.

Advantages of plant expression systems for generating heterologous proteins

Production costs

The cost of generating recombinant proteins in plants is approximately 2–10% of the cost of using microbial fermentation systems and 0.1% of the cost of using mammalian cell cultures (Chen et al. 2014). For instance, the production costs of butyrylcholinesterase using a plant platform such as N. benthamiana are approximately 20 times less than that derived from blood. Similarly, the cost per unit of producing cellulase enzymes for the production of ethanol using the plant-based system is reduced by approximately 30% compared to fungal cellulases (Tusé et al. 2014). It was estimated that the total cost of producing 1 g of human glutamic acid decarboxylase (hGAD65) serving as one of the major autoantigens in Type 1 autoimmune diabetes (T1D) could reach 700 Euros using insect cells whereas it costs less than 5 Euros for producing the same quantity of recombinant proteins using plant expression platform using T6 stable tobacco line (Gecchele et al. 2015). Ventria Bioscience is developing plant-produced rh-lactoferrin as the active ingredient in VEN100 for the treatment of antibiotic-associated diarrhea (AAD) and their Phase II studies has shown that incidence of AAD is reduced by 52%. Since AAD-associated healthcare costs are estimated at $1500 per high-risk patient, VEN100 could be considered cost-effective if the costs are brought to half, i.e., less than $750 per patient. Ventria Bioscience is able to bio-manufacture the product at less than $3.75/g and assuming 50% market penetration of 2.8 million U.S. prescriptions a year, 42 metric tons of recombinant protein would be required to satisfy U.S. demand. Manufacturing of rh-lactoferrin at the aforementioned low price and high production capacity is only possible with plant expression system (Broz et al. 2013).

Product authenticity

A majority of eukaryotic proteins expressed in a prokaryotic expression system such as E. coli do not fold correctly and tend to degrade or accumulate as inclusion bodies (Dirisala et al. 2012, 2013, 2015), which may hamper the biological activity of the protein. Proteins of human origin, with the exception of glycosylation, efficiently undergo all kinds of post-translational modification in plants. For therapeutic proteins such as serum albumin, insulin and colony-stimulating factors, glycosylation is not required for product performance (Gleba and Giritch 2011). For other proteins, such as recombinant erythropoietin, a specific glycosylation pattern is required for its performance (Obembe et al. 2011). This distinction is due to the differences in the glycoforms made by the human body. In addition, the glycosylation pattern contributes to the immunogenic potential for some recombinant proteins (Kim et al. 2014). Several changes in the glycosylation pathway are hence required to produce proteins with typical human glycan structures in plants (Bosch et al. 2013). To accomplish this, glycoengineered plants such as N. benthamiana, Physcomitrella and Lemna lacking xylose and fructose have been developed using RNAi technology or by mutagenesis (Decker et al. 2014).

High-level transgene expression

An efficient expression construct plays a major role in obtaining higher yields of recombinant proteins in transgenic plants by enhancing transcription and translation rates. For commercial applications, the chosen plant system should express the protein with high efficiency and they should meet the safety and regulatory requirements (Chen and Lai 2014). Transgenic plants have been utilized for producing therapeutic proteins, edible vaccines, antibodies for immunotherapy and proteins for diagnostics (Thomas and Walmsley 2014). In all these cases, the protein was either purified or the plant tissue was processed to a form that was suitable for topical application or ingestion thus reducing downstream processing. Stable nuclear transformation of antibody coding genes into the plant nuclear genome is the key step in reducing costs and simplifying production processes and is achieved either by Agrobacterium-mediated delivery or biolistics particle delivery system approaches. The recombinant proteins can be targeted to various organelles for standard eukaryotic post-translational modifications (Webster and Thomas 2012). Rapid production of therapeutic proteins is achieved by transient expression using viral coding sequences via A. tumefaciens or by agrofiltration involving infiltration of a suspension of recombinant Agrobacterium into plant tissue (Desai et al. 2010).

To achieve high levels of expression, the strength of key regulatory elements such as the promoter which drives the transcription is very important. Promoters such as the Cauliflower mosaic virus (CaMV) 35S promoter (Xu et al. 2014) or its enhanced version are widely used in dicots. In monocots, however, maize ubiquitin-1 promoter is preferred (Park et al. 2010). Other alternative constitutive promoters include those constructed from a combination of the octopine synthase (OcS) and the mannopine synthase (MaS) promoter sequence, e.g., the (ocs)3mas promoter (Veale et al. 2012). Seed-specific promoters such as fatty acid elongase, maize 27-kDa γ-zein promoter, barley hordein are also used for driving expression of genes (Makhzoum et al. 2013). The rice α-amylase 3D (RAmy3d) promoter is induced by sugar deprivation (Chung et al. 2014). The rate of transcription in cereals is increased by intron-mediated enhancement (Barlett et al. 2009). Vectors such as pEAQ are designed for recombinant protein expression by the Cowpea Mosaic virus hypertranslational expression system, which produces high yields of recombinant protein by its extremely high translational efficiency without the need for viral replication (Peyret and Lomonsoff 2013). Dugdale and his group have reported the In Plant Activation (INPACT) system, which is based on the replication machinery of tobacco yellow dwarf mastrevirus (Dugdale et al. 2013). This system involves transient gene expression from a stably transformed plant, combining the advantages of both means of expression. Optimization of codons in the gene to be cloned and expressed is based on host preference for codons and signal sequences also contribute in driving high-level expression of the genes (Makhzoum et al. 2013).

Very low contamination risks

Recombinant protein products generated using bacterial expression systems are more susceptible to contamination with endotoxins (Chen 2012), which, unlike those of plant-produced recombinant proteins, is more likely to render the final product unsafe for therapeutic use in humans. It is also of note that the health risks arising from contamination with potential human pathogens are minimal (Yusibov et al. 2011). For example, Type I collagen is found in all connective tissues. As such, it is the most abundant protein found in the human body (Gelse et al. 2003). Commercial collagen (pro-collagen) is produced from farm animals such as cows and pigs. It is used in regenerative medicine and has a growing demand globally (Ferreira et al. 2012). Unfortunately, however, commercial collagen produced from farm animals is liable to harbor human pathogens such as viruses or prions (Ferreira et al. 2012). Human recombinant type I pro-collagen has been successfully expressed in tobacco plants (Shoseyov et al. 2014).

Monoclonal antibodies (Mab) generated using plant expression systems

Although Mab produced using mammalian cell culture systems are effective in controlling a wide spectrum of diseases, long manufacturing times along with high production costs have limited their utility (Lahrtz 2015). Production of antibodies using plant-based platforms is a safe and economical alternative to conventional antibody expression (Mortimer et al. 2012). Plant expression systems have been developed that are amenable to the large-scale production of Mab in tightly regulated biocontainment facilities at a lower cost. The first Mab was successfully expressed in tobacco plants in 1989 (Hiatt et al. 1989). Subsequently, a variety of Mab and their derivatives, such as IgA, IgG, and single-chain variable fragments (scFv), have been produced in plants as shown in Table 1. For example, a low-cost plant-based production platform has been developed to produce Mab for rabies post-exposure prophylaxis which would greatly benefit low-income families in developing countries (Dolleweerd et al. 2014). The anti-Salmonella enterica scFv Mab produced in tobacco plants was proven to be effective (Makvandi-Nejad et al. 2005). Mab directed against the protective antigen (PA) of Bacillus anthracis—the causative agent of anthrax—were produced in N. benthamiana (Hull et al. 2005).
Table 1

Monoclonal antibodies generated through plant-based expression systems

Antigen target (Application)

Antibody type

Expression level


Low molecular weight phosphonate ester (P3)

Gamma and kappa immunoglobulin chains

1.3% of total leaf protein

Hiatt et al. (1989)

Conserved epitope of envelope protein of West Nile Virus (WNV)

Humanized anti-west Nile virus mAb


Lai et al. (2010)

Rabies virus

Chimeric antibody of mouse origin with human constant domains

280 mg/kg of fresh leaf weight

Dollyweerd et al. (2014)

Lipopolysaccharide (LPS) of S. enterica Paratyphi B

ScFv antibody

41.7 µg/g leaf tissue

Makvandi-Nejad et al. (2005)

Protective antigen (PA) of Bacillus anthracis

Humanized antibody


Hull et al. (2005)

Ricin toxin

Humanized monoclonal antibodies


Sully et al. (2014)

Rabies virus

mAb SO57 with or without an endoplasmic reticulum (ER)-retention peptide signal

0.4 and 1.2 mg of mAbP and mAbPK per kilogram of fresh leaves

Lee et al. (2013)

Antibodies produced by genetically modified crops (Plantibodies) generated against cell surface antigens of Streptococcus mutans were proven to be effective in reducing tooth decay in humans and other animal models. A recent study has shown that plantibodies generated in a robust plant-based platform neutralized the toxic effect of ricin toxin, which is considered to be a bioterrorism agent (Sully et al. 2014). Lee et al. (2013) reported that reprogramming of plant cell by addition of an ER retention signal to the Mab increased the expression of the rabies antibody which had virus-neutralizing activity comparable to that of native monoclonal antibody.

Production of vaccines

The development of transgenic plant vaccines is accomplished by separating and purifying the foreign protein produced in plant tissue. The antigens isolated and purified from plants can elicit the production of specific antibodies in animals (Piron et al. 2014). Alternatively, the protein can be expressed in the edible part of plants without downstream processing for oral intake (Jain et al. 2013). Using edible plants expressing heterologous antigens avoids the extraction and purification of the recombinant product and the biomolecule is protected by an external membrane for enhanced and prolonged action on mucosal surfaces (Kwon et al. 2013). One advantage of edible vaccines over traditional vaccination methods is their low cost. Other advantages include convenient storage and ease of intake. Using a plant expression system, an oral vaccine was generated for the first time with the ability to target both human immunodeficiency virus (HIV) and hepatitis B (Shchelkunov et al. 2006). When analyzed in experimental animals, this vaccine showed good efficacy.

Clinical trials

Taliglucerase alfa by Protalix Biotherapeutics (Carmiel, Israel) was the first plant-made biological product to be manufactured for treating Gaucher disease. Its subsequent approval by the Food and Drug Administration (FDA), USA, paved the way for other companies to invest in this field (Zimran et al. 2011). Protalix has successfully marketed the product under the brand name Elelyso, using the carrot as a host plant (Aviezer et al. 2009). Influenza vaccines are currently being developed by Medicago Inc. (USA & Canada) by expressing the hemagglutinin protein of H5N1 influenza in Nicotiana plants (Maxmen 2012). Their H5N1 influenza vaccine was reported to be safe and immunogenic when administered intramuscularly to humans. It is currently undergoing phase II clinical trials (Penney et al. 2011).

Another pharmaceutical company, Ventria Bioscience, has acquired the capability to produce commercial quantities of recombinant human albumin, which is being planned for toxicological and phase 1 safety studies (Stoger et al. 2014). The same company has also generated recombinant human lactoferrin and has successfully conducted phase II/III clinical trials (Yemets et al. 2014). Planet Company developed the world’s first clinically tested plantibody, CaroRx, which is specifically targeted towards Streptococcus mutans—the bacterium that causes tooth decay (Yusibov et al. 2011). Another company, Biolex, also announced that it has initiated the sale of its remaining phase III-ready asset, Locteron—a controlled-release interferon for the treatment of hepatitis B and C, produced using duckweed as the host plant (Jansen and Bruijne 2012). For enzyme-replacement therapy, Greenovation has developed two products, alpha-galactosidase and glucocerebrosidase, which are expected to reach clinical trial phases shortly (Paul 2015). Interestingly, the same company has also generated recombinant human keratinocyte growth factor. SemBioSys has produced insulin using safflower as host. This insulin (SBS-1000) has been demonstrated to be the bioequivalent of Humulin (commercially available insulin) (Paul and Ma 2011). The company is proceeding towards phase II clinical trials for this product. Another US-based company, Fraunhofer, has cleared phase I trials for its H1N1 influenza virus subunit vaccine produced in tobacco plants (Cummings et al. 2014). Biolex-manufactured recombinant interferon (IFN-α2b), using duckweed as host plant, is currently awaiting phase II trials (Thomas et al. 2011).

Road blocks

Although the expression of recombinant proteins or vaccines in plants for therapeutic purposes is on the increase (Table 2), there are many problems to be solved prior to commercialization (Sabalza et al. 2014). It is estimated that the purification costs for plant-derived pharmaceuticals to meet regulatory standards account for approximately 80% of their total cost—a significant impediment to commercial production (Fischer et al. 2013; Howard and Hood 2014). Although researchers are coming up with novel purification strategies, such as the fusion of protein-elastin-like polypeptides to facilitate simple purification, or cost-effective approaches such as aqueous two-phase partitioning for removing plant phenolics and alkaloids from plant extracts prior to capture chromatography (Platis and Labrou 2009) and membrane filtration. There is wide scope for efficient methodologies to be developed in this area (Hassouneh et al. 2012). Another constraint is the cost incurred by the setting up of containment facilities (Fischer et al. 2012). Some of the regulatory and biosafety issues concerning production of pharmaceutical proteins in plants are contamination of food and feed chain and risks associated with horizontal gene transfer (Obembe et al. 2011). Prodigene Company Inc has launched large-scale production of trypsin by transgenic maize and United States department of agriculture (USDA) has found that the follow on crops were contaminated as seeds from GE maize had voluntarily sprouted during the following season. Because of this incident the company was charged with hefty fine and it had incurred heavy losses (Fox 2003).
Table 2

List of some of the recombinant human pharmaceutical proteins produced using various plant expression platforms during the past 5 years demonstrating their significance

Pharmaceutical protein

Host plant




Maximum yield



Nicotiana tabacum (tobacco)

Treatment of anemia, rheumatoid arthritis, and other types of anemia

Glyceraldehydes-3-phosphate dehydrogenase (Gapc), Cauliflower mosaic virus (CaMV35S) and Nopaline synthase (NOS)


4 mg/kg TSP

Kittur et al. (2012)

Nicotiana benthamiana



85 mg/kg FW

Jez et al. (2013)

Insulin-like growth factor

Arabidopsis thaliana

Neuropathy, aging and cancer treatment



0.17% of TSP

Li et al. (2011)

Interleukin-10 (IL-10)

Oryza sativa (rice)

Immunoregulation and inflammation

Glutelin B1 (GluB1)


50 mg/kg FW

Fujiwara et al. (2010)

Nicotiana tabacum (tobacco)

CaMV 35S

Tobacco cell suspensions

762 mg/kg FW

Kaldis et al. (2013)

Interferon alfa 2b (IFN-α2b)

Nicotiana excelsior

Treatment of Hepatitis C, Herpetic infection

CaMV 35S


1.3 ± 0.6 × 102 IU/g FW

Sindarovska et al. (2010)

Daucus carota (carrot)

CaMV 35S

Tap root

16.5 × 103 IU/g FW

Luchakivskaya et al. (2011)


50 × 103 IU/g FW

Apolipoprotein A-I (Milano)

Carthamus tinctorius (saffflower)

Treatment of vascular diseases



7.3 g/kg

Nykiforuk et al. (2011)

Nicotiana tabacum (tobacco)

CaMV 35S


5 mg/kg FW

Chiaiese et al. (2011)

Gastric lipase

Nicotiana benthamiana

Treatment of cystic fibrosis, pancreatitis

Cowpea mosaic virus (CPMV)


0.5 mg/g of infiltrated tissue

Vardakoua et al. (2012)


Nicotiana benthamiana

Targeted towards malaria infection

CaMV 35S


12 mg/kg FW

Feller et al. (2013)


Oryza sativa (rice)

Drug carrier used in increasing drugs therapeutic index



10 g/kg dry weight

Zhang et al. (2010)


Nicotiana benthamiana

Regulation of immune responses, inflammation and hematopoiesis

CaMV 35S


7% of TSP

Nausch et al. (2012)

Human epidermal growth factor

Nicotiana benthamiana

Treatment of chronic diabetes ulcers

CaMV 35S


6.24% of TSP

Thomas and Walmsley (2014)

TSP total soluble protein, FW fresh weight

Future directions

Since the majority of the costs associated in generating recombinant proteins in plants is attributed to the purification process, designing efficient methodologies for reducing the number of purification steps required will greatly reduce total production costs (Wilken and Nikolov 2012). Cost-effective ways of setting up containment facilities or reducing maintenance costs can also bring down costs associated with the generation of recombinant proteins (Shanmugaraj and Ramalingam 2014; Paul 2015). Funds generated by government agencies can also help with the expenses associated with clinical trials, expediting the process of licensing recombinant pharmaceutical proteins (Howard and Hood 2014). Next-generation sequencing technologies may also facilitate the rapid development of transgenics, contributing to the enhanced production of recombinant proteins from plants (Egan et al. 2012). The production system for plant-made pharmaceuticals (PMP) must adhere to strict regulatory requirements laid down by Food and drug administration, USA and European agency for the evaluation of medicinal products and other agencies. These requirements for PMP vary between USA and EU (2008/27/EC; 1829/2003/EC) and were well reviewed by Sparrow et al. (2013). It would be good to have common regulatory framework on a global platform so as to eliminate trade barriers. Areas with scope for further improvement include improved suspension cultures and engineering of media by substituting media supplements on a trial and error basis to achieve greater productivity.

In conclusion, the advancement of technologies in this field can serve as an invaluable resource for the generation of recombinant pharmaceutical proteins, which will have a positive impact on human lives. Plant-made pharmaceuticals has shown great promise for therapeutic purposes with some of them tested and many others under investigation. Pharmaceutical products produced using plant-based expression systems can be easily expanded with lower infrastructure and production costs and are potentially free from animal pathogens. While the extraction and purification costs incurred for both plant and cell culture systems are similar, plant-made pharmaceuticals can be easily scaled up to produce tonnes of quantities with minimum production costs. Strict agronomic, licensing and manufacturing regulations need to be framed and are to be adhered for ensuring safety and efficacy of the therapeutic proteins.

Author contribution statement

VRD, GP and RRN designed the work. VRD wrote the manuscript. SK, PNR and SRKRS critically reviewed the drafted proposal and gave critical comments. VRD, PNR, SKN took an active part in revising the manuscript. All the authors approved the final version of the manuscript prior to submission.


The authors thank Vignan’s University, India for its in-house administrative and financial resources to execute this work. This work was supported in part by a Grant (ECR/2016/000304) from Science and Engineering Research Board, New Delhi, India under the young scientist scheme. We also thank the anonymous reviewers for their comments which helped us in improving the quality of manuscript. We thank Dr. David McMurray, Regents Professor, Texas A & M University System Health Science Center, College Station, USA for providing valuable comments.

Copyright information

© Franciszek Górski Institute of Plant Physiology, Polish Academy of Sciences, Kraków 2016

Authors and Affiliations

  • Vijaya R. Dirisala
    • 1
  • Rahul R. Nair
    • 4
  • Krupanidhi Srirama
    • 1
  • Prakash Narayana Reddy
    • 1
  • K. R. S. Sambasiva Rao
    • 2
  • N. Satya Sampath Kumar
    • 1
  • Giridhar Parvatam
    • 3
  1. 1.Department of BiotechnologyVignan’s Foundation for Science, Technology and Research University (VFSTRU)GunturIndia
  2. 2.Department of BiotechnologyAcharya Nagarjuna UniversityGunturIndia
  3. 3.Plant Cell Biotechnology DepartmentCentral Food Technological Research Institute (CFTRI)MysoreIndia
  4. 4.Administrative officeAushmath BiosciencesCoimbatoreIndia

Personalised recommendations